一种基于形式概念分析的问答系统答案抽取的研究

An approach to extract answer in question answering based on FCA

为了在问答系统中尝试回答更加复杂的问题,有必要存在一种原理性的方法来动态产生不同的问答策略。提出在问答系统中使用形式概念分析(FCA)来抽取答案。在抽取过程中,首先在常问问题集(FAQs)中寻找已经存在的问题,如果在FAQs中的答案不满足用户的需求,再通过搜索引擎获取相关的文档。接着利用这些文档中前N个文档构建概念格,进而采用概念匹配在格中抽取答案。对于不同的问题,使用不同的策略进行匹配和抽取答案。

As the number of available processing options continues to grow and our question answering （QA） systems attempt to answer more complicated questions, it is essential to have a principled method for dynamically generating different QA strategies. A question answering method was proposed using formal concept analysis （FCA） to extract answers. In the process of extraction, we firstly searched the frequently asked questions （ FAQs）. If the answers in the FAQs cannot satisfy the user＇s requirement, we continued to retrieve the relevant documents from Web by search engine. Then, the concept lattice was built by making use of the first N documents. Thereby, the concept matching was utilized to extract answer from the lattices. Different strategies were used to match and select answers to different questions.