基于Web的答疑系统的问答系统.doc

外文翻译--qa42：基于Web的答疑系统的问答系统--CS224N最终项目 qa42 Web-Based Question Answering System -Based Question Answering System CS224N Final Project 1 Introduction qa42 named in memory of Douglas Adams [8] is an open domain question-answering system that builds upon and extends prior work in [1] by exploiting the redundancy of the world wide web[9] qa42 returns specific answers to factoid questions rather than summaries as is done in search engines and traditional question-answering systems The current version of qa42 is geared to answer only questions for persons organizations locations dates and quantities Additional types of questions are planned for future work An outline of the process is diagrammed in Figure 1 The question to be answered is written into one or more search engine queries which are then sent to the Google search engine [10] Summaries returned by Google are scored against answer models also generated from the question Similar viable answers are clustered together and rescored based upon frequency qa42 presents the three answers that score the highest For the readers convenience examples and several tables are presented at the end of this document We now present the details 2 Question Processing 21 Parse Analysis As a first step the comtion Java class parses the question using [2] as the parser trained on data from [11] From the resultant parse tree it ascertains the pronoun type pronoun subtype main verb subject noun phrases and object noun phrases The pronoun type is defined as whowhom where when why how which what and other In general the pronoun type is taken to be the interrogative pronoun used in the question In the case where more than one interrogative pronoun appears in the question the outermost from the parse tree is used The other category is used when no such interrogative pronoun exists in the question This is commonly the case when the question is worded as an imperative An example is Name the designer of the shoe that spawned millions of