Get our free extension to see links to code for papers anywhere online!

Chrome logo Add to Chrome

Firefox logo Add to Firefox


Sense Vocabulary Compression through the Semantic Knowledge of WordNet for Neural Word Sense Disambiguation

May 14, 2019
Loïc Vial, Benjamin Lecouteux, Didier Schwab



In this article, we tackle the issue of the limited quantity of manually sense annotated corpora for the task of word sense disambiguation, by exploiting the semantic relationships between senses such as synonymy, hypernymy and hyponymy, in order to compress the sense vocabulary of Princeton WordNet, and thus reduce the number of different sense tags that must be observed to disambiguate all words of the lexical database. We propose two different methods that greatly reduces the size of neural WSD models, with the benefit of improving their coverage without additional training data, and without impacting their precision. In addition to our method, we present a new WSD system which relies on pre-trained BERT word vectors in order to achieve results that significantly outperform the state of the art on all WSD evaluation tasks.

* Accepted at the 10th Global WordNet Conference - GWC 2019. arXiv admin note: text overlap with arXiv:1811.00960 


Share this with someone who'll enjoy it:

   Access Paper Source



Share this with someone who'll enjoy it: