Journal of Harbin University of Science and Technology (Oct 2019)
A Method of Word Sense Disambiguation with Restricted Boltzmann Machine
Abstract
For polysemy phenomenon in Chinese, Restricted Boltzmann Machine (RBM) is adopted to determine the true meaning of ambiguous vocabulary where linguistic knowledge in context is used Word form, part of speech and semantic categories in four left and right lexical units adjacent to an ambiguous word are selected as disambiguation features At the same time, RBM is used to construct word sense disambiguation (WSD) model Training corpus in SemEval-2007: Task#5 and semantic annotation corpus in Harbin Institute of Technology are used to optimize parameters of RBM Test corpus in SemEval-2007: Task#5 is used to evaluate WSD model Experimental results show that compared with Bayesian word sense disambiguation classifier, disambiguation accuracy of WSD method with RBM is improved
Keywords