A new answer analysis approach for Chinese yes-no question

  • Authors:

  • Lei Xu;Yong Ren;Wangyi Ye

  • Affiliations:

  • Department of Electronic Engineering, Tsinghua University, Beijing, China;Department of Electronic Engineering, Tsinghua University, Beijing, China;Department of Electronic Engineering, Tsinghua University, Beijing, China

  • Venue:
  • AICI'11 Proceedings of the Third international conference on Artificial intelligence and computational intelligence - Volume Part III
  • Year:
  • 2011

Quantified Score

Hi-index 0.00

Visualization

Abstract

The accumulated question-answer archives in community-based question answering (CQA) services have produced a valuable repository for knowledge discovery. The yes-no question is a common type of question which has not been well studied in previous work. This paper proposed a novel approach to analyze answers to Chinese yes-no questions in CQA services. The analysis task was innovatively formed as a sentence selection problem: sentences that best expressed the answerer's opinion were selected. First, a conditional random field (CRF) based annotation model was proposed to split the question into several segments. Then a new score function, which combined position information of the sentence and segmentation information of the question, was designed to score all sentences of one answer. Experiment results validated the effectiveness of the CRF-based question segmentation method. The proposed answer scoring function was also proved to be more appropriate than other approaches with respect to satisfying user's information need.