基于词典的语料库词义标注 = Dictionary Based Corpus Sense Tagging

Abstract

This study is aiming to build a word sense tagged Chinese corpus. The thesis firstly probes into the effects of different linguistic disambiguation cues extracted from dictionary to practice word sense tagging. The study investigates how to use Pinyin to distinguish the homographs; how to employ part-of-speech to discriminate the ambiguities, and further examines the effect and limitation of using collocations as disambiguation cue. The study also investigates how to use word sense frequency distribution to improve baseline of automatic sense tagging. The thesis secondly analyzes the difficulties summed up in manual and automatic sense tagging. On the basis of the practice of sense tagged corpus construction, the study supports that a high precision word sense tagging procedure is realizable by using hybrid linguistic disambiguation cues; however, from the analysis of difficult parts in sense tagging, more studies in dictionary sense distinctions should be done to improve tagging consistency.