Skip to content

E
ease
Commit 98ec66ee by Vik Paruchuri
Options

Add needed files back in

parent e3283669
Showing with 3507 additions and 0 deletions
.gitignore 0 → 100644
.idea/
__pycache__/
models/
*.pyc
*~
tests/
_build/
build/
dist/
machine_learning.egg-info/
*.egg
LICENSE.txt 0 → 100644
GNU AFFERO GENERAL PUBLIC LICENSE
Version 3, 19 November 2007
Copyright (C) 2007 Free Software Foundation, Inc. <http://fsf.org/>
Everyone is permitted to copy and distribute verbatim copies
of this license document, but changing it is not allowed.
Preamble
The GNU Affero General Public License is a free, copyleft license for
software and other kinds of works, specifically designed to ensure
cooperation with the community in the case of network server software.
The licenses for most software and other practical works are designed
to take away your freedom to share and change the works. By contrast,
our General Public Licenses are intended to guarantee your freedom to
share and change all versions of a program--to make sure it remains free
software for all its users.
When we speak of free software, we are referring to freedom, not
price. Our General Public Licenses are designed to make sure that you
have the freedom to distribute copies of free software (and charge for
them if you wish), that you receive source code or can get it if you
want it, that you can change the software or use pieces of it in new
free programs, and that you know you can do these things.
Developers that use our General Public Licenses protect your rights
with two steps: (1) assert copyright on the software, and (2) offer
you this License which gives you legal permission to copy, distribute
and/or modify the software.
A secondary benefit of defending all users' freedom is that
improvements made in alternate versions of the program, if they
receive widespread use, become available for other developers to
incorporate. Many developers of free software are heartened and
encouraged by the resulting cooperation. However, in the case of
software used on network servers, this result may fail to come about.
The GNU General Public License permits making a modified version and
letting the public access it on a server without ever releasing its
source code to the public.
The GNU Affero General Public License is designed specifically to
ensure that, in such cases, the modified source code becomes available
to the community. It requires the operator of a network server to
provide the source code of the modified version running there to the
users of that server. Therefore, public use of a modified version, on
a publicly accessible server, gives the public access to the source
code of the modified version.
An older license, called the Affero General Public License and
published by Affero, was designed to accomplish similar goals. This is
a different license, not a version of the Affero GPL, but Affero has
released a new version of the Affero GPL which permits relicensing under
this license.
The precise terms and conditions for copying, distribution and
modification follow.
TERMS AND CONDITIONS
0. Definitions.
"This License" refers to version 3 of the GNU Affero General Public License.
"Copyright" also means copyright-like laws that apply to other kinds of
works, such as semiconductor masks.
"The Program" refers to any copyrightable work licensed under this
License. Each licensee is addressed as "you". "Licensees" and
"recipients" may be individuals or organizations.
To "modify" a work means to copy from or adapt all or part of the work
in a fashion requiring copyright permission, other than the making of an
exact copy. The resulting work is called a "modified version" of the
earlier work or a work "based on" the earlier work.
A "covered work" means either the unmodified Program or a work based
on the Program.
To "propagate" a work means to do anything with it that, without
permission, would make you directly or secondarily liable for
infringement under applicable copyright law, except executing it on a
computer or modifying a private copy. Propagation includes copying,
distribution (with or without modification), making available to the
public, and in some countries other activities as well.
To "convey" a work means any kind of propagation that enables other
parties to make or receive copies. Mere interaction with a user through
a computer network, with no transfer of a copy, is not conveying.
An interactive user interface displays "Appropriate Legal Notices"
to the extent that it includes a convenient and prominently visible
feature that (1) displays an appropriate copyright notice, and (2)
tells the user that there is no warranty for the work (except to the
extent that warranties are provided), that licensees may convey the
work under this License, and how to view a copy of this License. If
the interface presents a list of user commands or options, such as a
menu, a prominent item in the list meets this criterion.
1. Source Code.
The "source code" for a work means the preferred form of the work
for making modifications to it. "Object code" means any non-source
form of a work.
A "Standard Interface" means an interface that either is an official
standard defined by a recognized standards body, or, in the case of
interfaces specified for a particular programming language, one that
is widely used among developers working in that language.
The "System Libraries" of an executable work include anything, other
than the work as a whole, that (a) is included in the normal form of
packaging a Major Component, but which is not part of that Major
Component, and (b) serves only to enable use of the work with that
Major Component, or to implement a Standard Interface for which an
implementation is available to the public in source code form. A
"Major Component", in this context, means a major essential component
(kernel, window system, and so on) of the specific operating system
(if any) on which the executable work runs, or a compiler used to
produce the work, or an object code interpreter used to run it.
The "Corresponding Source" for a work in object code form means all
the source code needed to generate, install, and (for an executable
work) run the object code and to modify the work, including scripts to
control those activities. However, it does not include the work's
System Libraries, or general-purpose tools or generally available free
programs which are used unmodified in performing those activities but
which are not part of the work. For example, Corresponding Source
includes interface definition files associated with source files for
the work, and the source code for shared libraries and dynamically
linked subprograms that the work is specifically designed to require,
such as by intimate data communication or control flow between those
subprograms and other parts of the work.
The Corresponding Source need not include anything that users
can regenerate automatically from other parts of the Corresponding
Source.
The Corresponding Source for a work in source code form is that
same work.
2. Basic Permissions.
All rights granted under this License are granted for the term of
copyright on the Program, and are irrevocable provided the stated
conditions are met. This License explicitly affirms your unlimited
permission to run the unmodified Program. The output from running a
covered work is covered by this License only if the output, given its
content, constitutes a covered work. This License acknowledges your
rights of fair use or other equivalent, as provided by copyright law.
You may make, run and propagate covered works that you do not
convey, without conditions so long as your license otherwise remains
in force. You may convey covered works to others for the sole purpose
of having them make modifications exclusively for you, or provide you
with facilities for running those works, provided that you comply with
the terms of this License in conveying all material for which you do
not control copyright. Those thus making or running the covered works
for you must do so exclusively on your behalf, under your direction
and control, on terms that prohibit them from making any copies of
your copyrighted material outside their relationship with you.
Conveying under any other circumstances is permitted solely under
the conditions stated below. Sublicensing is not allowed; section 10
makes it unnecessary.
3. Protecting Users' Legal Rights From Anti-Circumvention Law.
No covered work shall be deemed part of an effective technological
measure under any applicable law fulfilling obligations under article
11 of the WIPO copyright treaty adopted on 20 December 1996, or
similar laws prohibiting or restricting circumvention of such
measures.
When you convey a covered work, you waive any legal power to forbid
circumvention of technological measures to the extent such circumvention
is effected by exercising rights under this License with respect to
the covered work, and you disclaim any intention to limit operation or
modification of the work as a means of enforcing, against the work's
users, your or third parties' legal rights to forbid circumvention of
technological measures.
4. Conveying Verbatim Copies.
You may convey verbatim copies of the Program's source code as you
receive it, in any medium, provided that you conspicuously and
appropriately publish on each copy an appropriate copyright notice;
keep intact all notices stating that this License and any
non-permissive terms added in accord with section 7 apply to the code;
keep intact all notices of the absence of any warranty; and give all
recipients a copy of this License along with the Program.
You may charge any price or no price for each copy that you convey,
and you may offer support or warranty protection for a fee.
5. Conveying Modified Source Versions.
You may convey a work based on the Program, or the modifications to
produce it from the Program, in the form of source code under the
terms of section 4, provided that you also meet all of these conditions:
a) The work must carry prominent notices stating that you modified
it, and giving a relevant date.
b) The work must carry prominent notices stating that it is
released under this License and any conditions added under section
7. This requirement modifies the requirement in section 4 to
"keep intact all notices".
c) You must license the entire work, as a whole, under this
License to anyone who comes into possession of a copy. This
License will therefore apply, along with any applicable section 7
additional terms, to the whole of the work, and all its parts,
regardless of how they are packaged. This License gives no
permission to license the work in any other way, but it does not
invalidate such permission if you have separately received it.
d) If the work has interactive user interfaces, each must display
Appropriate Legal Notices; however, if the Program has interactive
interfaces that do not display Appropriate Legal Notices, your
work need not make them do so.
A compilation of a covered work with other separate and independent
works, which are not by their nature extensions of the covered work,
and which are not combined with it such as to form a larger program,
in or on a volume of a storage or distribution medium, is called an
"aggregate" if the compilation and its resulting copyright are not
used to limit the access or legal rights of the compilation's users
beyond what the individual works permit. Inclusion of a covered work
in an aggregate does not cause this License to apply to the other
parts of the aggregate.
6. Conveying Non-Source Forms.
You may convey a covered work in object code form under the terms
of sections 4 and 5, provided that you also convey the
machine-readable Corresponding Source under the terms of this License,
in one of these ways:
a) Convey the object code in, or embodied in, a physical product
(including a physical distribution medium), accompanied by the
Corresponding Source fixed on a durable physical medium
customarily used for software interchange.
b) Convey the object code in, or embodied in, a physical product
(including a physical distribution medium), accompanied by a
written offer, valid for at least three years and valid for as
long as you offer spare parts or customer support for that product
model, to give anyone who possesses the object code either (1) a
copy of the Corresponding Source for all the software in the
product that is covered by this License, on a durable physical
medium customarily used for software interchange, for a price no
more than your reasonable cost of physically performing this
conveying of source, or (2) access to copy the
Corresponding Source from a network server at no charge.
c) Convey individual copies of the object code with a copy of the
written offer to provide the Corresponding Source. This
alternative is allowed only occasionally and noncommercially, and
only if you received the object code with such an offer, in accord
with subsection 6b.
d) Convey the object code by offering access from a designated
place (gratis or for a charge), and offer equivalent access to the
Corresponding Source in the same way through the same place at no
further charge. You need not require recipients to copy the
Corresponding Source along with the object code. If the place to
copy the object code is a network server, the Corresponding Source
may be on a different server (operated by you or a third party)
that supports equivalent copying facilities, provided you maintain
clear directions next to the object code saying where to find the
Corresponding Source. Regardless of what server hosts the
Corresponding Source, you remain obligated to ensure that it is
available for as long as needed to satisfy these requirements.
e) Convey the object code using peer-to-peer transmission, provided
you inform other peers where the object code and Corresponding
Source of the work are being offered to the general public at no
charge under subsection 6d.
A separable portion of the object code, whose source code is excluded
from the Corresponding Source as a System Library, need not be
included in conveying the object code work.
A "User Product" is either (1) a "consumer product", which means any
tangible personal property which is normally used for personal, family,
or household purposes, or (2) anything designed or sold for incorporation
into a dwelling. In determining whether a product is a consumer product,
doubtful cases shall be resolved in favor of coverage. For a particular
product received by a particular user, "normally used" refers to a
typical or common use of that class of product, regardless of the status
of the particular user or of the way in which the particular user
actually uses, or expects or is expected to use, the product. A product
is a consumer product regardless of whether the product has substantial
commercial, industrial or non-consumer uses, unless such uses represent
the only significant mode of use of the product.
"Installation Information" for a User Product means any methods,
procedures, authorization keys, or other information required to install
and execute modified versions of a covered work in that User Product from
a modified version of its Corresponding Source. The information must
suffice to ensure that the continued functioning of the modified object
code is in no case prevented or interfered with solely because
modification has been made.
If you convey an object code work under this section in, or with, or
specifically for use in, a User Product, and the conveying occurs as
part of a transaction in which the right of possession and use of the
User Product is transferred to the recipient in perpetuity or for a
fixed term (regardless of how the transaction is characterized), the
Corresponding Source conveyed under this section must be accompanied
by the Installation Information. But this requirement does not apply
if neither you nor any third party retains the ability to install
modified object code on the User Product (for example, the work has
been installed in ROM).
The requirement to provide Installation Information does not include a
requirement to continue to provide support service, warranty, or updates
for a work that has been modified or installed by the recipient, or for
the User Product in which it has been modified or installed. Access to a
network may be denied when the modification itself materially and
adversely affects the operation of the network or violates the rules and
protocols for communication across the network.
Corresponding Source conveyed, and Installation Information provided,
in accord with this section must be in a format that is publicly
documented (and with an implementation available to the public in
source code form), and must require no special password or key for
unpacking, reading or copying.
7. Additional Terms.
"Additional permissions" are terms that supplement the terms of this
License by making exceptions from one or more of its conditions.
Additional permissions that are applicable to the entire Program shall
be treated as though they were included in this License, to the extent
that they are valid under applicable law. If additional permissions
apply only to part of the Program, that part may be used separately
under those permissions, but the entire Program remains governed by
this License without regard to the additional permissions.
When you convey a copy of a covered work, you may at your option
remove any additional permissions from that copy, or from any part of
it. (Additional permissions may be written to require their own
removal in certain cases when you modify the work.) You may place
additional permissions on material, added by you to a covered work,
for which you have or can give appropriate copyright permission.
Notwithstanding any other provision of this License, for material you
add to a covered work, you may (if authorized by the copyright holders of
that material) supplement the terms of this License with terms:
a) Disclaiming warranty or limiting liability differently from the
terms of sections 15 and 16 of this License; or
b) Requiring preservation of specified reasonable legal notices or
author attributions in that material or in the Appropriate Legal
Notices displayed by works containing it; or
c) Prohibiting misrepresentation of the origin of that material, or
requiring that modified versions of such material be marked in
reasonable ways as different from the original version; or
d) Limiting the use for publicity purposes of names of licensors or
authors of the material; or
e) Declining to grant rights under trademark law for use of some
trade names, trademarks, or service marks; or
f) Requiring indemnification of licensors and authors of that
material by anyone who conveys the material (or modified versions of
it) with contractual assumptions of liability to the recipient, for
any liability that these contractual assumptions directly impose on
those licensors and authors.
All other non-permissive additional terms are considered "further
restrictions" within the meaning of section 10. If the Program as you
received it, or any part of it, contains a notice stating that it is
governed by this License along with a term that is a further
restriction, you may remove that term. If a license document contains
a further restriction but permits relicensing or conveying under this
License, you may add to a covered work material governed by the terms
of that license document, provided that the further restriction does
not survive such relicensing or conveying.
If you add terms to a covered work in accord with this section, you
must place, in the relevant source files, a statement of the
additional terms that apply to those files, or a notice indicating
where to find the applicable terms.
Additional terms, permissive or non-permissive, may be stated in the
form of a separately written license, or stated as exceptions;
the above requirements apply either way.
8. Termination.
You may not propagate or modify a covered work except as expressly
provided under this License. Any attempt otherwise to propagate or
modify it is void, and will automatically terminate your rights under
this License (including any patent licenses granted under the third
paragraph of section 11).
However, if you cease all violation of this License, then your
license from a particular copyright holder is reinstated (a)
provisionally, unless and until the copyright holder explicitly and
finally terminates your license, and (b) permanently, if the copyright
holder fails to notify you of the violation by some reasonable means
prior to 60 days after the cessation.
