Questions Parsing Subsystem Of Financial Q&A System

In recent years, with the subsequent weakness of the keyword search products,the related products of semantic understanding have continued to emerge, and thereare even some amazing products. This allows users of information search productshave higher demands: smarter, faster, more professional, and more accurate. Whichare based on the premise of query could be understand in a clearer, more in-depthway. Therefore, in the information search products, particularly in the questionanswering system, the parsing of query is very important.As the name suggests, the Finance Q&A System is an answering system ofclosed field. In the closed field (the financial field), as a structured knowledgerepository, ontology has a natural advantage in describing of the fixed semanticrelations. Therefore, my research is about the semantic parsing, which is based onthe ontology relations, of natural language questions in the financial field. Thissubsystem consists of four main parts: parsing the configuration file, preparation,semantic association, translation instructions. The first part is the configuration fileparsing subsystem. Profile information into the configuration information requiredfor follow-up parsing. The second part is the questions pre-processing part. Firstconvert the questions, and then word and extract the entry information, and finallygenerate a semantic node. The third part is the semantic node based on ontologicalrelations associate. After this step, the questions have been structured in the middleof the analytical results. The fourth part is the translation of the intermediate resultsof the semantic parsing query instructions for different data sources, as well as theoperating instructions of the query results. All instructions generated and submittedto Financial Coordinator subsystem of the question answering system queries andfollow-up operation. Now use the data source there are two: the ontology database,financial institutions, and version data services. Its corresponding query commandfor SParQL and specific XML format.As written, the regression test set of approximately8500financial andeconomic questions. Regression test set, parsing accuracy of93.55%.The averageresponse time is26milliseconds. Formally launched for all users random questions,parsing accuracy rate is more than80%.