An open-domain question recommended technology for large scale question answering pair corpus

作者：Duan Liguo^*; Chen Junjie

来源：Journal of Computational Information Systems, 2011, 7(9): 3291-3298.

• Accuracy rate
• High-accuracy
• Information Extraction
• Question Answering
• Question answering systems
• Semantic distance
• Semantic similarity
• Time complexity

The purpose of question recommended technology is to recommend question and answer of the highest similarity in Question Answering(QA) pair corpus to users. This paper introduces large-scale QA pair corpus constructed on Baidu ZhiDao platform and puts forward a two-phase-similarity-calculation method, which takes both time complexity and accuracy into account. With this method, question similarity is ranked first using the less time-consuming TF/IDF, and then the top-ranked questions are calculated and ranked again using high-accuracy semantic similarity method, and at last the question with the highest similarity in QA pair corpus is recommended to users, as well as its corresponding answer. We downloaded 16800 web pages in the experiments, retrieved 11678 pairs successfully with an accuracy rate of 86.14% and obtained preferable effects.

• 出版日期2011
• 单位太原理工大学