Moreover, your license from a particular copyright holder is
reinstated permanently if the copyright holder notifies you of the
violation by some reasonable means, this is the first time you have
received notice of violation of this License (for any work) from that
copyright holder, and you cure the violation prior to 30 days after
your receipt of the notice.
Termination of your rights under this section does not terminate the
licenses of parties who have received copies or rights from you under
this License. If your rights have been terminated and not permanently
reinstated, you do not qualify to receive new licenses for the same
material under section 10.
9. Acceptance Not Required for Having Copies.
You are not required to accept this License in order to receive or
run a copy of the Program. Ancillary propagation of a covered work
occurring solely as a consequence of using peer-to-peer transmission
to receive a copy likewise does not require acceptance. However,
nothing other than this License grants you permission to propagate or
modify any covered work. These actions infringe copyright if you do
not accept this License. Therefore, by modifying or propagating a
covered work, you indicate your acceptance of this License to do so.
10. Automatic Licensing of Downstream Recipients.
Each time you convey a covered work, the recipient automatically
receives a license from the original licensors, to run, modify and
propagate that work, subject to this License. You are not responsible
for enforcing compliance by third parties with this License.
An "entity transaction" is a transaction transferring control of an
organization, or substantially all assets of one, or subdividing an
organization, or merging organizations. If propagation of a covered
work results from an entity transaction, each party to that
transaction who receives a copy of the work also receives whatever
licenses to the work the party's predecessor in interest had or could
give under the previous paragraph, plus a right to possession of the
Corresponding Source of the work from the predecessor in interest, if
the predecessor has it or can get it with reasonable efforts.
You may not impose any further restrictions on the exercise of the
rights granted or affirmed under this License. For example, you may
not impose a license fee, royalty, or other charge for exercise of
rights granted under this License, and you may not initiate litigation
(including a cross-claim or counterclaim in a lawsuit) alleging that
any patent claim is infringed by making, using, selling, offering for
sale, or importing the Program or any portion of it.
11. Patents.
A "contributor" is a copyright holder who authorizes use under this
License of the Program or a work on which the Program is based. The
work thus licensed is called the contributor's "contributor version".
A contributor's "essential patent claims" are all patent claims
owned or controlled by the contributor, whether already acquired or
hereafter acquired, that would be infringed by some manner, permitted
by this License, of making, using, or selling its contributor version,
but do not include claims that would be infringed only as a
consequence of further modification of the contributor version. For
purposes of this definition, "control" includes the right to grant
patent sublicenses in a manner consistent with the requirements of
this License.
Each contributor grants you a non-exclusive, worldwide, royalty-free
patent license under the contributor's essential patent claims, to
make, use, sell, offer for sale, import and otherwise run, modify and
propagate the contents of its contributor version.
In the following three paragraphs, a "patent license" is any express
agreement or commitment, however denominated, not to enforce a patent
(such as an express permission to practice a patent or covenant not to
sue for patent infringement). To "grant" such a patent license to a
party means to make such an agreement or commitment not to enforce a
patent against the party.
If you convey a covered work, knowingly relying on a patent license,
and the Corresponding Source of the work is not available for anyone
to copy, free of charge and under the terms of this License, through a
publicly available network server or other readily accessible means,
then you must either (1) cause the Corresponding Source to be so
available, or (2) arrange to deprive yourself of the benefit of the
patent license for this particular work, or (3) arrange, in a manner
consistent with the requirements of this License, to extend the patent
license to downstream recipients. "Knowingly relying" means you have
actual knowledge that, but for the patent license, your conveying the
covered work in a country, or your recipient's use of the covered work
in a country, would infringe one or more identifiable patents in that
country that you have reason to believe are valid.
If, pursuant to or in connection with a single transaction or
arrangement, you convey, or propagate by procuring conveyance of, a
covered work, and grant a patent license to some of the parties
receiving the covered work authorizing them to use, propagate, modify
or convey a specific copy of the covered work, then the patent license
you grant is automatically extended to all recipients of the covered
work and works based on it.
A patent license is "discriminatory" if it does not include within
the scope of its coverage, prohibits the exercise of, or is
conditioned on the non-exercise of one or more of the rights that are
specifically granted under this License. You may not convey a covered
work if you are a party to an arrangement with a third party that is
in the business of distributing software, under which you make payment
to the third party based on the extent of your activity of conveying
the work, and under which the third party grants, to any of the
parties who would receive the covered work from you, a discriminatory
patent license (a) in connection with copies of the covered work
conveyed by you (or copies made from those copies), or (b) primarily
for and in connection with specific products or compilations that
contain the covered work, unless you entered into that arrangement,
or that patent license was granted, prior to 28 March 2007.
Nothing in this License shall be construed as excluding or limiting
any implied license or other defenses to infringement that may
otherwise be available to you under applicable patent law.
12. No Surrender of Others' Freedom.
If conditions are imposed on you (whether by court order, agreement or
otherwise) that contradict the conditions of this License, they do not
excuse you from the conditions of this License. If you cannot convey a
covered work so as to satisfy simultaneously your obligations under this
License and any other pertinent obligations, then as a consequence you may
not convey it at all. For example, if you agree to terms that obligate you
to collect a royalty for further conveying from those to whom you convey
the Program, the only way you could satisfy both those terms and this
License would be to refrain entirely from conveying the Program.
13. Remote Network Interaction; Use with the GNU General Public License.
Notwithstanding any other provision of this License, if you modify the
Program, your modified version must prominently offer all users
interacting with it remotely through a computer network (if your version
supports such interaction) an opportunity to receive the Corresponding
Source of your version by providing access to the Corresponding Source
from a network server at no charge, through some standard or customary
means of facilitating copying of software. This Corresponding Source
shall include the Corresponding Source for any work covered by version 3
of the GNU General Public License that is incorporated pursuant to the
following paragraph.
Notwithstanding any other provision of this License, you have
permission to link or combine any covered work with a work licensed
under version 3 of the GNU General Public License into a single
combined work, and to convey the resulting work. The terms of this
License will continue to apply to the part which is the covered work,
but the work with which it is combined will remain governed by version
3 of the GNU General Public License.
14. Revised Versions of this License.
The Free Software Foundation may publish revised and/or new versions of
the GNU Affero General Public License from time to time. Such new versions
will be similar in spirit to the present version, but may differ in detail to
address new problems or concerns.
Each version is given a distinguishing version number. If the
Program specifies that a certain numbered version of the GNU Affero General
Public License "or any later version" applies to it, you have the
option of following the terms and conditions either of that numbered
version or of any later version published by the Free Software
Foundation. If the Program does not specify a version number of the
GNU Affero General Public License, you may choose any version ever published
by the Free Software Foundation.
If the Program specifies that a proxy can decide which future
versions of the GNU Affero General Public License can be used, that proxy's
public statement of acceptance of a version permanently authorizes you
to choose that version for the Program.
Later license versions may give you additional or different
permissions. However, no additional obligations are imposed on any
author or copyright holder as a result of your choosing to follow a
later version.
15. Disclaimer of Warranty.
THERE IS NO WARRANTY FOR THE PROGRAM, TO THE EXTENT PERMITTED BY
APPLICABLE LAW. EXCEPT WHEN OTHERWISE STATED IN WRITING THE COPYRIGHT
HOLDERS AND/OR OTHER PARTIES PROVIDE THE PROGRAM "AS IS" WITHOUT WARRANTY
OF ANY KIND, EITHER EXPRESSED OR IMPLIED, INCLUDING, BUT NOT LIMITED TO,
THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
PURPOSE. THE ENTIRE RISK AS TO THE QUALITY AND PERFORMANCE OF THE PROGRAM
IS WITH YOU. SHOULD THE PROGRAM PROVE DEFECTIVE, YOU ASSUME THE COST OF
ALL NECESSARY SERVICING, REPAIR OR CORRECTION.
16. Limitation of Liability.
IN NO EVENT UNLESS REQUIRED BY APPLICABLE LAW OR AGREED TO IN WRITING
WILL ANY COPYRIGHT HOLDER, OR ANY OTHER PARTY WHO MODIFIES AND/OR CONVEYS
THE PROGRAM AS PERMITTED ABOVE, BE LIABLE TO YOU FOR DAMAGES, INCLUDING ANY
GENERAL, SPECIAL, INCIDENTAL OR CONSEQUENTIAL DAMAGES ARISING OUT OF THE
USE OR INABILITY TO USE THE PROGRAM (INCLUDING BUT NOT LIMITED TO LOSS OF
DATA OR DATA BEING RENDERED INACCURATE OR LOSSES SUSTAINED BY YOU OR THIRD
PARTIES OR A FAILURE OF THE PROGRAM TO OPERATE WITH ANY OTHER PROGRAMS),
EVEN IF SUCH HOLDER OR OTHER PARTY HAS BEEN ADVISED OF THE POSSIBILITY OF
SUCH DAMAGES.
17. Interpretation of Sections 15 and 16.
If the disclaimer of warranty and limitation of liability provided
above cannot be given local legal effect according to their terms,
reviewing courts shall apply local law that most closely approximates
an absolute waiver of all civil liability in connection with the
Program, unless a warranty or assumption of liability accompanies a
copy of the Program in return for a fee.
END OF TERMS AND CONDITIONS
How to Apply These Terms to Your New Programs
If you develop a new program, and you want it to be of the greatest
possible use to the public, the best way to achieve this is to make it
free software which everyone can redistribute and change under these terms.
To do so, attach the following notices to the program. It is safest
to attach them to the start of each source file to most effectively
state the exclusion of warranty; and each file should have at least
the "copyright" line and a pointer to where the full notice is found.
<one line to give the program's name and a brief idea of what it does.>
Copyright (C) <year> <name of author>
This program is free software: you can redistribute it and/or modify
it under the terms of the GNU Affero General Public License as published by
the Free Software Foundation, either version 3 of the License, or
(at your option) any later version.
This program is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
GNU Affero General Public License for more details.
You should have received a copy of the GNU Affero General Public License
along with this program. If not, see <http://www.gnu.org/licenses/>.
Also add information on how to contact you by electronic and paper mail.
If your software can interact with users remotely through a computer
network, you should also make sure that it provides a way for users to
get its source. For example, if your program is a web application, its
interface could display a "Source" link that leads users to an archive
of the code. There are many ways you could offer source, and different
solutions will be better for different programs; see section 13 for the
specific requirements.
You should also get your employer (if you work as a programmer) or school,
if any, to sign a "copyright disclaimer" for the program, if necessary.
For more information on this, and how to apply and follow the GNU AGPL, see
<http://www.gnu.org/licenses/>.
Readme.md 0 → 100644
ML
====================
Overview
---------------------
This is a repo with functions that can score arbitrary free text and numeric predictors.
This is licensed under the AGPL, please see LICENSE.txt for details.
The goal here is to provide a high-performance, scalable solution that can predict targets from arbitrary values.
Note that this is a library. You will need to implement your own code to make it runnable. The ml-service-api repo in
the edX github organization is an API wrapper for this code. See http://github.com/edx/ml-service-api for more information.
How to Contribute
-----------------------
Contributions are very welcome. The easiest way is to fork this repo, and then make a pull request from your fork.
The current backlog is in the issues section. Please feel free to open new issues or work on existing ones.
Detailed Information
-------------------------
Please look in the docs folder for more detailed documentation. There is a README there that explains how to build
and view the docs.
\ No newline at end of file
apt-packages.txt 0 → 100644
python-pip
gfortran
libblas3gf
libblas-dev
liblapack3gf
liblapack-dev
libatlas-base-dev
libxml2-dev
libxslt1-dev
aspell
python
\ No newline at end of file
docs/Makefile 0 → 100644
# Makefile for Sphinx documentation
#
# You can set these variables from the command line.
SPHINXOPTS =
SPHINXBUILD = sphinx-build
PAPER =
BUILDDIR = _build
# Internal variables.
PAPEROPT_a4 = -D latex_paper_size=a4
PAPEROPT_letter = -D latex_paper_size=letter
ALLSPHINXOPTS = -d $(BUILDDIR)/doctrees $(PAPEROPT_$(PAPER)) $(SPHINXOPTS) .
# the i18n builder cannot share the environment and doctrees with the others
I18NSPHINXOPTS = $(PAPEROPT_$(PAPER)) $(SPHINXOPTS) .
.PHONY: help clean html dirhtml singlehtml pickle json htmlhelp qthelp devhelp epub latex latexpdf text man changes linkcheck doctest gettext
help:
@echo "Please use \`make <target>' where <target> is one of"
@echo " html to make standalone HTML files"
@echo " dirhtml to make HTML files named index.html in directories"
@echo " singlehtml to make a single large HTML file"
@echo " pickle to make pickle files"
@echo " json to make JSON files"
@echo " htmlhelp to make HTML files and a HTML help project"
@echo " qthelp to make HTML files and a qthelp project"
@echo " devhelp to make HTML files and a Devhelp project"
@echo " epub to make an epub"
@echo " latex to make LaTeX files, you can set PAPER=a4 or PAPER=letter"
@echo " latexpdf to make LaTeX files and run them through pdflatex"
@echo " text to make text files"
@echo " man to make manual pages"
@echo " texinfo to make Texinfo files"
@echo " info to make Texinfo files and run them through makeinfo"
@echo " gettext to make PO message catalogs"
@echo " changes to make an overview of all changed/added/deprecated items"
@echo " linkcheck to check all external links for integrity"
@echo " doctest to run all doctests embedded in the documentation (if enabled)"
clean:
-rm -rf $(BUILDDIR)/*
html:
$(SPHINXBUILD) -b html $(ALLSPHINXOPTS) $(BUILDDIR)/html
@echo
@echo "Build finished. The HTML pages are in $(BUILDDIR)/html."
dirhtml:
$(SPHINXBUILD) -b dirhtml $(ALLSPHINXOPTS) $(BUILDDIR)/dirhtml
@echo
@echo "Build finished. The HTML pages are in $(BUILDDIR)/dirhtml."
singlehtml:
$(SPHINXBUILD) -b singlehtml $(ALLSPHINXOPTS) $(BUILDDIR)/singlehtml
@echo
@echo "Build finished. The HTML page is in $(BUILDDIR)/singlehtml."
pickle:
$(SPHINXBUILD) -b pickle $(ALLSPHINXOPTS) $(BUILDDIR)/pickle
@echo
@echo "Build finished; now you can process the pickle files."
json:
$(SPHINXBUILD) -b json $(ALLSPHINXOPTS) $(BUILDDIR)/json
@echo
@echo "Build finished; now you can process the JSON files."
htmlhelp:
$(SPHINXBUILD) -b htmlhelp $(ALLSPHINXOPTS) $(BUILDDIR)/htmlhelp
@echo
@echo "Build finished; now you can run HTML Help Workshop with the" \
".hhp project file in $(BUILDDIR)/htmlhelp."
qthelp:
$(SPHINXBUILD) -b qthelp $(ALLSPHINXOPTS) $(BUILDDIR)/qthelp
@echo
@echo "Build finished; now you can run "qcollectiongenerator" with the" \
".qhcp project file in $(BUILDDIR)/qthelp, like this:"
@echo "# qcollectiongenerator $(BUILDDIR)/qthelp/MLAPI.qhcp"
@echo "To view the help file:"
@echo "# assistant -collectionFile $(BUILDDIR)/qthelp/MLAPI.qhc"
devhelp:
$(SPHINXBUILD) -b devhelp $(ALLSPHINXOPTS) $(BUILDDIR)/devhelp
@echo
@echo "Build finished."
