| In natural language, the polysemy phenomenon is universal. So it is very difficult for the machine to process natural language. Word sense disambiguation(WSD) is an important problem in natural language processing. The task of word sense disambiguation is to determine the exact meaning of an ambiguous words in a specific context. It has important roles in information retrieval, machine translation, text categorization and topic tracking. Until now, word sense disambiguation has been a complicated problem for researchers of computing language.Supervised word sense disambiguation based on machine learning theory is researched in this paper. This method is a mainstream methodology in the field of word sense disambiguation. This method has strong extensibility and flexibility. It can also deal with different languages and be suitable for the development of language. In this paper, research contents are divided into following parts:Firstly, the development of word sense disambiguation is elaborated. The representative methods are listed. Authoritative evaluation system of word sense disambiguation is introduced. The unresolved problems affecting WSD are expounded.Secondly, corpus and dictionary which is used in process of disambiguation are introduced. Analytic process of corpus and performance of the corpus are intuoduced. Extraction and selection of disambiguation features are studied. Two mapping methods are studied for evaluating the correctness of the classifier. After analyzing the semantic classification of Tongyici Cilin, two layers of semantic code are chosen as disambiguation feature. Hidden Markov model(HMM) is optimized. The classifier of word sense disambiguation is built.Finally, the process of disambiguation is taken as a decoding problem of hidden Markov model. A disambiguation algorithm based on Viterbi is proposed. The calculation process of the algorithm is explained in detail. Two experiments are designed for evaluating the method which is proposed in this paper. |
PDF Full Text Request