Automatic QA System For Software Testing

In the digital information age, people show a never stronger needs for the information,especially after the internet wave swept all over the world,this needs is to the extreme.The appearing of the massive internet provides us with the endless learning resource.With the development and popularization of the search engine thchnology,as long as we input the key words,we can easily find what we want.However,the constantly increasing of the online information gives us the convenience,it also brings the trouble for us. The search engine returned to us only the large number of relevant web pages rather than accurate answers.How can we exactly find what we want from the vast web pages is a huge challenge for the search engine.For overcoming the shortcomings of the search engines,many domestic and foreign scholars devote to exploring new search technique, by this way.the automatic question answering system came into being.It not only allows users to ask questions in natural language sentences,but also can returnn a simple and accurate answer to users,rather than a lot of relevant web pages.This paper designed a Chinese automatic question answering system for the field of sofeware testing,it mainly contains three parts: Chinese word segmentation module, problems'comprehention module,information retrieval module. The Chinese word segmentation is the key technology to the Chinese question answering system, the quality of the segmentation results directly affects to the degree of the users'question comprehension.In this module,the paper raised an improvement method of ambiguous fragments discovery. The problems'comprehention is to understand the purpose of users'questions,this system adopt two ways to deal with users'questions.The first step,take users'questions to match the database questions to find the answer as soon as possible. Based on BM string matching algorithm,this article made use of the continuous series of characters are not in the model string need not be compared to change the comparing order,and decrease the comparing time.The results indicated that the new algorithm imporved the efficiency of BM string matching algorithm.The second step,for question of the first step matching failure,the article uses own search engine which the author developed to search relevant questions from the software testing forum to find the appropriate answer.The task of the information retrieval is to use keywords which extracted from the front to search relevant documents from the document library,and return the most relevant documents.