@echo "To view the help file:"
@echo "# mkdir -p $$HOME/.local/share/devhelp/MLAPI"
@echo "# ln -s $(BUILDDIR)/devhelp $$HOME/.local/share/devhelp/MLAPI"
@echo "# devhelp"
epub:
$(SPHINXBUILD) -b epub $(ALLSPHINXOPTS) $(BUILDDIR)/epub
@echo
@echo "Build finished. The epub file is in $(BUILDDIR)/epub."
latex:
$(SPHINXBUILD) -b latex $(ALLSPHINXOPTS) $(BUILDDIR)/latex
@echo
@echo "Build finished; the LaTeX files are in $(BUILDDIR)/latex."
@echo "Run \`make' in that directory to run these through (pdf)latex" \
"(use \`make latexpdf' here to do that automatically)."
latexpdf:
$(SPHINXBUILD) -b latex $(ALLSPHINXOPTS) $(BUILDDIR)/latex
@echo "Running LaTeX files through pdflatex..."
$(MAKE) -C $(BUILDDIR)/latex all-pdf
@echo "pdflatex finished; the PDF files are in $(BUILDDIR)/latex."
text:
$(SPHINXBUILD) -b text $(ALLSPHINXOPTS) $(BUILDDIR)/text
@echo
@echo "Build finished. The text files are in $(BUILDDIR)/text."
man:
$(SPHINXBUILD) -b man $(ALLSPHINXOPTS) $(BUILDDIR)/man
@echo
@echo "Build finished. The manual pages are in $(BUILDDIR)/man."
texinfo:
$(SPHINXBUILD) -b texinfo $(ALLSPHINXOPTS) $(BUILDDIR)/texinfo
@echo
@echo "Build finished. The Texinfo files are in $(BUILDDIR)/texinfo."
@echo "Run \`make' in that directory to run these through makeinfo" \
"(use \`make info' here to do that automatically)."
info:
$(SPHINXBUILD) -b texinfo $(ALLSPHINXOPTS) $(BUILDDIR)/texinfo
@echo "Running Texinfo files through makeinfo..."
make -C $(BUILDDIR)/texinfo info
@echo "makeinfo finished; the Info files are in $(BUILDDIR)/texinfo."
gettext:
$(SPHINXBUILD) -b gettext $(I18NSPHINXOPTS) $(BUILDDIR)/locale
@echo
@echo "Build finished. The message catalogs are in $(BUILDDIR)/locale."
changes:
$(SPHINXBUILD) -b changes $(ALLSPHINXOPTS) $(BUILDDIR)/changes
@echo
@echo "The overview file is in $(BUILDDIR)/changes."
linkcheck:
$(SPHINXBUILD) -b linkcheck $(ALLSPHINXOPTS) $(BUILDDIR)/linkcheck
@echo
@echo "Link check complete; look for any errors in the above output " \
"or in $(BUILDDIR)/linkcheck/output.txt."
doctest:
$(SPHINXBUILD) -b doctest $(ALLSPHINXOPTS) $(BUILDDIR)/doctest
@echo "Testing of doctests in the sources finished, look at the " \
"results in $(BUILDDIR)/doctest/output.txt."
docs/README 0 → 100644
This directory contains documentation that can be built into HTML using sphinx (http://sphinx.pocoo.org/).
Sphinx uses ReST (reStructuredText) as the source for its documentation files.
To create an HTML version of the docs:
* Install Sphinx (``pip install Sphinx``)
* In this directory, type ``make html`` (or ``make.bat html`` on
Windows) at a shell prompt.
The documentation in _build/html/index.html can then be viewed in a web browser.
\ No newline at end of file
docs/conf.py 0 → 100644
# -*- coding: utf-8 -*-
#
# ML API documentation build configuration file, created by
# sphinx-quickstart on Fri Mar 1 09:51:10 2013.
#
# This file is execfile()d with the current directory set to its containing dir.
#
# Note that not all possible configuration values are present in this
# autogenerated file.
#
# All configuration values have a default; values that are commented out
# serve to show the default.
import sys, os
# If extensions (or modules to document with autodoc) are in another directory,
# add these directories to sys.path here. If the directory is relative to the
# documentation root, use os.path.abspath to make it absolute, like shown here.
sys.path.append(os.path.abspath('.'))
sys.path.append(os.path.abspath('../'))
sys.path.append(os.path.abspath('../machine-learning'))
sys.path.append(os.path.abspath('../../'))
# -- General configuration -----------------------------------------------------
# If your documentation needs a minimal Sphinx version, state it here.
#needs_sphinx = '1.0'
# Add any Sphinx extension module names here, as strings. They can be extensions
# coming with Sphinx (named 'sphinx.ext.*') or your custom ones.
extensions = ['sphinx.ext.autodoc', 'sphinx.ext.intersphinx', 'sphinx.ext.todo', 'sphinx.ext.coverage', 'sphinx.ext.pngmath', 'sphinx.ext.mathjax', 'sphinx.ext.ifconfig', 'sphinx.ext.viewcode']
# Add any paths that contain templates here, relative to this directory.
templates_path = ['_templates']
# The suffix of source filenames.
source_suffix = '.rst'
# The encoding of source files.
#source_encoding = 'utf-8-sig'
# The master toctree document.
master_doc = 'index'
# General information about the project.
project = u'ML'
copyright = u'2013, edX'
# The version info for the project you're documenting, acts as replacement for
# |version| and |release|, also used in various other places throughout the
# built documents.
#
# The short X.Y version.
version = '.01'
# The full version, including alpha/beta/rc tags.
release = '.01'
# The language for content autogenerated by Sphinx. Refer to documentation
# for a list of supported languages.
#language = None
# There are two options for replacing |today|: either, you set today to some
# non-false value, then it is used:
#today = ''
# Else, today_fmt is used as the format for a strftime call.
#today_fmt = '%B %d, %Y'
# List of patterns, relative to source directory, that match files and
# directories to ignore when looking for source files.
exclude_patterns = ['_build']
# The reST default role (used for this markup: `text`) to use for all documents.
#default_role = None
# If true, '()' will be appended to :func: etc. cross-reference text.
#add_function_parentheses = True
# If true, the current module name will be prepended to all description
# unit titles (such as .. function::).
#add_module_names = True
# If true, sectionauthor and moduleauthor directives will be shown in the
# output. They are ignored by default.
#show_authors = False
# The name of the Pygments (syntax highlighting) style to use.
pygments_style = 'sphinx'
# A list of ignored prefixes for module index sorting.
#modindex_common_prefix = []
# -- Options for HTML output ---------------------------------------------------
# The theme to use for HTML and HTML Help pages. See the documentation for
# a list of builtin themes.
html_theme = 'default'
# Theme options are theme-specific and customize the look and feel of a theme
# further. For a list of options available for each theme, see the
# documentation.
#html_theme_options = {}
# Add any paths that contain custom themes here, relative to this directory.
#html_theme_path = []
# The name for this set of Sphinx documents. If None, it defaults to
# "<project> v<release> documentation".
#html_title = None
# A shorter title for the navigation bar. Default is the same as html_title.
#html_short_title = None
# The name of an image file (relative to this directory) to place at the top
# of the sidebar.
#html_logo = None
# The name of an image file (within the static path) to use as favicon of the
# docs. This file should be a Windows icon file (.ico) being 16x16 or 32x32
# pixels large.
#html_favicon = None
# Add any paths that contain custom static files (such as style sheets) here,
# relative to this directory. They are copied after the builtin static files,
# so a file named "default.css" will overwrite the builtin "default.css".
html_static_path = ['_static']
# If not '', a 'Last updated on:' timestamp is inserted at every page bottom,
# using the given strftime format.
#html_last_updated_fmt = '%b %d, %Y'
# If true, SmartyPants will be used to convert quotes and dashes to
# typographically correct entities.
#html_use_smartypants = True
# Custom sidebar templates, maps document names to template names.
#html_sidebars = {}
# Additional templates that should be rendered to pages, maps page names to
# template names.
#html_additional_pages = {}
# If false, no module index is generated.
#html_domain_indices = True
# If false, no index is generated.
#html_use_index = True
# If true, the index is split into individual pages for each letter.
#html_split_index = False
# If true, links to the reST sources are added to the pages.
#html_show_sourcelink = True
# If true, "Created using Sphinx" is shown in the HTML footer. Default is True.
#html_show_sphinx = True
# If true, "(C) Copyright ..." is shown in the HTML footer. Default is True.
#html_show_copyright = True
# If true, an OpenSearch description file will be output, and all pages will
# contain a <link> tag referring to it. The value of this option must be the
# base URL from which the finished HTML is served.
#html_use_opensearch = ''
# This is the file name suffix for HTML files (e.g. ".xhtml").
#html_file_suffix = None
# Output file base name for HTML help builder.
htmlhelp_basename = 'MLdoc'
# -- Options for LaTeX output --------------------------------------------------
latex_elements = {
# The paper size ('letterpaper' or 'a4paper').
#'papersize': 'letterpaper',
# The font size ('10pt', '11pt' or '12pt').
#'pointsize': '10pt',
# Additional stuff for the LaTeX preamble.
#'preamble': '',
}
# Grouping the document tree into LaTeX files. List of tuples
# (source start file, target name, title, author, documentclass [howto/manual]).
latex_documents = [
('index', 'ML.tex', u'ML API Documentation',
u'edX', 'manual'),
]
# The name of an image file (relative to this directory) to place at the top of
# the title page.
#latex_logo = None
# For "manual" documents, if this is true, then toplevel headings are parts,
# not chapters.
#latex_use_parts = False
# If true, show page references after internal links.
#latex_show_pagerefs = False
# If true, show URL addresses after external links.
#latex_show_urls = False
# Documents to append as an appendix to all manuals.
#latex_appendices = []
# If false, no module index is generated.
#latex_domain_indices = True
# -- Options for manual page output --------------------------------------------
# One entry per manual page. List of tuples
# (source start file, name, description, authors, manual section).
man_pages = [
('index', 'ml', u'ML Documentation',
[u'edX'], 1)
]
# If true, show URL addresses after external links.
#man_show_urls = False
# -- Options for Texinfo output ------------------------------------------------
# Grouping the document tree into Texinfo files. List of tuples
# (source start file, target name, title, author,
# dir menu entry, description, category)
texinfo_documents = [
('index', 'ML', u'ML Documentation',
u'edX', 'ML', 'One line description of project.',
'Miscellaneous'),
]
# Documents to append as an appendix to all manuals.
#texinfo_appendices = []
# If false, no module index is generated.
#texinfo_domain_indices = True
# How to display URL addresses: 'footnote', 'no', or 'inline'.
#texinfo_show_urls = 'footnote'
# Example configuration for intersphinx: refer to the Python standard library.
intersphinx_mapping = {'http://docs.python.org/': None}
docs/index.rst 0 → 100644
.. ML documentation master file, created by
sphinx-quickstart on Fri Mar 1 09:51:10 2013.
You can adapt this file completely to your liking, but it should at least
contain the root `toctree` directive.
ML Documentation
==================================
Overview
---------------------------------
.. toctree::
:maxdepth: 1
overview/description
overview/goals
overview/contributing
Installation and Usage
---------------------------------
.. toctree::
:maxdepth: 1
installation/installation_overview
installation/usage
Module Documentation
---------------------------------
.. toctree::
:maxdepth: 1
project
Indices and tables
==================
* :ref:`genindex`
* :ref:`modindex`
* :ref:`search`
docs/installation/installation_overview.rst 0 → 100644
===============================================
Installation Overview
===============================================
Notes on how to install:
1. cd DIRECTORY_YOU_INSTALLED_TO. Make sure that you install to the folder machine-learning!
2. sudo apt-get update
3. sudo apt-get upgrade gcc
4. sudo xargs -a apt-packages.txt apt-get install
5. Activate your virtual env (if you have one)
6. pip install -r pre-requirements.txt
7. pip install -r requirements.txt
8. python -m nltk.downloader maxent_treebank_pos_tagger wordnet
9. sudo mv /path/to/nltk_data /usr/share
docs/installation/usage.rst 0 → 100644
==================================
Usage
==================================
This repo offers the ability to create models and to grade new text. There are additional lower-level functions, as well.
Essay Grading
-------------------------------------
Essay grading can be done via the "grade" function in grade.py and the "create" function in create.py. Call the create function, and pass in the appropriate data (see documentation there), in order to obtain a created model. That model can then be used in conjunction with the "grade" function to get scores for new text.
