Knowledge Discovery Of English Modal Verb Can Based On Formal Concept Analysis

Generally speaking, ambiguity widely exists in Natural Language Processing (NLP)and word sense classification. The exact meaning of an ambiguous word in certain contextneeds to be identified. Therefore, Word Sense Disambiguation (WSD), a senseclassification research method, plays a significant role in such fields as machinetranslation, information retrieval, speech recognition and linguistics.With the development of machine leaning and corpus linguistic, researches aboutconstructing effective WSD models of English modal verbs have been carried on withabundant achievements. The studies of visualizing the hierarchical relations of the senseclassification influencing features and further discovering knowledge, however, are rare.Thus, the thesis attempts to explore the underlying relations of the sense classificationinfluencing features and tries to visualize the hierarchical relations by using FormalConcept Analysis (FCA) in the process of WSD of English modal verb can. Moreover, theuniversality and specificity usage of linguistic features are presented.Based on a1.2million words corpus, two semantic features and four syntacticfeatures are extracted. A WSD model for English modal verb can by FCA is constructed.The experiment results show that the accuracy for WSD is94.66%by using the model,while the accuracy for the model is92.66%. Therefore, the effectiveness of the WSDmodel by FCA is verified. The thesis discovers the knowledge about the hierarchicalrelations by observing the generated Hasse diagrams.Furthermore, the study realizes the visualization of the hierarchical relations for eachsense of English modal verb can by FCA as well. The contribution degree of each featureto the WSD of English modal verb can in the further experiments is also testified. Theexperiment results indicate that the semantic features have universality usage, whilesyntactic features have specificity usage for the sense classification of English modal verbcan. For sense ability and sense possibility, the syntactic sentence factivity contributesmore to the sense disambiguation compared with other selected linguistic features. Forsense permission, however, semantic features are more influential than syntactic features.The research has both theoretical significance and practical implications. On onehand, the theoretical significance is obvious for studies of English model verbs. On theother hand, the study provides another effective method for WSD of English modal verbs,which will have a practical and positive support for feature identification of machine translation.