Research On Geographical Knowledge Question Solving Method Based On Semantic Logic Reasoning

At traditional question answering system based on retrieving text library to generate the question answer, these methods are not applicable when processing logical reasoning needed questions. To solving these questions, we need to use non-retrieving methods. In this paper, we introduced a QA System dedicated to logical reasoning on geography. The system could accept some logical reasoning geographical questions and show results. The form of question could be a simple question or a complex calculation questions of senior middle school in geography. We designed logic description language, which used for expressing geographical area logical relationship. Implement specially appointed keywords lexical label, and translate question to logic description form, which can be recognized by reasoning engine, it will make QA system have stronger capacity in definitive environment of geographical knowledge.The contents of this paper mainly include the following aspects:Build geographical logic reasoning knowledge base. We analyzed particularity and approaches of geographic knowledge and implement description language on logic geography. The description language could Guide reasoning engine to reason knowledge of geography. Due to the particularities of geography, In spatial reasoning, we developed classification of geographic entity on methods of design reasoning engine. Each type of geographic entities was described by its physical properties and related knowledge bases were prepared.Formalize the question. When QA System was dealing with questions, grammar structure were analyzed by extracting the question keyword and marking them lexical labels. In questions, Semantic logic were firstly achieved by grammar structure. Then questions were translated into logic description form that could be distinguished by reasoning engine.Geographical knowledge logical reasoning. At the same time, in this project, reasoning space was constructed by reasoning engine based on geographic entities in question, and by the geographical knowledge logic to build the corresponding "linkage reachable space", It may make entity attribute linkage modified strictly. In reasoning process, methods and tactics adopted by reasoning engine were also be introduced.In this paper, we use 58 geography examination questions of senior middle school and 48 general reasoning geography questions. Evaluate Mephisto QA System with Baidu QA system and Wolfram Alpha. In QA System, recognition rate for questions was 0.698. Precision was 0.770. Recall was 0.537. F1 was 0.633. This system performance was better than Baidu QA system and International famous reasoning system Wolfram Alpha in the same test set.