Font Size: a A A

The Research Of Word Segmentation Based On Statistics To The Domain And Its Application In Production Design

Posted on:2011-06-09Degree:MasterType:Thesis
Country:ChinaCandidate:H ZhouFull Text:PDF
GTID:2132360302993487Subject:Measuring and Testing Technology and Instruments
Abstract/Summary:PDF Full Text Request
This paper designs and implements a word segmentation model, face to the Test Measurement Technique domain based on the statistical word segmentation, and applied to product design process. The model relies on the whole field of Chinese language understanding system implemented in the form of a natural language description of the design requirements of the user to the computer to recognize the conceptual design requirements or design parameters of the transformation.The paper has analyzed the characteristics and the difficulty of Chinese word segmentation, and then puts forward a new word segmentation model combing the mechanical word segmentation method, corpus-based statistical segmentation method and knowledge representation according to the existing word segmentation methods. Finally, put this segmentation module to the entire Chinese understanding system. In mechanical word segmentation stage, this segmentation system has provided all possible results, and ruled out a preliminary ambiguity. In the statistical phase, the model uses corpus-based statistical method to the results of the initial segmentation for further ambiguity processing, calculating their co-occurrence degree and matching situation and to feedback to the dictionary to improve word segmentation system and the natural language understanding system of self-perfect and benign interaction. Meanwhile, combined the conceptual dependency knowledge expression method, this segmentation module carries on the result of reasonably priority. Finally, the combination of the actual use of the word model is applied to product design of user needs analysis.
Keywords/Search Tags:Natural Language Understanding, Word Segmentation, Statistics, Corpus, Knowledge Representation
PDF Full Text Request
Related items