摘要
在对比传统词频相似度模型的基础上 ,提出并实现了一种基于加权语义相似度模型的自动问答系统 .首先 ,利用语义树中词语间的距离和语义树的高度来计算词语间的语义相似度 ,然后利用词语间的语义相似度和词语的权重进一步计算用户问题与答案库中问题间的语义相似度 .基于此模型的自动问答系统能够接受用自然语言描述的问题 ,通过语义相似度的计算 ,自动地返回相关答案 .实验表明 ,本文提出的基于语义树的加权语义相似度模型与传统的词频相似度模型相比 ,准确率有明显提高 .
Contrasting with traditional similarity model of word frequency, an automatic question answering system based on weighted semantic similarity model is proposed and implemented in this paper. First, the question answering system calculates semantic similarity by using distance between words in semantic tree and the height of semantic tree, then this system uses semantic similarity between words and the weight of word to calculate semantic similarity between users' questions and questions in answer database. The automatic question answering system from this model can accept questions in natural language and return correlative answers automatically. Test result shows that compared with traditional similarity model of word frequency, the weighted semantic similarity model improves the precision obviously.
出处
《东南大学学报（自然科学版）》
EI
CAS
CSCD
北大核心
2004年第5期609-612,共4页
Journal of Southeast University：Natural Science Edition
基金
国家"十五"重大攻关资助项目 (2 0 0 1BA10 1A0 5 -0 4) .
关键词
自动问答系统
权重
语义树
语义相似度
Computer programming languages
Database systems
Semantics
