Research On Question Answering System Based On Text Pattern Reasoning

Question Answering System (QAS) is an interesting subject to research in the current domain of natural language processing (NLP), because it not only allows the user to ask questions using natural language, but also return precise answers instead of related web pages. So, compared with traditional Search Engine (SE) which depends on key word matching (for example, Google), Question Answering System (QAS) can retrieve the most precise answer to satisfy the user's demand of searching.However, current Question Answering System (QAS) is not precise enough. The reason lies in the fact that there is almost no pattern reasoning techniques in current Question Answering System (QAS) and there is almost no Question Answering System (QAS) that vastly use reasoning rules. This problem must be solved to improve the quality of Question Answering System (QAS).So, based on the research of antecedent researchers, this dissertation proposed the idea of"Pattern Reasoning (PR)", and proposed a new model of Question Answering System (QAS) with the name of"Surface Structure Pattern Reasoning (SSPR)". This model has the following characteristics:â‘ . Do little or even nothing of manual work to formalize original text but take original text as repository. Compared with traditional techniques that build repository manually, this technique will save mass manpower and material resource.â‘¡. Adopt the mechanism of logical reasoning between text sentences. Through this mechanism, Question Answering System (QAS) can gain new knowledge to answer the user's question. That's to say, based on current knowledge, through logical reasoning mechanism, Question Answering System (QAS) can gain new knowledge instead of simply searching existed answer which lies in the text. Compared with traditional techniques, this technique will greatly improve the intelligence of Question Answering System (QAS).â‘¢. Give precise answer to the user's question. Compared with traditional techniques that simply return related web pages or passages, this technique will greatly improve the preciseness of Question Answering System (QAS).