Arbitrary sets of predictors and text scoring
----------------------------------------------------
This repo can also be used to compute scores for arbitrary sets of numeric and textual predictors. For example, you could predict whether the stock market will rise or fall tomorrow by passing in a set of article headlines, article text, and publication times. Use the functions "create_generic" in create.py and "grade_generic" in grade.py to do this.
docs/make.bat 0 → 100644
@ECHO OFF
REM Command file for Sphinx documentation
if "%SPHINXBUILD%" == "" (
set SPHINXBUILD=sphinx-build
)
set BUILDDIR=_build
set ALLSPHINXOPTS=-d %BUILDDIR%/doctrees %SPHINXOPTS% .
set I18NSPHINXOPTS=%SPHINXOPTS% .
if NOT "%PAPER%" == "" (
set ALLSPHINXOPTS=-D latex_paper_size=%PAPER% %ALLSPHINXOPTS%
set I18NSPHINXOPTS=-D latex_paper_size=%PAPER% %I18NSPHINXOPTS%
)
if "%1" == "" goto help
if "%1" == "help" (
:help
echo.Please use `make ^<target^>` where ^<target^> is one of
echo. html to make standalone HTML files
echo. dirhtml to make HTML files named index.html in directories
echo. singlehtml to make a single large HTML file
echo. pickle to make pickle files
echo. json to make JSON files
echo. htmlhelp to make HTML files and a HTML help project
echo. qthelp to make HTML files and a qthelp project
echo. devhelp to make HTML files and a Devhelp project
echo. epub to make an epub
echo. latex to make LaTeX files, you can set PAPER=a4 or PAPER=letter
echo. text to make text files
echo. man to make manual pages
echo. texinfo to make Texinfo files
echo. gettext to make PO message catalogs
echo. changes to make an overview over all changed/added/deprecated items
echo. linkcheck to check all external links for integrity
echo. doctest to run all doctests embedded in the documentation if enabled
goto end
)
if "%1" == "clean" (
for /d %%i in (%BUILDDIR%\*) do rmdir /q /s %%i
del /q /s %BUILDDIR%\*
goto end
)
if "%1" == "html" (
%SPHINXBUILD% -b html %ALLSPHINXOPTS% %BUILDDIR%/html
if errorlevel 1 exit /b 1
echo.
echo.Build finished. The HTML pages are in %BUILDDIR%/html.
goto end
)
if "%1" == "dirhtml" (
%SPHINXBUILD% -b dirhtml %ALLSPHINXOPTS% %BUILDDIR%/dirhtml
if errorlevel 1 exit /b 1
echo.
echo.Build finished. The HTML pages are in %BUILDDIR%/dirhtml.
goto end
)
if "%1" == "singlehtml" (
%SPHINXBUILD% -b singlehtml %ALLSPHINXOPTS% %BUILDDIR%/singlehtml
if errorlevel 1 exit /b 1
echo.
echo.Build finished. The HTML pages are in %BUILDDIR%/singlehtml.
goto end
)
if "%1" == "pickle" (
%SPHINXBUILD% -b pickle %ALLSPHINXOPTS% %BUILDDIR%/pickle
if errorlevel 1 exit /b 1
echo.
echo.Build finished; now you can process the pickle files.
goto end
)
if "%1" == "json" (
%SPHINXBUILD% -b json %ALLSPHINXOPTS% %BUILDDIR%/json
if errorlevel 1 exit /b 1
echo.
echo.Build finished; now you can process the JSON files.
goto end
)
if "%1" == "htmlhelp" (
%SPHINXBUILD% -b htmlhelp %ALLSPHINXOPTS% %BUILDDIR%/htmlhelp
if errorlevel 1 exit /b 1
echo.
echo.Build finished; now you can run HTML Help Workshop with the ^
.hhp project file in %BUILDDIR%/htmlhelp.
goto end
)
if "%1" == "qthelp" (
%SPHINXBUILD% -b qthelp %ALLSPHINXOPTS% %BUILDDIR%/qthelp
if errorlevel 1 exit /b 1
echo.
echo.Build finished; now you can run "qcollectiongenerator" with the ^
.qhcp project file in %BUILDDIR%/qthelp, like this:
echo.^> qcollectiongenerator %BUILDDIR%\qthelp\MLAPI.qhcp
echo.To view the help file:
echo.^> assistant -collectionFile %BUILDDIR%\qthelp\MLAPI.ghc
goto end
)
if "%1" == "devhelp" (
%SPHINXBUILD% -b devhelp %ALLSPHINXOPTS% %BUILDDIR%/devhelp
if errorlevel 1 exit /b 1
echo.
echo.Build finished.
goto end
)
if "%1" == "epub" (
%SPHINXBUILD% -b epub %ALLSPHINXOPTS% %BUILDDIR%/epub
if errorlevel 1 exit /b 1
echo.
echo.Build finished. The epub file is in %BUILDDIR%/epub.
goto end
)
if "%1" == "latex" (
%SPHINXBUILD% -b latex %ALLSPHINXOPTS% %BUILDDIR%/latex
if errorlevel 1 exit /b 1
echo.
echo.Build finished; the LaTeX files are in %BUILDDIR%/latex.
goto end
)
if "%1" == "text" (
%SPHINXBUILD% -b text %ALLSPHINXOPTS% %BUILDDIR%/text
if errorlevel 1 exit /b 1
echo.
echo.Build finished. The text files are in %BUILDDIR%/text.
goto end
)
if "%1" == "man" (
%SPHINXBUILD% -b man %ALLSPHINXOPTS% %BUILDDIR%/man
if errorlevel 1 exit /b 1
echo.
echo.Build finished. The manual pages are in %BUILDDIR%/man.
goto end
)
if "%1" == "texinfo" (
%SPHINXBUILD% -b texinfo %ALLSPHINXOPTS% %BUILDDIR%/texinfo
if errorlevel 1 exit /b 1
echo.
echo.Build finished. The Texinfo files are in %BUILDDIR%/texinfo.
goto end
)
if "%1" == "gettext" (
%SPHINXBUILD% -b gettext %I18NSPHINXOPTS% %BUILDDIR%/locale
if errorlevel 1 exit /b 1
echo.
echo.Build finished. The message catalogs are in %BUILDDIR%/locale.
goto end
)
if "%1" == "changes" (
%SPHINXBUILD% -b changes %ALLSPHINXOPTS% %BUILDDIR%/changes
if errorlevel 1 exit /b 1
echo.
echo.The overview file is in %BUILDDIR%/changes.
goto end
)
if "%1" == "linkcheck" (
%SPHINXBUILD% -b linkcheck %ALLSPHINXOPTS% %BUILDDIR%/linkcheck
if errorlevel 1 exit /b 1
echo.
echo.Link check complete; look for any errors in the above output ^
or in %BUILDDIR%/linkcheck/output.txt.
goto end
)
if "%1" == "doctest" (
%SPHINXBUILD% -b doctest %ALLSPHINXOPTS% %BUILDDIR%/doctest
if errorlevel 1 exit /b 1
echo.
echo.Testing of doctests in the sources finished, look at the ^
results in %BUILDDIR%/doctest/output.txt.
goto end
)
:end
docs/overview/contributing.rst 0 → 100644
===============================================
Contributing
===============================================
We welcome contributions! In order to contribute:
1. Find something that would be useful to contribute, or check the github issues tracker to see if there is anything to work on.
2. Fork the repository and add in your new feature.
3. Put in a pull request to merge your feature into this repo.
Please feel free to contact vik@edx.org if you have any questions about contributing.
\ No newline at end of file
docs/overview/description.rst 0 → 100644
===============================================
Description
===============================================
The ML repo allows anyone to use machine-learning based automated classification. This automated classification can work on both free text (essays, content, etc), and on numeric values.
Let's say that you have 10000 user reviews for 15 books (ie "I loved this!", "I didn't like it.", and so on). What you really want to do is use the user reviews to get an aggregate score for each book that indicates how well-received it is. But, in your haste to collect the data, you forgot to get scores from the users. In this case, the text of the user reviews is your predictor, and the score that you want to collect from each user for each book is the target variable.
So, how do you turn the text into numbers? One very straightforward way is to just label each of the reviews by hand on a scale from 0 (the user didn't like it at all) to 5 (they really loved it). But, somewhere around review 200 you are going to start to get very sick of the whole process. A less labor intensive way is to use automated classification.
If you choose to use automated classification for this task, you will score some reasonable subset of the reviews (if you score more, the classification will be more accurate, but 200 should be fine as a baseline). Once you have your subset, which can also be called a "training" set, you will be able to "train" a machine learning model that learns how to map your scores to the text of the reviews. It will then be able to automatically score the rest of the 9800 reviews. Let's say you also want to take the user's activity level into account in order to weight the score. You can add in a numeric predictor in addition to your existing text predictor (the review text itself) in order to predict the target variable (score).
This repo gives you a nice, clean way to do that via convenience functions grade, grade_generic, create, and create_generic.
docs/overview/goals.rst 0 → 100644
===============================================
edX Goals
===============================================
Why is this open source? The algorithms and code in this repository offer a fast and flexible way to predict almost anything that you have data for. The goals in open sourcing this are to encourage widespread adoption and to get useful contributions that improve and expand on this code.
\ No newline at end of file
docs/project.rst 0 → 100644
##############
ML
##############
ML Model Creation
-------------------------------------
.. automodule:: create
:members:
ML Grading
--------------------------------------
.. automodule:: grade
:members:
Essay Set
--------------------------------------
.. automodule:: essay_set
:members:
Feature Extractor
--------------------------------------
.. automodule:: feature_extractor
:members:
Predictor Set
--------------------------------------
.. automodule:: predictor_set
:members:
Predictor Extractor
--------------------------------------
.. automodule:: predictor_extractor
:members:
Utility Functions
--------------------------------------
.. automodule:: util_functions
:members:
full_requirements.txt 0 → 100644
-r pre-requirements.txt
-r requirements.txt
\ No newline at end of file
install_system_req.sh 0 → 100644
#!/usr/bin/env bash
# posix compliant sanity check
if [ -z $BASH ] || [ $BASH = "/bin/sh" ]; then
echo "Please use the bash interpreter to run this script"
exit 1
fi
error() {
printf '\E[31m'; echo "$@"; printf '\E[0m'
}
output() {
printf '\E[36m'; echo "$@"; printf '\E[0m'
}
### START
DIR="$( cd "$( dirname "${BASH_SOURCE[0]}" )" && pwd )"
BREW_FILE=$DIR/"brew-formulas.txt"
APT_PKGS_FILE=$DIR/"apt-packages.txt"
case `uname -s` in
[Ll]inux)
command -v lsb_release &>/dev/null || {
error "Please install lsb-release."
exit 1
}
distro=`lsb_release -cs`
case $distro in
maya|lisa|natty|oneiric|precise|quantal)
output "Installing Ubuntu requirements"
# DEBIAN_FRONTEND=noninteractive is required for silent mysql-server installation
export DEBIAN_FRONTEND=noninteractive
# install packages listed in APT_PKGS_FILE
cat $APT_PKGS_FILE | xargs sudo apt-get -y install
;;
*)
error "Unsupported distribution - $distro"
exit 1
;;
esac
;;
Darwin)
if [[ ! -w /usr/local ]]; then
cat<<EO
You need to be able to write to /usr/local for
the installation of brew and brew packages.
Either make sure the group you are in (most likely 'staff')
can write to that directory or simply execute the following
and re-run the script:
$ sudo chown -R $USER /usr/local
EO
exit 1
fi
output "Installing OSX requirements"
if [[ ! -r $BREW_FILE ]]; then
error "$BREW_FILE does not exist, needed to install brew"
exit 1
fi
# brew errors if the package is already installed
for pkg in $(cat $BREW_FILE); do
grep $pkg <(brew list) &>/dev/null || {
output "Installing $pkg"
brew install $pkg
}
done
# paths where brew likes to install python scripts
PATH=/usr/local/share/python:/usr/local/bin:$PATH
command -v pip &>/dev/null || {
output "Installing pip"
easy_install pip
}
if ! grep -Eq ^1.7 <(virtualenv --version 2>/dev/null); then
output "Installing virtualenv >1.7"
pip install 'virtualenv>1.7' virtualenvwrapper
fi
command -v coffee &>/dev/null || {
output "Installing coffee script"
curl --insecure https://npmjs.org/install.sh | sh
npm install -g coffee-script
}
;;
*)
error "Unsupported platform"
exit 1
;;
esac
machine_learning/create.py 0 → 100644
"""
Functions that create a machine learning model from training data
"""
import os
import sys
import logging
from statsd import statsd
import numpy
#Define base path and add to sys path
base_path = os.path.dirname(__file__)
sys.path.append(base_path)
one_up_path = os.path.abspath(os.path.join(os.path.dirname(__file__), '..//'))
sys.path.append(one_up_path)
#Import modules that are dependent on the base path
import model_creator
import util_functions
import predictor_set
import predictor_extractor
#Make a log
log = logging.getLogger(__name__)
@statsd.timed('open_ended_assessment.machine_learning.creator.time')
def create(text,score,prompt_string):
"""
Creates a machine learning model from input text, associated scores, a prompt, and a path to the model
TODO: Remove model path argument, it is needed for now to support legacy code
text - A list of strings containing the text of the essays
score - a list of integers containing score values
prompt_string - the common prompt for the set of essays
"""
#Initialize a results dictionary to return
results = {'errors': [],'success' : False, 'cv_kappa' : 0, 'cv_mean_absolute_error': 0,
'feature_ext' : "", 'classifier' : "", 'algorithm' : util_functions.AlgorithmTypes.classification,
'score' : score, 'text' : text, 'prompt' : prompt_string}
if len(text)!=len(score):
msg = "Target and text lists must be same length."
results['errors'].append(msg)
log.exception(msg)
return results
#Decide what algorithm to use (regression or classification)
try:
#Count the number of unique score points in the score list
if len(util_functions.f7(list(score)))>5:
type = util_functions.AlgorithmTypes.regression
else:
type = util_functions.AlgorithmTypes.classification
except:
type = util_functions.AlgorithmTypes.regression
try:
#Create an essay set object that encapsulates all the essays and alternate representations (tokens, etc)
e_set = model_creator.create_essay_set(text, score, prompt_string)
except:
msg = "essay set creation failed."
results['errors'].append(msg)
log.exception(msg)
try:
#Gets features from the essay set and computes error
feature_ext, classifier, cv_error_results = model_creator.extract_features_and_generate_model(e_set, type=type)
results['cv_kappa']=cv_error_results['kappa']
results['cv_mean_absolute_error']=cv_error_results['mae']
results['feature_ext']=feature_ext
results['classifier']=classifier
results['algorithm'] = type
results['success']=True
except:
msg = "feature extraction and model creation failed."
results['errors'].append(msg)
log.exception(msg)
#Count number of successful/unsuccessful creations
statsd.increment("open_ended_assessment.machine_learning.creator_count",
tags=["success:{0}".format(results['success'])])
return results
def create_generic(numeric_values, textual_values, target, algorithm = util_functions.AlgorithmTypes.regression):
"""
Creates a model from a generic list numeric values and text values
numeric_values - A list of lists that are the predictors
textual_values - A list of lists that are the predictors
(each item in textual_values corresponds to the similarly indexed counterpart in numeric_values)
target - The variable that we are trying to predict. A list of integers.
algorithm - the type of algorithm that will be used
"""
#Initialize a result dictionary to return.
results = {'errors': [],'success' : False, 'cv_kappa' : 0, 'cv_mean_absolute_error': 0,
'feature_ext' : "", 'classifier' : "", 'algorithm' : algorithm}
if len(numeric_values)!=len(textual_values) or len(numeric_values)!=len(target):
msg = "Target, numeric features, and text features must all be the same length."
results['errors'].append(msg)
log.exception(msg)
return results
try:
#Initialize a predictor set object that encapsulates all of the text and numeric predictors
pset = predictor_set.PredictorSet(type="train")
for i in xrange(0, len(numeric_values)):
pset.add_row(numeric_values[i], textual_values[i], target[i])
except:
msg = "predictor set creation failed."
results['errors'].append(msg)
log.exception(msg)
try:
#Extract all features and then train a classifier with the features
feature_ext, classifier, cv_error_results = model_creator.extract_features_and_generate_model_predictors(pset, algorithm)
results['cv_kappa']=cv_error_results['kappa']
results['cv_mean_absolute_error']=cv_error_results['mae']
results['feature_ext']=feature_ext
results['classifier']=classifier
results['success']=True
except:
msg = "feature extraction and model creation failed."
results['errors'].append(msg)
log.exception(msg)
#Count number of successful/unsuccessful creations
statsd.increment("open_ended_assessment.machine_learning.creator_count",
tags=["success:{0}".format(results['success'])])
return results
\ No newline at end of file
machine_learning/data/essaycorpus.txt 0 → 100644
This source diff could not be displayed because it is too large. You can view the blob instead.
machine_learning/data/good_pos_ngrams.p 0 → 100644
This source diff could not be displayed because it is too large. You can view the blob instead.
machine_learning/essay_set.py 0 → 100644
"""
Defines an essay set object, which encapsulates essays from training and test sets.
