Language identification using Random Indexing

This is the result of a project course in Uppsala University where I tested language identification using Random Indexing models.

A number of different candidate models where tested, and are to be described below. The best one was the ShortestPath model, which produced an F-score of 99.43% given the training data.

The experiments are built upon PyDSM, a library made for exploring distributional semantic models. Make sure you have this installed first.

You also need to have Expy installed. This is for saving your experiment results as well as calculating result measures.

Everything is tested on Python 3.4.

When you are sure you have the necessary libraries installed, you should be able to reproduce the results by typing:

python evaluate.py config.shortestpath

This uses the configuration found in config/shortestpath.py when running the experiment:

{'dimensionality': '2000',
 'directed': 'True',
 'num_indices': '8',
 'ordered': 'False',
 'rimodel': "<class 'models.ShortestPath'>",
 'test_path': '/Users/jimmy/dev/projects/rilangid/resources/test/reproduce/',
 'train': 'True',
 'window_size': '(100, 100)'}

After a long while, you should get a result similar to this:

Precision: 0.9968449289229714
Recall: 0.991852487135506
F-score: 0.9943254868891508