摘要：The paper presents a flexible system for extracting features and creating training and test examples for solving the all-words sense disambiguation (WSD) task. The system allows integrating word and sense embeddings as part of an example description. The system possesses two unique features distinguishing it from all similar WSD systems—the ability to construct a special compressed representation for word embeddings and the ability to construct training and test sets of examples with different data granularity. The first feature allows generation of data sets with quite small dimensionality, which can be used for training highly accurate classifiers of different types. The second feature allows generating sets of examples that can be used for training classifiers specialized in disambiguating a concrete word, words belonging to the same part-of-speech (POS) category or all open class words. Intensive experimentation has shown that classifiers trained on examples created by the system outperform the standard baselines for measuring the behaviour of all-words WSD classifiers.
关键词：word sense disambiguation; word embedding; classification; neural networks; random forest; deep forest; JRip word sense disambiguation ; word embedding ; classification ; neural networks ; random forest ; deep forest ; JRip