Performs spell and grammar checking, tokenization, and stemming.
"""
import numpy
import nltk
import sys
import random
import os
import logging
base_path = os.path.dirname(__file__)
sys.path.append(base_path)
import util_functions
if not base_path.endswith("/"):
base_path=base_path+"/"
log=logging.getLogger(__name__)
MAXIMUM_ESSAY_LENGTH=20000
class EssaySet(object):
def __init__(self, type="train"):
"""
Initialize variables and check essay set type
"""
if(type != "train" and type != "test"):
type = "train"
self._type = type
self._score=[]
self._text=[]
self._id=[]
self._clean_text=[]
self._tokens=[]
self._pos=[]
self._clean_stem_text=[]
self._generated = []
self._prompt = ""
self._spelling_errors=[]
self._markup_text=[]
def add_essay(self, essay_text, essay_score, essay_generated=0):
"""
Add new (essay_text,essay_score) pair to the essay set.
essay_text must be a string.
essay_score must be an int.
essay_generated should not be changed by the user.
Returns a confirmation that essay was added.
"""
# Get maximum current essay id, or set to 0 if this is the first essay added
if(len(self._id) > 0):
max_id = max(self._id)
else:
max_id = 0
# Verify that essay_score is an int, essay_text is a string, and essay_generated equals 0 or 1
try:
essay_text=essay_text.encode('ascii', 'ignore')
if len(essay_text)<5:
essay_text="Invalid essay."
except:
log.exception("Could not parse essay into ascii.")
try:
#Try conversion of types
essay_score=int(essay_score)
essay_text=str(essay_text)
except:
#Nothing needed here, will return error in any case.
log.exception("Invalid type for essay score : {0} or essay text : {1}".format(type(essay_score),type(essay_text)))
if isinstance(essay_score,int) and isinstance(essay_text, basestring)\
and (essay_generated == 0 or essay_generated == 1):
self._id.append(max_id + 1)
self._score.append(essay_score)
# Clean text by removing non digit/work/punctuation characters
try:
essay_text=str(essay_text.encode('ascii', 'ignore'))
except:
essay_text = (essay_text.decode('utf-8','replace')).encode('ascii','ignore')
cleaned_essay=util_functions.sub_chars(essay_text).lower()
if(len(cleaned_essay)>MAXIMUM_ESSAY_LENGTH):
cleaned_essay=cleaned_essay[0:MAXIMUM_ESSAY_LENGTH]
self._text.append(cleaned_essay)
# Spell correct text using aspell
cleaned_text,spell_errors,markup_text=util_functions.spell_correct(self._text[len(self._text) - 1])
self._clean_text.append(cleaned_text)
self._spelling_errors.append(spell_errors)
self._markup_text.append(markup_text)
# Tokenize text
self._tokens.append(nltk.word_tokenize(self._clean_text[len(self._clean_text) - 1]))
# Part of speech tag text
self._pos.append(nltk.pos_tag(self._clean_text[len(self._clean_text) - 1].split(" ")))
self._generated.append(essay_generated)
# Stem spell corrected text
porter = nltk.PorterStemmer()
por_toks = " ".join([porter.stem(w) for w in self._tokens[len(self._tokens) - 1]])
self._clean_stem_text.append(por_toks)
ret = "text: " + self._text[len(self._text) - 1] + " score: " + str(essay_score)
else:
raise util_functions.InputError(essay_text, "arguments need to be in format "
"(text,score). text needs to be string,"
" score needs to be int.")
def update_prompt(self, prompt_text):
"""
Update the default prompt string, which is "".
prompt_text should be a string.
Returns the prompt as a confirmation.
"""
if(type(prompt_text) == type("text")):
self._prompt = util_functions.sub_chars(prompt_text)
ret = self._prompt
else:
raise util_functions.InputError(prompt_text, "Invalid prompt. Need to enter a string value.")
return ret
def generate_additional_essays(self, e_text, e_score, dict=None, max_syns=3):
"""
Substitute synonyms to generate extra essays from existing ones.
This is done to increase the amount of training data.
Should only be used with lowest scoring essays.
e_text is the text of the original essay.
e_score is the score of the original essay.
dict is a fixed dictionary (list) of words to replace.
max_syns defines the maximum number of additional essays to generate. Do not set too high.
"""
random.seed(1)
e_toks = nltk.word_tokenize(e_text)
all_syns = []
for word in e_toks:
synonyms = util_functions.get_wordnet_syns(word)
if(len(synonyms) > max_syns):
synonyms = random.sample(synonyms, max_syns)
all_syns.append(synonyms)
new_essays = []
for i in range(0, max_syns):
syn_toks = e_toks
for z in range(0, len(e_toks)):
if len(all_syns[z]) > i and (dict == None or e_toks[z] in dict):
syn_toks[z] = all_syns[z][i]
new_essays.append(" ".join(syn_toks))
for z in xrange(0, len(new_essays)):
self.add_essay(new_essays[z], e_score, 1)
\ No newline at end of file
machine_learning/external_code/fisher/LICENSE.txt 0 → 100644
Redistribution and use in source and binary forms, with or without modification, are permitted provided that the following conditions are met:
Redistributions of source code must retain the above copyright notice, this list of conditions and the following disclaimer.
Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution.
The names of its contributors may not be used to endorse or promote products derived from this software without specific prior written permission.
THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL JEET SUKUMARAN OR MARK T. HOLDER BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
\ No newline at end of file
machine_learning/external_code/fisher/fisher.py 0 → 100644
#! /usr/bin/env python
##############################################################################
# Following functions have been taken from the DendroPy library from:
##
## DendroPy Phylogenetic Computing Library.
##
## Copyright 2010 Jeet Sukumaran and Mark T. Holder.
## All rights reserved.
##
## See "LICENSE.txt" for terms and conditions of usage.
##
## If you use this work or any portion thereof in published work,
## please cite it as:
##
## Sukumaran, J. and M. T. Holder. 2010. DendroPy: a Python library
## for phylogenetic computing. Bioinformatics 26: 1569-1571.
##
##############################################################################
import math
## From dendropy.mathlib.probability
def hypergeometric_pmf(x, m, n, k):
"""
Given a population consisting of `m` items of class M and `n` items of class N,
this returns the probability of observing `x` items of class M when sampling
`k` times without replacement from the entire population (i.e., {M,N})
p(x) = (choose(m, x) * choose(n, k-x)) / choose(m+n, k)
"""
# following fails with 'OverflowError: long int too large to convert to
# float' with large numbers
# return float(binomial_coefficient(m, x) * binomial_coefficient(n, k-x))/binomial_coefficient(m+n, k)
a = math.log(binomial_coefficient(m, x))
b = math.log(binomial_coefficient(n, k-x))
c = math.log(binomial_coefficient(m+n, k))
return math.exp(a+b-c)
## From dendropy.mathlib.probability
def binomial_coefficient(population, sample):
"Returns `population` choose `sample`."
s = max(sample, population - sample)
assert s <= population
assert population > -1
if s == population:
return 1
numerator = 1
denominator = 1
for i in xrange(s+1, population + 1):
numerator *= i
denominator *= (i - s)
return numerator/denominator
## From dendropy.mathlib.statistics
class FishersExactTest(object):
"""
Given a 2x2 table:
+---+---+
| a | b |
+---+---+
| c | d |
+---+---+
represented by a list of lists::
[[a,b],[c,d]]
this calculates the sum of the probability of this table and all others
more extreme under the null hypothesis that there is no association between
the categories represented by the vertical and horizontal axes.
"""
def probability_of_table(table):
"""
Given a 2x2 table:
+---+---+
| a | b |
+---+---+
| c | d |
+---+---+
represented by a list of lists::
[[a,b],[c,d]]
this returns the probability of this table under the null hypothesis of
no association between rows and columns, which was shown by Fisher to be
a hypergeometric distribution:
p = ( choose(a+b, a) * choose(c+d, c) ) / choose(a+b+c+d, a+c)
"""
a = table[0][0]
b = table[0][1]
c = table[1][0]
d = table[1][1]
return hypergeometric_pmf(a, a+b, c+d, a+c)
probability_of_table = staticmethod(probability_of_table)
def __init__(self, table):
self.table = table
self.flat_table = [table[0][0], table[0][1], table[1][0], table[1][1]]
self.min_value = min(self.flat_table)
self.max_value = max(self.flat_table)
def _rotate_cw(self, table):
"""
Returns a copy of table such that all the values
are rotated clockwise once.
"""
return [ [ table[1][0], table[0][0] ],
[table[1][1], table[0][1] ] ]
def _min_rotation(self):
"""
Returns copy of self.table such that the smallest value is in the first
(upper left) cell.
"""
table = [list(self.table[0]), list(self.table[1])]
while table[0][0] != self.min_value:
table = self._rotate_cw(table)
return table
def _max_rotation(self):
"""
Returns copy of self.table such that the largest value is in the first
(upper left) cell.
"""
table = [list(self.table[0]), list(self.table[1])]
while table[0][0] != self.max_value:
table = self._rotate_cw(table)
return table
def _sum_left_tail(self):
# left_tail_tables = self._get_left_tail_tables()
# p_vals = [ self.probability_of_table(t) for t in left_tail_tables ]
p_vals = self._get_left_tail_probs()
return sum(p_vals)
def _sum_right_tail(self):
# right_tail_tables = self._get_right_tail_tables()
# p_vals = [ self.probability_of_table(t) for t in right_tail_tables ]
p_vals = self._get_right_tail_probs()
return sum(p_vals)
def _get_left_tail_probs(self):
table = self._min_rotation()
row_totals = [sum(table[0]), sum(table[1])]
col_totals = [table[0][0] + table[1][0], table[0][1] + table[1][1]]
p_vals = []
while True:
table[0][0] -= 1
if table[0][0] < 0:
break
table[0][1] = row_totals[0] - table[0][0]
table[1][0] = col_totals[0] - table[0][0]
table[1][1] = row_totals[1] - table[1][0]
p_vals.append(self.probability_of_table(table))
return p_vals
def _get_right_tail_probs(self):
table = self._min_rotation()
row_totals = [sum(table[0]), sum(table[1])]
col_totals = [table[0][0] + table[1][0], table[0][1] + table[1][1]]
p_vals = []
while True:
table[0][0] += 1
table[0][1] = row_totals[0] - table[0][0]
if table[0][1] < 0:
break
table[1][0] = col_totals[0] - table[0][0]
if table[1][0] < 0:
break
table[1][1] = row_totals[1] - table[1][0]
if table[1][1] < 0:
break
p_vals.append(self.probability_of_table(table))
return p_vals
def _get_left_tail_tables(self):
table = self._min_rotation()
row_totals = [sum(table[0]), sum(table[1])]
col_totals = [table[0][0] + table[1][0], table[0][1] + table[1][1]]
left_tail_tables = []
while True:
table[0][0] -= 1
if table[0][0] < 0:
break
table[0][1] = row_totals[0] - table[0][0]
table[1][0] = col_totals[0] - table[0][0]
table[1][1] = row_totals[1] - table[1][0]
left_tail_tables.append([list(table[0]), list(table[1])])
return left_tail_tables
def _get_right_tail_tables(self):
table = self._min_rotation()
row_totals = [sum(table[0]), sum(table[1])]
col_totals = [table[0][0] + table[1][0], table[0][1] + table[1][1]]
right_tail_tables = []
while True:
table[0][0] += 1
table[0][1] = row_totals[0] - table[0][0]
if table[0][1] < 0:
break
table[1][0] = col_totals[0] - table[0][0]
if table[1][0] < 0:
break
table[1][1] = row_totals[1] - table[1][0]
if table[1][1] < 0:
break
right_tail_tables.append([list(table[0]), list(table[1])])
return right_tail_tables
def left_tail_p(self):
"""
Returns the sum of probabilities of this table and all others more
extreme.
"""
return self.probability_of_table(self.table) + self._sum_left_tail()
def right_tail_p(self):
"""
Returns the sum of probabilities of this table and all others more
extreme.
"""
return self.probability_of_table(self.table) + self._sum_right_tail()
def two_tail_p(self):
"""
Returns the sum of probabilities of this table and all others more
extreme.
"""
p0 = self.probability_of_table(self.table)
all_p_vals = self._get_left_tail_probs() + self._get_right_tail_probs()
p_vals = []
for p in all_p_vals:
if p <= p0:
p_vals.append(p)
return sum(p_vals) + p0
def assert_almost_equal(v1, v2, prec=8):
if abs(v1-v2) <= 10**(-prec):
print "OK: {} == {}".format(v1, v2)
else:
print "FAIL: {} != {}".format(v1, v2)
if __name__ == "__main__":
table = [[12, 5], [29, 2]]
ft = FishersExactTest(table)
assert_almost_equal(ft.left_tail_p(), 0.044554737835078267)
assert_almost_equal(ft.right_tail_p(), 0.99452520602190897)
assert_almost_equal(ft.two_tail_p(), 0.08026855207410688)
\ No newline at end of file
machine_learning/feature_extractor.py 0 → 100644
"""
Extracts features from training set and test set essays
"""
import numpy
import re
import nltk
import sys
from sklearn.feature_extraction.text import CountVectorizer
import pickle
import os
from itertools import chain
import copy
import operator
import logging
base_path = os.path.dirname(__file__)
sys.path.append(base_path)
from essay_set import EssaySet
import util_functions
if not base_path.endswith("/"):
base_path=base_path+"/"
log = logging.getLogger(__name__)
#Paths to needed data files
NGRAM_PATH = base_path + "data/good_pos_ngrams.p"
ESSAY_CORPUS_PATH = util_functions.ESSAY_CORPUS_PATH
class FeatureExtractor(object):
def __init__(self):
self._good_pos_ngrams = self.get_good_pos_ngrams()
self.dict_initialized = False
self._spell_errors_per_character=0
self._grammar_errors_per_character=0
def initialize_dictionaries(self, e_set, max_feats2 = 200):
"""
Initializes dictionaries from an essay set object
Dictionaries must be initialized prior to using this to extract features
e_set is an input essay set
returns a confirmation of initialization
"""
if(hasattr(e_set, '_type')):
if(e_set._type == "train"):
#normal text (unstemmed) useful words/bigrams
nvocab = util_functions.get_vocab(e_set._text, e_set._score, max_feats2 = max_feats2)
#stemmed and spell corrected vocab useful words/ngrams
svocab = util_functions.get_vocab(e_set._clean_stem_text, e_set._score, max_feats2 = max_feats2)
#dictionary trained on proper vocab
self._normal_dict = CountVectorizer(ngram_range=(1,2), vocabulary=nvocab)
#dictionary trained on proper vocab
self._stem_dict = CountVectorizer(ngram_range=(1,2), vocabulary=svocab)
self.dict_initialized = True
#Average spelling errors in set. needed later for spelling detection
self._mean_spelling_errors=sum(e_set._spelling_errors)/float(len(e_set._spelling_errors))
self._spell_errors_per_character=sum(e_set._spelling_errors)/float(sum([len(t) for t in e_set._text]))
#Gets the number and positions of grammar errors
good_pos_tags,bad_pos_positions=self._get_grammar_errors(e_set._pos,e_set._text,e_set._tokens)
self._grammar_errors_per_character=(sum(good_pos_tags)/float(sum([len(t) for t in e_set._text])))
#Generate bag of words features
bag_feats=self.gen_bag_feats(e_set)
#Sum of a row of bag of words features (topical words in an essay)
f_row_sum=numpy.sum(bag_feats[:,:])
#Average index of how "topical" essays are
self._mean_f_prop=f_row_sum/float(sum([len(t) for t in e_set._text]))
ret = "ok"
else:
raise util_functions.InputError(e_set, "needs to be an essay set of the train type.")
