Similarity computation of Chinese question based on chunk

Zheng Tao Yu*, Lei Hu, Li Huang, Jing Hui Deng, Shi Ping Tang

*Corresponding author for this work

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

4 Citations (Scopus)

Abstract

The currently similarity computation methods of Chinese sentence and their shortcomings are analyzed at first. According to the characteristic of the Chinese question sentence, Chinese question general chunk and special chunk are defined, and then a similarity computation method of Chinese question based on chunk is proposed. In this method, the semantic similarity of words is computed on the basis of HowNet. General chunk is recognized by chunk parsing theory and HMM learning method, and special chunk is retrieved with some heuristic rule or SVM learning methods, then the similarity of each chunk in the two question sentences is computed separately, then the similarity computation of question sentence is realized, which is based on chunk similarity. Finally, the experiment result of question similarity computation method shows that the method proposed in the paper gets a better performance than the others.

Original languageEnglish
Title of host publicationProceedings of the 2006 International Conference on Machine Learning and Cybernetics
Pages17-22
Number of pages6
DOIs
Publication statusPublished - 2006
Event2006 International Conference on Machine Learning and Cybernetics - Dalian, China
Duration: 13 Aug 200616 Aug 2006

Publication series

NameProceedings of the 2006 International Conference on Machine Learning and Cybernetics
Volume2006

Conference

Conference2006 International Conference on Machine Learning and Cybernetics
Country/TerritoryChina
CityDalian
Period13/08/0616/08/06

Keywords

  • Chunk similarity
  • General chunk
  • Sentence similarity
  • Special chunk
  • Word similarity

Fingerprint

Dive into the research topics of 'Similarity computation of Chinese question based on chunk'. Together they form a unique fingerprint.

Cite this