Corpus-driven development of a large-scale lexicon of lexical semantics structures

点击翻译按钮获取中文摘要

We aim at the automatic analysis of semantic structures of natural language texts, and developed a large-scale lexicon of lexical semantics, which represents meanings a word inherently owns. Conventional lexicons cannot assign semantic structures to unknown words (words that do not exist in a lexicon). Therefore, this research proposed a method for representing word-to-semantics mappings as a probabilistic model. We empirically demonstrated that we can obtain a probabilistic model of a large-scale lexicon by using an existing lexicon as training data and statistics extracted from large texts as features.