摘要: 句子相似度算法是基于常问问题集的问答系统(FAQ)的关键。针对汉语中一词多义现象,提出一种改进的基于知网的词义消歧算法,确定词语在不同上下文环境的义项号,利用知网系统及义项号,使用改进的相似度计算方法进行相似度计算模块设计。结合实际应用,实现一个实际的FAQ系统。实验证明,改进的词义消歧方法提高了消歧的精度和速度。而词义消歧的引入提高了问答系统的精度和速度。
关键词:
常问问答系统,
知网系统,
词义消歧
Abstract: Sentences similarity arithmetic is the key of Frequency Asked Question(FAQ). There are many muti-sense words in Chinese. This paper proposes a new arithmetic of word sense distinguish based on HowNet, uses the advanced method to the sentences similarity and realizes a real FAQ system. Word sense distinguish uses the proposed method to do the word sense distinguish. The sentences similarity is got with HowNet, and a real FAQ system is realized. Experimental results show that the proposed method can effectively increase the accuracy of the FAQ system.
Key words:
Frequency Asked Question(FAQ),
HowNet system,
word sense distinguish
中图分类号:
李 辉;张 琦;卢湖川;杨德礼. 基于知网的中文常问问答系统[J]. 计算机工程, 2008, 34(23): 62-64,6.
LI Hui; ZHANG Qi; LU Hu-chuan; YANG De-li. Chinese Frequency Asked Questions Based on HowNet[J]. Computer Engineering, 2008, 34(23): 62-64,6.