Automatic Understanding Of Natural Language Questions For Chinese Querying Bases On Ontology

As the Internet increasing and the amount of information carried by,how quickly and efficiently find the information users needs is the direction of Internet development. Storing structured data is a efficiently way to find useful information,so appears more and more Knowledge base and Ontology base.Use these massive knowledge,especially expertise knowledge of some areas,can be targeted to answer users’questions.And Knowledge usually stored as RDF triples.If we want to take advantage of these vast amounts of data,we need a professional query.Because of the user input is natural languages questions,we need to get the reasonable query after understanding the natural language questions.Understand and analysis of questions if not only the first step in answering system,but the quality of understand the sentence was intended to impact on the merits of the question answering system.The processing of the answering system generally includes three subsystems:The analysis of natural language questions subsystem,information retrieval subsystem and answer extraction subsystem.The main problem of analysis of natural language questions subsystems is analysis and classification the user input sentence and extract the user’s intention and expressing the semantic information in some form or other. Information retrieval subsystem search the answer according to the output of the analysis question subsystem,and find out the correct representation of knowledge or the range of contain the answer.The main job of answer extraction subsystem is to filter the search results and change the retrieved knowledge representation to accurate natural language in natural language generation algorithm.Then returned to the natural language answer to the user.This thesis is mainly concerned pretreatments of the question sentences and contructing the semantic query graph according to the tree of the questions’syntactic analysis. The aim of this is to change the natural language question to SPARQL queries so that machine can understand and then realize the searching of the ontology base.The content of the thesis research mainly has the following several aspects:(1) Constructing domain ontology base. This thesis built Guilin domain ontology and use the standardized describes ontology language OWL2to design it.(2) Preprocessing the natural language question that the user gives.It includes these steps: segmentation, named entity recognition and systactic analysis.And adding the words of Guilin tourism areas in the module of segmentation to improve the precision rate.Then use parser to analysis the question get parsing tree.(3) Using the results of parsing tree to constructed the graph of query semantic, and nouns constructed into node and verbs constructed into edges. The graph of query semantic is a map of describing the relationship of entity that user’s question.(4) The mapping of query semantic and domain ontology.The nodes of semantic graph maps into the entity of ontology, and edges maps into the relation of ontology, then generates query sentence of SPARQL language.When the node or the edge do not exist in the ontology base, we use the "The forest of synonyms word" to expaning the point or edge.Each node of the candidate assemble that were expanded may generate a query of SPARQL language.