Word sense disambiguation is an important researchproblem in natural language processing field. For the phenomenon that a Chineseword has many senses

recurrent neural network (RNN) is used to determine truemeaning of ambiguous word with its context. Target ambiguous word is viewed ascenter and its four adjacent word units are extracted. Word

part.of.speechand semantic categories are extracted as disambiguation features. Based ondisambiguation features

recurrent neural network is used to construct wordsense disambiguation classifier. Training corpus in SemEval.2007:Task5 and semantic annotation corpus in Harbin Institute of Technology are usedto optimize parameters of RNN. Test corpus in SemEval.2007: Task#5 is applied to testword sense disambiguation classifier. Experimental results show that theproposed method can improve accuracy of word sense disambiguation.