else:
raise util_functions.InputError(e_set, "wrong input. need an essay set object")
return ret
def get_good_pos_ngrams(self):
"""
Gets a list of gramatically correct part of speech sequences from an input file called essaycorpus.txt
Returns the list and caches the file
"""
if(os.path.isfile(NGRAM_PATH)):
good_pos_ngrams = pickle.load(open(NGRAM_PATH, 'rb'))
elif os.path.isfile(ESSAY_CORPUS_PATH):
essay_corpus = open(ESSAY_CORPUS_PATH).read()
essay_corpus = util_functions.sub_chars(essay_corpus)
good_pos_ngrams = util_functions.regenerate_good_tokens(essay_corpus)
pickle.dump(good_pos_ngrams, open(NGRAM_PATH, 'wb'))
else:
#Hard coded list in case the needed files cannot be found
good_pos_ngrams=['NN PRP', 'NN PRP .', 'NN PRP . DT', 'PRP .', 'PRP . DT', 'PRP . DT NNP', '. DT',
'. DT NNP', '. DT NNP NNP', 'DT NNP', 'DT NNP NNP', 'DT NNP NNP NNP', 'NNP NNP',
'NNP NNP NNP', 'NNP NNP NNP NNP', 'NNP NNP NNP .', 'NNP NNP .', 'NNP NNP . TO',
'NNP .', 'NNP . TO', 'NNP . TO NNP', '. TO', '. TO NNP', '. TO NNP NNP',
'TO NNP', 'TO NNP NNP']
return good_pos_ngrams
def _get_grammar_errors(self,pos,text,tokens):
"""
Internal function to get the number of grammar errors in given text
pos - part of speech tagged text (list)
text - normal text (list)
tokens - list of lists of tokenized text
"""
word_counts = [max(len(t),1) for t in tokens]
good_pos_tags = []
min_pos_seq=2
max_pos_seq=4
bad_pos_positions=[]
for i in xrange(0, len(text)):
pos_seq = [tag[1] for tag in pos[i]]
pos_ngrams = util_functions.ngrams(pos_seq, min_pos_seq, max_pos_seq)
long_pos_ngrams=[z for z in pos_ngrams if z.count(' ')==(max_pos_seq-1)]
bad_pos_tuples=[[z,z+max_pos_seq] for z in xrange(0,len(long_pos_ngrams)) if long_pos_ngrams[z] not in self._good_pos_ngrams]
bad_pos_tuples.sort(key=operator.itemgetter(1))
to_delete=[]
for m in reversed(xrange(len(bad_pos_tuples)-1)):
start, end = bad_pos_tuples[m]
for j in xrange(m+1, len(bad_pos_tuples)):
lstart, lend = bad_pos_tuples[j]
if lstart >= start and lstart <= end:
bad_pos_tuples[m][1]=bad_pos_tuples[j][1]
to_delete.append(j)
fixed_bad_pos_tuples=[bad_pos_tuples[z] for z in xrange(0,len(bad_pos_tuples)) if z not in to_delete]
bad_pos_positions.append(fixed_bad_pos_tuples)
overlap_ngrams = [z for z in pos_ngrams if z in self._good_pos_ngrams]
if (len(pos_ngrams)-len(overlap_ngrams))>0:
divisor=len(pos_ngrams)/len(pos_seq)
else:
divisor=1
good_pos_tags.append((len(pos_ngrams)-len(overlap_ngrams))/divisor)
return good_pos_tags,bad_pos_positions
def gen_length_feats(self, e_set):
"""
Generates length based features from an essay set
Generally an internal function called by gen_feats
Returns an array of length features
e_set - EssaySet object
"""
text = e_set._text
lengths = [len(e) for e in text]
word_counts = [max(len(t),1) for t in e_set._tokens]
comma_count = [e.count(",") for e in text]
ap_count = [e.count("'") for e in text]
punc_count = [e.count(".") + e.count("?") + e.count("!") for e in text]
chars_per_word = [lengths[m] / float(word_counts[m]) for m in xrange(0, len(text))]
good_pos_tags,bad_pos_positions= self._get_grammar_errors(e_set._pos,e_set._text,e_set._tokens)
good_pos_tag_prop = [good_pos_tags[m] / float(word_counts[m]) for m in xrange(0, len(text))]
length_arr = numpy.array((
lengths, word_counts, comma_count, ap_count, punc_count, chars_per_word, good_pos_tags,
good_pos_tag_prop)).transpose()
return length_arr.copy()
def gen_bag_feats(self, e_set):
"""
Generates bag of words features from an input essay set and trained FeatureExtractor
Generally called by gen_feats
Returns an array of features
e_set - EssaySet object
"""
if(hasattr(self, '_stem_dict')):
sfeats = self._stem_dict.transform(e_set._clean_stem_text)
nfeats = self._normal_dict.transform(e_set._text)
bag_feats = numpy.concatenate((sfeats.toarray(), nfeats.toarray()), axis=1)
else:
raise util_functions.InputError(self, "Dictionaries must be initialized prior to generating bag features.")
return bag_feats.copy()
def gen_feats(self, e_set):
"""
Generates bag of words, length, and prompt features from an essay set object
returns an array of features
e_set - EssaySet object
"""
bag_feats = self.gen_bag_feats(e_set)
length_feats = self.gen_length_feats(e_set)
prompt_feats = self.gen_prompt_feats(e_set)
overall_feats = numpy.concatenate((length_feats, prompt_feats, bag_feats), axis=1)
overall_feats = overall_feats.copy()
return overall_feats
def gen_prompt_feats(self, e_set):
"""
Generates prompt based features from an essay set object and internal prompt variable.
Generally called internally by gen_feats
Returns an array of prompt features
e_set - EssaySet object
"""
prompt_toks = nltk.word_tokenize(e_set._prompt)
expand_syns = []
for word in prompt_toks:
synonyms = util_functions.get_wordnet_syns(word)
expand_syns.append(synonyms)
expand_syns = list(chain.from_iterable(expand_syns))
prompt_overlap = []
prompt_overlap_prop = []
for j in e_set._tokens:
tok_length=len(j)
if(tok_length==0):
tok_length=1
prompt_overlap.append(len([i for i in j if i in prompt_toks]))
prompt_overlap_prop.append(prompt_overlap[len(prompt_overlap) - 1] / float(tok_length))
expand_overlap = []
expand_overlap_prop = []
for j in e_set._tokens:
tok_length=len(j)
if(tok_length==0):
tok_length=1
expand_overlap.append(len([i for i in j if i in expand_syns]))
expand_overlap_prop.append(expand_overlap[len(expand_overlap) - 1] / float(tok_length))
prompt_arr = numpy.array((prompt_overlap, prompt_overlap_prop, expand_overlap, expand_overlap_prop)).transpose()
return prompt_arr.copy()
def gen_feedback(self, e_set, features=None):
"""
Generate feedback for a given set of essays
e_set - EssaySet object
features - optionally, pass in a matrix of features extracted from e_set using FeatureExtractor
in order to get off topic feedback.
Returns a list of lists (one list per essay in e_set)
e_set - EssaySet object
"""
#Set ratio to modify thresholds for grammar/spelling errors
modifier_ratio=1.05
#Calc number of grammar and spelling errors per character
set_grammar,bad_pos_positions=self._get_grammar_errors(e_set._pos,e_set._text,e_set._tokens)
set_grammar_per_character=[set_grammar[m]/float(len(e_set._text[m])+.1) for m in xrange(0,len(e_set._text))]
set_spell_errors_per_character=[e_set._spelling_errors[m]/float(len(e_set._text[m])+.1) for m in xrange(0,len(e_set._text))]
#Iterate through essays and create a feedback dict for each
all_feedback=[]
for m in xrange(0,len(e_set._text)):
#Be very careful about changing these messages!
individual_feedback={'grammar' : "Grammar: Ok.",
'spelling' : "Spelling: Ok.",
'markup_text' : "",
'grammar_per_char' : set_grammar_per_character[m],
'spelling_per_char' : set_spell_errors_per_character[m],
'too_similar_to_prompt' : False,
}
markup_tokens=e_set._markup_text[m].split(" ")
#This loop ensures that sequences of bad grammar get put together into one sequence instead of staying
#disjointed
bad_pos_starts=[z[0] for z in bad_pos_positions[m]]
bad_pos_ends=[z[1]-1 for z in bad_pos_positions[m]]
for z in xrange(0,len(markup_tokens)):
if z in bad_pos_starts:
markup_tokens[z]='<bg>' + markup_tokens[z]
elif z in bad_pos_ends:
markup_tokens[z]=markup_tokens[z] + "</bg>"
if(len(bad_pos_ends)>0 and len(bad_pos_starts)>0 and len(markup_tokens)>1):
if max(bad_pos_ends)>(len(markup_tokens)-1) and max(bad_pos_starts)<(len(markup_tokens)-1):
markup_tokens[len(markup_tokens)-1]+="</bg>"
#Display messages if grammar/spelling errors greater than average in training set
if set_grammar_per_character[m]>(self._grammar_errors_per_character*modifier_ratio):
individual_feedback['grammar']="Grammar: More grammar errors than average."
if set_spell_errors_per_character[m]>(self._spell_errors_per_character*modifier_ratio):
individual_feedback['spelling']="Spelling: More spelling errors than average."
#Test topicality by calculating # of on topic words per character and comparing to the training set
#mean. Requires features to be passed in
if features is not None:
f_row_sum=numpy.sum(features[m,12:])
f_row_prop=f_row_sum/len(e_set._text[m])
if f_row_prop<(self._mean_f_prop/1.5) or len(e_set._text[m])<20:
individual_feedback['topicality']="Topicality: Essay may be off topic."
if(features[m,9]>.6):
individual_feedback['prompt_overlap']="Prompt Overlap: Too much overlap with prompt."
individual_feedback['too_similar_to_prompt']=True
log.debug(features[m,9])
#Create string representation of markup text
markup_string=" ".join(markup_tokens)
individual_feedback['markup_text']=markup_string
all_feedback.append(individual_feedback)
return all_feedback
machine_learning/grade.py 0 → 100644
"""
Functions to score specified data using specified ML models
"""
import sys
import pickle
import os
import numpy
import logging
from statsd import statsd
#Append sys to base path to import the following modules
base_path = os.path.dirname(__file__)
sys.path.append(base_path)
#Depend on base path to be imported
from essay_set import EssaySet
import predictor_extractor
import predictor_set
import util_functions
#Imports needed to unpickle grader data
import feature_extractor
import sklearn.ensemble
import math
log = logging.getLogger(__name__)
@statsd.timed('open_ended_assessment.machine_learning.grader.time')
def grade(grader_data,submission):
"""
Grades a specified submission using specified models
grader_data - A dictionary:
{
'model' : trained model,
'extractor' : trained feature extractor,
'prompt' : prompt for the question,
'algorithm' : algorithm for the question,
}
submission - The student submission (string)
"""
#Initialize result dictionary
results = {'errors': [],'tests': [],'score': 0, 'feedback' : "", 'success' : False, 'confidence' : 0}
has_error=False
grader_set=EssaySet(type="test")
#This is to preserve legacy functionality
if 'algorithm' not in grader_data:
grader_data['algorithm'] = util_functions.AlgorithmTypes.classification
try:
#Try to add essay to essay set object
grader_set.add_essay(str(submission),0)
grader_set.update_prompt(str(grader_data['prompt']))
except:
results['errors'].append("Essay could not be added to essay set:{0}".format(submission))
has_error=True
#Try to extract features from submission and assign score via the model
try:
grader_feats=grader_data['extractor'].gen_feats(grader_set)
feedback=grader_data['extractor'].gen_feedback(grader_set,grader_feats)[0]
results['score']=int(grader_data['model'].predict(grader_feats)[0])
except :
results['errors'].append("Could not extract features and score essay.")
has_error=True
#Try to determine confidence level
try:
results['confidence'] = get_confidence_value(grader_data['algorithm'], grader_data['model'], grader_feats, results['score'], grader_data['score'])
except:
#If there is an error getting confidence, it is not a show-stopper, so just log
log.exception("Problem generating confidence value")
if not has_error:
#If the essay is just a copy of the prompt, return a 0 as the score
if(feedback['too_similar_to_prompt']):
results['score']=0
results['correct']=False
results['success']=True
#Generate short form output--number of problem areas identified in feedback
#Add feedback to results if available
results['feedback'] = {}
if 'topicality' in feedback and 'prompt_overlap' in feedback:
results['feedback'].update({
'topicality' : feedback['topicality'],
'prompt-overlap' : feedback['prompt_overlap'],
})
results['feedback'].update(
{
'spelling' : feedback['spelling'],
'grammar' : feedback['grammar'],
'markup-text' : feedback['markup_text'],
}
)
else:
#If error, success is False.
results['success']=False
#Count number of successful/unsuccessful gradings
statsd.increment("open_ended_assessment.machine_learning.grader_count",
tags=["success:{0}".format(results['success'])])
return results
def grade_generic(grader_data, numeric_features, textual_features):
"""
Grades a set of numeric and textual features using a generic model
grader_data -- dictionary containing:
{
'algorithm' - Type of algorithm to use to score
}
numeric_features - list of numeric features to predict on
textual_features - list of textual feature to predict on
"""
results = {'errors': [],'tests': [],'score': 0, 'success' : False, 'confidence' : 0}
has_error=False
#Try to find and load the model file
grader_set=predictor_set.PredictorSet(type="test")
#Try to add essays to essay set object
try:
grader_set.add_row(numeric_features, textual_features,0)
except:
results['errors'].append("Row could not be added to predictor set:{0} {1}".format(numeric_features, textual_features))
has_error=True
#Try to extract features from submission and assign score via the model
try:
grader_feats=grader_data['extractor'].gen_feats(grader_set)
results['score']=grader_data['model'].predict(grader_feats)[0]
except :
results['errors'].append("Could not extract features and score essay.")
