WSD using the Lesk Algorithm with Word2Vec Embeddings

This is an unsupervised approach for WSD that uses pre-trained Word2Vec embeddings for computing the overlaps/similarities in the Lesk algorithm. It's evaluated on the SemCor 3.0 dataset.

Usage:

See code/main.ipynb and run it cell by cell

(This was done as a course assignment for CS626: Speech and Natural Language Processing and the Web, Autumn 2021, IIT Bombay)