OpenCog Unsupervised Language Learning
Switch branches/tags
Nothing to show
Clone or download
Pull request Compare This branch is 203 commits ahead, 11 commits behind opencog:master.
alexei-gl Merge pull request #144 from alexei-gl/master
Workaround for grammar-tester to handle combinatorial explosions
Latest commit cbea398 Dec 10, 2018

README.md

SingularityNET Unsupervised Language Learning project

Dependencies

  • Anaconda 3 Python 3.6
  • numpy, pandas, scipy, scikit-learn, cython
  • SparseSVD Anaconda package
  • Jupyter notebook
  • Matplotlib

Download source code

Fork https://github.com/singnet/language-learning.git to your_repo

$ git clone https://github.com/your_repo/language-learning.git

Anaconda virtual environment

We use Ubuntu 16.04 LTS and miniconda3version of Anaconda. Please check Anaconda guides

Create virtual environment

$ cd ~/language-learning
$ conda env create -f conda-env-ull.yml

ull environment includes necessary packages, matplotlib and Jupyter notebook.
You can add packages and update environment at your own risk.

Other environments:

  • conda-env-ull-cli.yml -- simplified for CLI: no Jupyter notebook, matplotlib.
  • conda-env-ull-dev.yml -- development environment with extended package set.

Update environment:

Update with new environment file from Github repository recommended:

$ cd ~/language-learning
$ git pull
$ conda env update -n ull -f conda-env-ull.yml --prune

The --prune key would force remove packages not specified in the .yml file. If you have added come packages to the environment, you would rather let them prune and add after the update. Otherwise vwrsion conflicts might occur.
You might need to reinstall Grammar Tester after environment update.

Grammar Tester

Installation

From language-learning directory run:

$ source activate ull
$ pip install .

If for some reason you are not using virtual environment or using Python 2.x along with Python 3.x make sure you run pip3 instead:

$ pip3 install .

opencog-ull package will be installed to your virtual environment. Command line scripts from src/cli-scripts are copied to /bin subdirectory in your virtual environment.

To uninstall the package type:

$ pip uninstall opencog-ull

Running command line scripts

Command line scripts (which are located in src/cli-scripts) can be run from any location. In activated virtual environment type the name of the script you need to run.

Calling library functions from within your code

If you are going to use grammar tester from within your own code see src/samples for use cases.

Jupyter notebooks

Running on a local machine (with opencog-ull istalled)

$ cd ~/language-learning
$ source activate ull
$ jupyter notebook

Check sample notebooks in /notebooks directory.

Runing Jupyter notebooks on a server

Terminal:

$ ssh -L 8000:localhost:8888 login@server.ip.address  
$ screen  
sh-4.3$ cd language-learning  
sh-4.3$ source activate ull  
(ull) sh-4.3$ jupyter notebook --no-browser --port=8888
#...
[...NotebookApp] The Jupyter Notebook is running at:
[...NotebookApp] http://localhost:8888/?token=(copy_this_token)  

Browser: http://localhost:8000/?token=(token_copied_in_the_terminal)