Design And Implemenation Of Question Parsing Subsystem Of The Financial Question Answering System

Recent years, more and more investors are stepping into the area of finance byinvesting stocks or funds. In order to get the rich profit through rational investing,they may eager to get more information about the market. However, both thesoftware used to speculate in the stock market and the traditional search engine cannot meet the requirements of searching efficiently and precisely nowadays. For thisreason, the financial question answering system comes and it can help users searchthe related information by entering the question in a format of natural language intothe search box.The financial question answering system is an automatic system just in therestricted area of finance, while the parsing of the questions which users search isthe key point of implementing the system. This dissertation develops the questionparsing subsystem which mainly consists of four parts: the analysis of configuredocument, the analysis of words, syntactic analysis and semantic analysis. Theanalysis of configure document in XML format is the base of question parsing. Itmainly describes how the index and the content of syntactic pattern defined andaccessed, besides it also describe the representation of the result of configuredocument analysis in the memory. The analysis of words and expression mainlydescribes how to combine the Chinese expressions of number, time and other namedentities, use the regular expression matching and other methods to construct thesemantic node list which consists of the generation, merging and analysis of number,time, indexs, key words, the abbreviation of stocks and other various kinds of namedentity nodes from the result of word segmentation while retaining the ambiguity ofwords. Syntactic analysis which connects the single node and semantic tree is thekey step of constructing the subsystem of question parsing. This part mainlydescribes how to use syntactic analysis methods based on semantic, implement thespecific process of syntactic pattern matching which is started by the predicate head,this process used the related strategies of syntactic ambiguity resolution andcomplementation for ellipsis. Semantic analysis is the further process of the resultscoming from syntactic analysis. It further understands the purpose of userssearching by index disambiguation, property of index binding, property of indexcomplementing. The semantic tree constructed by this part is the output result ofquestion parsing, and it is also the semantic representation of the purpose of userssearching, the effective input of structural searching system, implemented thesemantic search truly. Meanwhile, in the design and implementations, using the Spring framework to simplify the control of parsing steps.The system runs well and robust on-line which reaches the analysis correct rateof92%and the average of response time of the system is just around70ms, whichassume the task of the financial answering system well and meet the requirements ofsearching efficiently and precisely.