文章基本信息

标题：COMPARISON OF DISTRIBUTIONAL SEMANTIC MODELS FOR RECOGNIZING TEXTUAL ENTAILMENT
本地全文：下载
作者：YUDI WIBISONO ; DWI HENDRATMO WIDYANTORO ; NUR ULFA MAULIDEVI 等
期刊名称：Journal of Theoretical and Applied Information Technology
印刷版ISSN：1992-8645
电子版ISSN：1817-3195
出版年度：2016
卷号：93
期号：2
出版社：Journal of Theoretical and Applied
摘要：Recognizing Textual Entailment (RTE) is an important task in many natural language processing. In this paper we investigate the effectiveness of distributional semantic model (DSM) in RTE task. Word2Vec and GloVe are recent methods that suitable for learning DSM using a large corpus and vocabulary. Seven distributional semantic models (DSM) generated using Word2Vec and GloVe were compared to get the best performer for RTE. To our knowledge, this paper is the first study of various DSM on RTE. We found that DSM improves entailment accuracy, with the best DSM is GloVe trained with 42 billion tokens taken from Common Crawl corpus. We also found the size of vocabulary size in DSM does not guarantee higher accuracy.
关键词：Recognizing Textual Entailment; Distributional Semantic Model; Text Alignment