一种基于知网语义相似度计算的应用研究

An application research based on computing of semantic similarities of HowNet

随着常见问题集(FAQ)的积累,问句数量日益增大,逐页浏览的知识获取方式很难满足用户的实际需求,也浪费用户大量的精力和时间,而基于FAQ的自动问答系统是解决该问题的一种有效途径。本文设计了QuickFaq自动问答系统框架,采用基于知网的语义相似度计算模块,找出相似问句并返回答案。实验表明,该方法提高了问答系统的召回率和不匹配率。

Following the accumulation of FAQ,the number of question sentence increased quickly. The FAQ system hardly satisfied the actual demands of users and wasted a lot of time. The automatic QA system based on FAQ can solve this problem efficiently. We propose QuickFaq automatic QA system based on computing of semantic similarities of HowNet. The system which is proposed can find similar sentence and return answers. Experimental results show that this method improves the recall rate and mismatching rate in QA system.