| As we all know, ambiguity is a widely-existing phenomenon in human language. In order to get rid of it, various modern computer techniques are already applied to automatically identify the appropriate sense of an ambiguous word in a given context by taking semantic, contextual and even common knowledge beyond the text. Such kind of researches is called Word Sense Disambiguation. It, one of the most heating and hardest problems in the field of Natural Language Processing, has a decisive influence on many branches of natural language processing, such as Machine Translation, Information Retrieval, Text Categorization, etc. Currently, the methodology in the research of word sense disambiguation has got remarkable progress including the construction of knowledge base, knowledge acquirement, feature selection and learning algorithms. However, it is noted that the object of such research is still mainly focused on regular nouns and verbs. Few of them can touch those words like English modal verbs which are more ambiguous, fuzzy and sensitive to context. But in nature, the object of word sense disambiguation is language, which means it can not get full development without the help of linguistic researches. It can be implied that linguistic investigation on some phenomena and the nature of language is supposed to contribute to the advance of word sense disambiguation in depth and at the same time development of this issue will also can serve linguistic researches and release linguists from burdensome and mechanical tagging of corpora. Therefore, this research has both of theoretical and practical meaning on linguistic studies.This thesis presents a disambiguation model for English modal verb may with a method of artificial neural network which has been widely applied in various fields of natural science due to its features of self-organization, self-adaptation, etc. By learning from a corpus with a scale of half-million words, this model can achieve an accuracy of 78%.
|
PDF Full Text Request