Hey Kenneth and Akshay,
I have installed conceptnet and all its dependencies on Windows
(cygwin), Windows (native) and Ubuntu and been using the REST client
all day and found it a bit slow. Is it only because of the server or
is it because you are using sql-lite to store it?
Heres how I did it:
Software versions I installed
Windows(native install)
Divisi 0.5 ( Library for reasoning by analgy and association over
semantic networks )
-- install python 2.5.4
http://www.python.org/download/ and add
python to the path in environment variables
-- install easy_install ( setup tools - makes your life much
easier ) setuptools-0.6c9.win32-py2.5.exe (md5)
http://pypi.python.org/pypi/setuptools
-- install min32gw (Automated MinGW Installer MinGW 5.1.4) for a
proper GNU Compiler and add it to the path in environment variables
http://sourceforge.net/project/showfiles.php?group_id=2435
-- install numpy
command C:\easy_install numpy ( make sure you have a proper
fortran compiler that numpy requires )
-- install ipython ( optional )
-- download divisi 0.5 from
http://divisi.media.mit.edu/dist/ and
change the svd folder with the one supplied by akshay in the first
post
and extract it and navigate to the folder containing setup.py and
type in
setup.py build -c mingw32 install
ConceptNet
-- install pystemmer 1.015
download pystemmer and extract it like and navigate to the folder
containing setup.py and type in
setup.py build -c mingw32 install
http://pypi.python.org/pypi/PyStemmer/1.0.1
--install conceptnet by typing in
easy_install conceptnet
Test test.py file to test if installation was all correct
Make sure you download sqlite.db of the database and the db_config
file .. make sure they are in the same folder
python test.py
if you get asserations, then you have successfully installed it .
............
Windows(Using cygwin)
1. install Cygwin with these options
-- autoconf
-- gcc compiler
-- gnu fortran compiler
-- libtool
-- make
-- python dev
-- python numpy 1.2
After installation great the group and user passwd as mentioned by
cygwin file
2. install easy_install ( setup tools - makes your life much
easier ) setuptools-0.6c9-py2.4.egg(md5)
sh setuptools-0.6c9-py2.4.egg
http://pypi.python.org/pypi/setuptools
3. install numpy
command C:\easy_install numpy ( make sure you have a proper
fortran compiler that numpy requires )
4. install ipython ( optional )
5. download divisi 0.5 from
http://divisi.media.mit.edu/dist/
easy_install divisi
ConceptNet
--install conceptnet by typing in
easy_install conceptnet
and there you go now
Test test.py file to test if installation was all correct
Make sure you download sqlite.db of the database and the db_config
file .. make sure they are in the same folder
python test.py
if you get asserations, then you have successfully installed it .
test.py
----------------------------------------------------------------
from csc.conceptnet4.models import Concept
print '-----dog assertions -----'
dog = Concept.get('dog', 'en')
for fwd in dog.get_assertions_forward():
print fwd
---------------------------------------------------------------
On Jun 2, 8:43 pm, Kenneth Arnold <
kenneth.arn...@gmail.com> wrote:
> How did you resolve your Fortran issues? (Divisi only depends on
> numpy's Fortran components for SVDs of dense tensors, I think.)
>
> Great to hear another masters project about common sense! Do keep us
> (i.e, this list, or
concept...@media.mit.edu if you don't want to
> advertise) informed about your project.
>
> btw, if you're just experimenting, you can play with our basic
> REST-ish client. See the new API docs page
> (
http://groups.google.com/group/conceptnet-users/web/api-documentation...)
> for a link to it. The functionality is limited to what has occurred to
> us to expose; if you want something new, it's pretty easy to add. We
> also fully support the sqlite dumps, which are really easy to get set
> up. I recommend using the latest ConceptNet and divisi branches on
> Launchpad if you can get bzr to work, because a lot of stuff is
> fixed/improved since the last release.
>
> -Ken
>