![]() |
ИСТИНА |
Войти в систему Регистрация |
ИСТИНА ЦЭМИ РАН |
||
We present a system that allows exploration of word senses induced from unlabeled text corpora for the purposes of linguistic research. The system performs word sense induction by clustering lexical substitutes generated by a neural language model for an ambiguous word. It provides convenient user interface to study clusters of the word usages, often corresponding to different senses of this word. The system allows exploring word senses expressed in a given corpus, as well as the difference between word senses in two corpora. This may be useful for studying lexical semantic change over time or between domains.