has_error=True
#Try to determine confidence level
try:
results['confidence'] = get_confidence_value(grader_data['algorithm'], grader_data['model'], grader_feats, results['score'])
except:
#If there is an error getting confidence, it is not a show-stopper, so just log
log.exception("Problem generating confidence value")
#Count number of successful/unsuccessful gradings
statsd.increment("open_ended_assessment.machine_learning.grader_count",
tags=["success:{0}".format(results['success'])])
if not has_error:
results['success'] = True
return results
def get_confidence_value(algorithm,model,grader_feats,score, scores):
"""
Determines a confidence in a certain score, given proper input parameters
algorithm- from util_functions.AlgorithmTypes
model - a trained model
grader_feats - a row of features used by the model for classification/regression
score - The score assigned to the submission by a prior model
"""
min_score=min(numpy.asarray(scores))
max_score=max(numpy.asarray(scores))
if algorithm == util_functions.AlgorithmTypes.classification:
#If classification, predict with probability, which gives you a matrix of confidences per score point
raw_confidence=model.predict_proba(grader_feats)[0,(score-min_score)]
#TODO: Normalize confidence somehow here
confidence=raw_confidence
else:
raw_confidence = model.predict(grader_feats)[0]
confidence = max(raw_confidence - math.floor(raw_confidence), math.ceil(raw_confidence) - raw_confidence)
return confidence
machine_learning/model_creator.py 0 → 100644
#Provides interface functions to create and save models
import numpy
import re
import nltk
import sys
from sklearn.feature_extraction.text import CountVectorizer
import pickle
import os
import sklearn.ensemble
from itertools import chain
base_path = os.path.dirname(__file__)
sys.path.append(base_path)
from essay_set import EssaySet
import util_functions
import feature_extractor
import logging
import predictor_extractor
log=logging.getLogger()
def read_in_test_data(filename):
"""
Reads in test data file found at filename.
filename must be a tab delimited file with columns id, dummy number column, score, dummy score, text
returns the score and the text
"""
id, e_set, score, score2, text = [], [], [], [], []
combined_raw = open(filename).read()
raw_lines = combined_raw.splitlines()
for row in xrange(1, len(raw_lines)):
id1, set1, score1, score12, text1 = raw_lines[row].strip().split("\t")
id.append(int(id1))
text.append(text1)
e_set.append(int(set1))
score.append(int(score1))
score2.append(int(score12))
return score, text
def read_in_test_prompt(filename):
"""
Reads in the prompt from a text file
Returns string
"""
prompt_string = open(filename).read()
return prompt_string
def read_in_test_data_twocolumn(filename,sep=","):
"""
Reads in a two column version of the test data.
Filename must point to a delimited file.
In filename, the first column should be integer score data.
The second column should be string text data.
Sep specifies the type of separator between fields.
"""
score, text = [], []
combined_raw = open(filename).read()
raw_lines = combined_raw.splitlines()
for row in xrange(1, len(raw_lines)):
score1, text1 = raw_lines[row].strip().split("\t")
text.append(text1)
score.append(int(score1))
return score, text
def create_essay_set(text, score, prompt_string, generate_additional=True):
"""
Creates an essay set from given data.
Text should be a list of strings corresponding to essay text.
Score should be a list of scores where score[n] corresponds to text[n]
Prompt string is just a string containing the essay prompt.
Generate_additional indicates whether to generate additional essays at the minimum score point or not.
"""
x = EssaySet()
for i in xrange(0, len(text)):
x.add_essay(text[i], score[i])
if score[i] == min(score) and generate_additional == True:
x.generate_additional_essays(x._clean_text[len(x._clean_text) - 1], score[i])
x.update_prompt(prompt_string)
return x
def get_cv_error(clf,feats,scores):
"""
Gets cross validated error for a given classifier, set of features, and scores
clf - classifier
feats - features to feed into the classified and cross validate over
scores - scores associated with the features -- feature row 1 associates with score 1, etc.
"""
results={'success' : False, 'kappa' : 0, 'mae' : 0}
try:
cv_preds=util_functions.gen_cv_preds(clf,feats,scores)
err=numpy.mean(numpy.abs(numpy.array(cv_preds)-scores))
kappa=util_functions.quadratic_weighted_kappa(list(cv_preds),scores)
results['mae']=err
results['kappa']=kappa
results['success']=True
except ValueError:
#If this is hit, everything is fine. It is hard to explain why the error occurs, but it isn't a big deal.
log.exception("Not enough classes (0,1,etc) in each cross validation fold.")
except:
log.exception("Error getting cv error estimates.")
return results
def get_algorithms(type):
"""
Gets two classifiers for each type of algorithm, and returns them. First for predicting, second for cv error.
type - one of util_functions.AlgorithmTypes
"""
if type == util_functions.AlgorithmTypes.classification:
clf = sklearn.ensemble.GradientBoostingClassifier(n_estimators=100, learn_rate=.05,
max_depth=4, random_state=1,min_samples_leaf=3)
clf2=sklearn.ensemble.GradientBoostingClassifier(n_estimators=100, learn_rate=.05,
max_depth=4, random_state=1,min_samples_leaf=3)
else:
clf = sklearn.ensemble.GradientBoostingRegressor(n_estimators=100, learn_rate=.05,
max_depth=4, random_state=1,min_samples_leaf=3)
clf2=sklearn.ensemble.GradientBoostingRegressor(n_estimators=100, learn_rate=.05,
max_depth=4, random_state=1,min_samples_leaf=3)
return clf, clf2
def extract_features_and_generate_model_predictors(predictor_set, type=util_functions.AlgorithmTypes.regression):
"""
Extracts features and generates predictors based on a given predictor set
predictor_set - a PredictorSet object that has been initialized with data
type - one of util_functions.AlgorithmType
"""
if(algorithm not in [util_functions.AlgorithmTypes.regression, util_functions.AlgorithmTypes.classification]):
algorithm = util_functions.AlgorithmTypes.regression
f = predictor_extractor.PredictorExtractor()
f.initialize_dictionaries(predictor_set)
train_feats = f.gen_feats(predictor_set)
clf,clf2 = get_algorithms(type)
cv_error_results=get_cv_error(clf2,train_feats,predictor_set._target)
try:
set_score = numpy.asarray(predictor_set._target, dtype=numpy.int)
clf.fit(train_feats, set_score)
except ValueError:
log.exception("Not enough classes (0,1,etc) in sample.")
set_score[0]=1
set_score[1]=0
clf.fit(train_feats, set_score)
return f, clf, cv_error_results
def extract_features_and_generate_model(essays, type=util_functions.AlgorithmTypes.regression):
"""
Feed in an essay set to get feature vector and classifier
essays must be an essay set object
additional array is an optional argument that can specify
a numpy array of values to add in
returns a trained FeatureExtractor object and a trained classifier
"""
f = feature_extractor.FeatureExtractor()
f.initialize_dictionaries(essays)
train_feats = f.gen_feats(essays)
set_score = numpy.asarray(essays._score, dtype=numpy.int)
if len(util_functions.f7(list(set_score)))>5:
type = util_functions.AlgorithmTypes.regression
else:
type = util_functions.AlgorithmTypes.classification
clf,clf2 = get_algorithms(type)
cv_error_results=get_cv_error(clf2,train_feats,essays._score)
try:
clf.fit(train_feats, set_score)
except ValueError:
log.exception("Not enough classes (0,1,etc) in sample.")
set_score[0]=1
set_score[1]=0
clf.fit(train_feats, set_score)
return f, clf, cv_error_results
def dump_model_to_file(prompt_string, feature_ext, classifier, text, score, model_path):
"""
Writes out a model to a file.
prompt string is a string containing the prompt
feature_ext is a trained FeatureExtractor object
classifier is a trained classifier
model_path is the path of write out the model file to
"""
model_file = {'prompt': prompt_string, 'extractor': feature_ext, 'model': classifier, 'text' : text, 'score' : score}
pickle.dump(model_file, file=open(model_path, "w"))
def create_essay_set_and_dump_model(text,score,prompt,model_path,additional_array=None):
"""
Function that creates essay set, extracts features, and writes out model
See above functions for argument descriptions
"""
essay_set=create_essay_set(text_score,prompt)
feature_ext,clf=extract_features_and_generate_model(essay_set,additional_array)
dump_model_to_file(prompt,feature_ext,clf,model_path)
machine_learning/predictor_extractor.py 0 → 100644
"""
Extracts features for an arbitrary set of textual and numeric inputs
"""
import numpy
import re
import nltk
import sys
from sklearn.feature_extraction.text import CountVectorizer
import pickle
import os
from itertools import chain
import copy
import operator
import logging
import math
from feature_extractor import FeatureExtractor
#Append to path and then import things that depend on path
base_path = os.path.dirname(__file__)
sys.path.append(base_path)
from essay_set import EssaySet
import util_functions
if not base_path.endswith("/"):
base_path=base_path+"/"
log = logging.getLogger(__name__)
class PredictorExtractor(object):
def __init__(self):
self._extractors = []
self._initialized = False
def initialize_dictionaries(self, p_set):
"""
Initialize dictionaries with the textual inputs in the PredictorSet object
p_set - PredictorSet object that has had data fed in
"""
success = False
if not (hasattr(p_set, '_type')):
error_message = "needs to be an essay set of the train type."
log.exception(error_message)
raise util_functions.InputError(p_set, error_message)
if not (p_set._type == "train"):
error_message = "needs to be an essay set of the train type."
log.exception(error_message)
raise util_functions.InputError(p_set, error_message)
div_length=len(p_set._essay_sets)
if div_length==0:
div_length=1
#Ensures that even with a large amount of input textual features, training time stays reasonable
max_feats2 = int(math.floor(200/div_length))
for i in xrange(0,len(p_set._essay_sets)):
self._extractors.append(FeatureExtractor())
self._extractors[i].initialize_dictionaries(p_set._essay_sets[i], max_feats2=max_feats2)
self._initialized = True
success = True
return success
def gen_feats(self, p_set):
"""
Generates features based on an iput p_set
p_set - PredictorSet
"""
if self._initialized!=True:
error_message = "Dictionaries have not been initialized."
log.exception(error_message)
raise util_functions.InputError(p_set, error_message)
textual_features = []
for i in xrange(0,len(p_set._essay_sets)):
textual_features.append(self._extractors[i].gen_feats(p_set._essay_sets[i]))
textual_matrix = numpy.concatenate(textual_features, axis=1)
predictor_matrix = numpy.array(p_set._numeric_features)
print textual_matrix.shape
print predictor_matrix.shape
overall_matrix = numpy.concatenate((textual_matrix, predictor_matrix), axis=1)
return overall_matrix.copy()
machine_learning/predictor_set.py 0 → 100644
import numpy
import nltk
import sys
import random
import os
import logging
import essay_set
base_path = os.path.dirname(__file__)
sys.path.append(base_path)
import util_functions
if not base_path.endswith("/"):
base_path=base_path+"/"
log=logging.getLogger(__name__)
class PredictorSet(object):
def __init__(self, type = "train"):
"""
Initialize variables and check essay set type
"""
if(type != "train" and type != "test"):
type = "train"
self._type = type
self._target=[]
self._textual_features=[]
self._numeric_features=[]
self._essay_sets=[]
def add_row(self, numeric_features, textual_features, target):
#Basic input checking
if not isinstance(target, (int, long, float)):
error_message = "Target is not a numeric value."
log.exception(error_message)
raise util_functions.InputError(target, error_message)
if not isinstance(numeric_features, list):
error_message = "Numeric features are not a list."
log.exception(error_message)
raise util_functions.InputError(numeric_features, error_message)
if not isinstance(textual_features, list):
error_message = "Textual features are not a list."
log.exception(error_message)
raise util_functions.InputError(textual_features, error_message)
#Do some length checking for parameters
if len(self._numeric_features)>0:
numeric_length = len(self._numeric_features[-1])
current_numeric_length = len(numeric_features)
if numeric_length != current_numeric_length:
error_message = "Numeric features are an improper length."
log.exception(error_message)
raise util_functions.InputError(numeric_features, error_message)
if len(self._textual_features)>0:
textual_length = len(self._textual_features[-1])
current_textual_length = len(textual_features)
if textual_length != current_textual_length:
error_message = "Textual features are an improper length."
log.exception(error_message)
raise util_functions.InputError(textual_features, error_message)
#Now check to see if text features and numeric features are individually correct
for i in xrange(0,len(numeric_features)):
try:
numeric_features[i] = float(numeric_features[i])
except:
error_message = "Numeric feature {0} not numeric.".format(numeric_features[i])
log.exception(error_message)
raise util_functions.InputError(numeric_features, error_message)
for i in xrange(0,len(textual_features)):
try:
textual_features[i] = str(textual_features[i].encode('ascii', 'ignore'))
except:
error_message = "Textual feature {0} not string.".format(textual_features[i])
log.exception(error_message)
raise util_functions.InputError(textual_features, error_message)
#Create essay sets for textual features if needed
if len(self._textual_features)==0:
for i in xrange(0,len(textual_features)):
self._essay_sets.append(essay_set.EssaySet(type=self._type))
#Add numeric and textual features
self._numeric_features.append(numeric_features)
self._textual_features.append(textual_features)
#Add targets
self._target.append(target)
#Add textual features to essay sets
for i in xrange(0,len(textual_features)):
self._essay_sets[i].add_essay(textual_features[i], target)
machine_learning/util_functions.py 0 → 100644
#Collection of misc functions needed to support essay_set.py and feature_extractor.py.
#Requires aspell to be installed and added to the path
from external_code.fisher import fisher
aspell_path = "aspell"
import re
import os
from sklearn.feature_extraction.text import CountVectorizer
import numpy
from itertools import chain
import math
import nltk
import pickle
import logging
import sys
log=logging.getLogger(__name__)
base_path = os.path.dirname(__file__)
sys.path.append(base_path)
if not base_path.endswith("/"):
base_path=base_path+"/"
#Paths to needed data files
ESSAY_CORPUS_PATH = base_path + "data/essaycorpus.txt"
ESSAY_COR_TOKENS_PATH = base_path + "data/essay_cor_tokens.p"
class AlgorithmTypes(object):
"""
Defines what types of algorithm can be used
"""
regression = "regression"
classification = "classifiction"
def create_model_path(model_path):
"""
Creates a path to model files
model_path - string
"""
if not model_path.startswith("/") and not model_path.startswith("models/"):
model_path="/" + model_path
if not model_path.startswith("models"):
model_path = "models" + model_path
if not model_path.endswith(".p"):
model_path+=".p"
return model_path
def sub_chars(string):
"""
Strips illegal characters from a string. Used to sanitize input essays.
Removes all non-punctuation, digit, or letter characters.
Returns sanitized string.
string - string
"""
#Define replacement patterns
sub_pat = r"[^A-Za-z\.\?!,';:]"
char_pat = r"\."
com_pat = r","
ques_pat = r"\?"
excl_pat = r"!"
sem_pat = r";"
col_pat = r":"
whitespace_pat = r"\s{1,}"
#Replace text. Ordering is very important!
nstring = re.sub(sub_pat, " ", string)
nstring = re.sub(char_pat," .", nstring)
nstring = re.sub(com_pat, " ,", nstring)
nstring = re.sub(ques_pat, " ?", nstring)
nstring = re.sub(excl_pat, " !", nstring)
nstring = re.sub(sem_pat, " ;", nstring)
nstring = re.sub(col_pat, " :", nstring)
nstring = re.sub(whitespace_pat, " ", nstring)
return nstring
def spell_correct(string):
"""
Uses aspell to spell correct an input string.
Requires aspell to be installed and added to the path.
Returns the spell corrected string if aspell is found, original string if not.
string - string
"""
#Create a temp file so that aspell could be used
f = open('tmpfile', 'w')
f.write(string)
f_path = os.path.abspath(f.name)
f.close()
try:
p = os.popen(aspell_path + " -a < " + f_path + " --sug-mode=ultra")
except:
log.exception("Could not find aspell, so could not spell correct!")
#Return original string if aspell fails
return string,0, string
#Aspell returns a list of incorrect words with the above flags
incorrect = p.readlines()
p.close()
incorrect_words = list()
correct_spelling = list()
for i in range(1, len(incorrect)):
if(len(incorrect[i]) > 10):
#Reformat aspell output to make sense
match = re.search(":", incorrect[i])
if hasattr(match, "start"):
begstring = incorrect[i][2:match.start()]
begmatch = re.search(" ", begstring)
begword = begstring[0:begmatch.start()]
sugstring = incorrect[i][match.start() + 2:]
sugmatch = re.search(",", sugstring)
if hasattr(sugmatch, "start"):
sug = sugstring[0:sugmatch.start()]
incorrect_words.append(begword)
correct_spelling.append(sug)
#Create markup based on spelling errors
newstring = string
markup_string = string
already_subbed=[]
for i in range(0, len(incorrect_words)):
sub_pat = r"\b" + incorrect_words[i] + r"\b"
sub_comp = re.compile(sub_pat)
newstring = re.sub(sub_comp, correct_spelling[i], newstring)
if incorrect_words[i] not in already_subbed:
markup_string=re.sub(sub_comp,'<bs>' + incorrect_words[i] + "</bs>", markup_string)
already_subbed.append(incorrect_words[i])
return newstring,len(incorrect_words),markup_string
def ngrams(tokens, min_n, max_n):
"""
Generates ngrams(word sequences of fixed length) from an input token sequence.
tokens is a list of words.
min_n is the minimum length of an ngram to return.
max_n is the maximum length of an ngram to return.
returns a list of ngrams (words separated by a space)
"""
all_ngrams = list()
n_tokens = len(tokens)
for i in xrange(n_tokens):
for j in xrange(i + min_n, min(n_tokens, i + max_n) + 1):
all_ngrams.append(" ".join(tokens[i:j]))
return all_ngrams
def f7(seq):
"""
Makes a list unique
"""
seen = set()
seen_add = seen.add
return [x for x in seq if x not in seen and not seen_add(x)]
def count_list(the_list):
"""
Generates a count of the number of times each unique item appears in a list
"""
count = the_list.count
result = [(item, count(item)) for item in set(the_list)]
result.sort()
return result
def regenerate_good_tokens(string):
"""
Given an input string, part of speech tags the string, then generates a list of
ngrams that appear in the string.
Used to define grammatically correct part of speech tag sequences.
Returns a list of part of speech tag sequences.
"""
toks = nltk.word_tokenize(string)
pos_string = nltk.pos_tag(toks)
pos_seq = [tag[1] for tag in pos_string]
pos_ngrams = ngrams(pos_seq, 2, 4)
sel_pos_ngrams = f7(pos_ngrams)
return sel_pos_ngrams
def get_vocab(text, score, max_feats=750, max_feats2=200):
"""
Uses a fisher test to find words that are significant in that they separate
high scoring essays from low scoring essays.
text is a list of input essays.
score is a list of scores, with score[n] corresponding to text[n]
max_feats is the maximum number of features to consider in the first pass
max_feats2 is the maximum number of features to consider in the second (final) pass
Returns a list of words that constitute the significant vocabulary
"""
dict = CountVectorizer(ngram_range=(1,2), max_features=max_feats)
dict_mat = dict.fit_transform(text)
set_score = numpy.asarray(score, dtype=numpy.int)
med_score = numpy.median(set_score)
new_score = set_score
if(med_score == 0):
med_score = 1
new_score[set_score < med_score] = 0
new_score[set_score >= med_score] = 1
fish_vals = []
for col_num in range(0, dict_mat.shape[1]):
loop_vec = dict_mat.getcol(col_num).toarray()
good_loop_vec = loop_vec[new_score == 1]
bad_loop_vec = loop_vec[new_score == 0]
good_loop_present = len(good_loop_vec[good_loop_vec > 0])
good_loop_missing = len(good_loop_vec[good_loop_vec == 0])
bad_loop_present = len(bad_loop_vec[bad_loop_vec > 0])
bad_loop_missing = len(bad_loop_vec[bad_loop_vec == 0])
fish_val = fisher.FishersExactTest.probability_of_table(
[[good_loop_present, bad_loop_present], [good_loop_missing, bad_loop_missing]])
fish_vals.append(fish_val)
cutoff = 1
if(len(fish_vals) > max_feats2):
cutoff = sorted(fish_vals)[max_feats2]
good_cols = numpy.asarray([num for num in range(0, dict_mat.shape[1]) if fish_vals[num] <= cutoff])
getVar = lambda searchList, ind: [searchList[i] for i in ind]
vocab = getVar(dict.get_feature_names(), good_cols)
return vocab
def edit_distance(s1, s2):
"""
Calculates string edit distance between string 1 and string 2.
Deletion, insertion, substitution, and transposition all increase edit distance.
"""
d = {}
lenstr1 = len(s1)
lenstr2 = len(s2)
for i in xrange(-1, lenstr1 + 1):
d[(i, -1)] = i + 1
for j in xrange(-1, lenstr2 + 1):
d[(-1, j)] = j + 1
for i in xrange(lenstr1):
for j in xrange(lenstr2):
if s1[i] == s2[j]:
cost = 0
else:
cost = 1
d[(i, j)] = min(
d[(i - 1, j)] + 1, # deletion
d[(i, j - 1)] + 1, # insertion
d[(i - 1, j - 1)] + cost, # substitution
)
if i and j and s1[i] == s2[j - 1] and s1[i - 1] == s2[j]:
d[(i, j)] = min(d[(i, j)], d[i - 2, j - 2] + cost) # transposition
return d[lenstr1 - 1, lenstr2 - 1]
class Error(Exception):
pass
class InputError(Error):
def __init__(self, expr, msg):
self.expr = expr
self.msg = msg
def gen_cv_preds(clf, arr, sel_score, num_chunks=3):
"""
Generates cross validated predictions using an input classifier and data.
clf is a classifier that implements that implements the fit and predict methods.
arr is the input data array (X)
sel_score is the target list (y). y[n] corresponds to X[n,:]
num_chunks is the number of cross validation folds to use
Returns an array of the predictions where prediction[n] corresponds to X[n,:]
"""
cv_len = int(math.floor(len(sel_score) / num_chunks))
chunks = []
for i in range(0, num_chunks):
range_min = i * cv_len
range_max = ((i + 1) * cv_len)
if i == num_chunks - 1:
range_max = len(sel_score)
chunks.append(range(range_min, range_max))
preds = []
set_score = numpy.asarray(sel_score, dtype=numpy.int)
chunk_vec = numpy.asarray(range(0, len(chunks)))
for i in xrange(0, len(chunks)):
loop_inds = list(
chain.from_iterable([chunks[int(z)] for z, m in enumerate(range(0, len(chunks))) if int(z) != i]))
sim_fit = clf.fit(arr[loop_inds], set_score[loop_inds])
preds.append(list(sim_fit.predict(arr[chunks[i]])))
all_preds = list(chain(*preds))
return(all_preds)
def gen_model(clf, arr, sel_score):
"""
Fits a classifier to data and a target score
clf is an input classifier that implements the fit method.
arr is a data array(X)
sel_score is the target list (y) where y[n] corresponds to X[n,:]
sim_fit is not a useful return value. Instead the clf is the useful output.
"""
set_score = numpy.asarray(sel_score, dtype=numpy.int)
sim_fit = clf.fit(arr, set_score)
return(sim_fit)
def gen_preds(clf, arr):
"""
Generates predictions on a novel data array using a fit classifier
clf is a classifier that has already been fit
arr is a data array identical in dimension to the array clf was trained on
Returns the array of predictions.
"""
if(hasattr(clf, "predict_proba")):
ret = clf.predict(arr)
# pred_score=preds.argmax(1)+min(x._score)
else:
ret = clf.predict(arr)
return ret
def calc_list_average(l):
"""
Calculates the average value of a list of numbers
Returns a float
"""
total = 0.0
for value in l:
total += value
return total / len(l)
stdev = lambda d: (sum((x - 1. * sum(d) / len(d)) ** 2 for x in d) / (1. * (len(d) - 1))) ** .5
def quadratic_weighted_kappa(rater_a, rater_b, min_rating=None, max_rating=None):
"""
Calculates kappa correlation between rater_a and rater_b.
Kappa measures how well 2 quantities vary together.
rater_a is a list of rater a scores
rater_b is a list of rater b scores
min_rating is an optional argument describing the minimum rating possible on the data set
max_rating is an optional argument describing the maximum rating possible on the data set
Returns a float corresponding to the kappa correlation
"""
assert(len(rater_a) == len(rater_b))
if min_rating is None:
min_rating = min(rater_a + rater_b)
if max_rating is None:
max_rating = max(rater_a + rater_b)
conf_mat = confusion_matrix(rater_a, rater_b,
min_rating, max_rating)
num_ratings = len(conf_mat)
num_scored_items = float(len(rater_a))
hist_rater_a = histogram(rater_a, min_rating, max_rating)
hist_rater_b = histogram(rater_b, min_rating, max_rating)
numerator = 0.0
denominator = 0.0
if(num_ratings > 1):
for i in range(num_ratings):
for j in range(num_ratings):
expected_count = (hist_rater_a[i] * hist_rater_b[j]
/ num_scored_items)
d = pow(i - j, 2.0) / pow(num_ratings - 1, 2.0)
numerator += d * conf_mat[i][j] / num_scored_items
denominator += d * expected_count / num_scored_items
return 1.0 - numerator / denominator
else:
return 1.0
def confusion_matrix(rater_a, rater_b, min_rating=None, max_rating=None):
"""
Generates a confusion matrix between rater_a and rater_b
A confusion matrix shows how often 2 values agree and disagree
See quadratic_weighted_kappa for argument descriptions
"""
assert(len(rater_a) == len(rater_b))
if min_rating is None:
min_rating = min(rater_a)
if max_rating is None:
max_rating = max(rater_a)
num_ratings = int(max_rating - min_rating + 1)
conf_mat = [[0 for i in range(num_ratings)]
for j in range(num_ratings)]
for a, b in zip(rater_a, rater_b):
conf_mat[a - min_rating][b - min_rating] += 1
return conf_mat
def histogram(ratings, min_rating=None, max_rating=None):
"""
Generates a frequency count of each rating on the scale
ratings is a list of scores
Returns a list of frequencies
"""
if min_rating is None:
min_rating = min(ratings)
if max_rating is None:
max_rating = max(ratings)
num_ratings = int(max_rating - min_rating + 1)
hist_ratings = [0 for x in range(num_ratings)]
for r in ratings:
hist_ratings[r - min_rating] += 1
return hist_ratings
def get_wordnet_syns(word):
"""
Utilize wordnet (installed with nltk) to get synonyms for words
word is the input word
returns a list of unique synonyms
"""
synonyms = []
regex = r"_"
pat = re.compile(regex)
synset = nltk.wordnet.wordnet.synsets(word)
for ss in synset:
for swords in ss.lemma_names:
synonyms.append(pat.sub(" ", swords.lower()))
synonyms = f7(synonyms)
return synonyms
def get_separator_words(toks1):
"""
Finds the words that separate a list of tokens from a background corpus
Basically this generates a list of informative/interesting words in a set
toks1 is a list of words
Returns a list of separator words
"""
tab_toks1 = nltk.FreqDist(word.lower() for word in toks1)
if(os.path.isfile(ESSAY_COR_TOKENS_PATH)):
toks2 = pickle.load(open(ESSAY_COR_TOKENS_PATH, 'rb'))
else:
essay_corpus = open(ESSAY_CORPUS_PATH).read()
essay_corpus = sub_chars(essay_corpus)
toks2 = nltk.FreqDist(word.lower() for word in nltk.word_tokenize(essay_corpus))
pickle.dump(toks2, open(ESSAY_COR_TOKENS_PATH, 'wb'))
sep_words = []
for word in tab_toks1.keys():
tok1_present = tab_toks1[word]
if(tok1_present > 2):
tok1_total = tab_toks1._N
tok2_present = toks2[word]
tok2_total = toks2._N
fish_val = fisher.FishersExactTest.probability_of_table(
[[tok1_present, tok2_present], [tok1_total, tok2_total]])
if(fish_val < .001 and tok1_present / float(tok1_total) > (tok2_present / float(tok2_total)) * 2):
sep_words.append(word)
sep_words = [w for w in sep_words if not w in nltk.corpus.stopwords.words("english") and len(w) > 5]
return sep_words
def encode_plus(s):
"""
Literally encodes the plus sign
input is a string
returns the string with plus signs encoded
"""
regex = r"\+"
pat = re.compile(regex)
return pat.sub("%2B", s)
def getMedian(numericValues):
"""
Gets the median of a list of values
Returns a float/int
"""
theValues = sorted(numericValues)
if len(theValues) % 2 == 1:
return theValues[(len(theValues) + 1) / 2 - 1]
else:
lower = theValues[len(theValues) / 2 - 1]
upper = theValues[len(theValues) / 2]
return (float(lower + upper)) / 2
\ No newline at end of file
requirements.txt 0 → 100644
boto==2.6.0
coverage==3.5.3
dogstatsd-python==0.2
lxml==3.0.1
mock==0.8.0
nltk==2.0.3
nose==1.2.1
scipy==0.11.0
path.py==3.0
pip
pylint==0.26.0
pytz==2012h
scikit-learn==0.12.1
sphinx
django-sphinx-autodoc
setup.py 0 → 100644
from setuptools import setup, find_packages
with open('requirements.txt') as f:
required = f.read().splitlines()
setup(
name = "machine-learning",
version = "0.1",
packages=['machine_learning', 'machine_learning.external_code', 'machine_learning.data', 'machine_learning.external_code.fisher'],
package_data = {
'': ['*.txt', '*.rst', '*.p'],
},
author = "Vik Paruchuri",
author_email = "vik@edx.org",
description = "Machine learning based automated text classification for essay scoring.",
license = "AGPL",
keywords = "ml machine learning nlp essay education",
url = "https://github.com/edx/machine-learning",
include_package_data = True,
)
\ No newline at end of file
Markdown is supported
0% or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment