Contextualized Word Embeddings for Word Sense Induction for Russian Language

Word Sense Induction (WSI) is the process of automatically discovering multiple senses or meanings of a word. WSI task can be described as grouping contexts of a given word in accordance with its senses which are not provided beforehand. Modern WSI systems are given small text fragments only and should cluster them into some unidentified number of clusters. In the present work contextualized word embeddings such as provided by ELMo or BERT are applied to the WSI task for the Russian language. Our hypothesis is that since embeddings produced by these models depend on context, we presumed that they could be able to capture the particular word meaning used in a particular sentence. We have tested it on the 3 datasets available for Russian language Word Sense Induction task. We created a WSI system for Russian language based on clustering context-dependent word embeddings constructed by pre-trained language models. For coarse-grained wiki-wiki dataset and more fine-grained bts-rnc dataset from RUSSE'18 task our best model performed results comparable to state-of-the-art (ARI=0.789 and ARI=0.3032, respectively). It has also outperformed another approaches based on context clustering, that use context-independent word representations. Moreover, our experiment proved that using BERT as word embedder for homonyms sense induction is more efficient than exploiting BERT —— in a more standard way —— as language model.