Research And Implementation Of Method For Component Testing Information Semantic Retrieval System Based On Ontology

With the component reuse was increasingly concerned by the academic and practice community, more attention has been paid to the component testing which has become a major means of component quality assurance. Large amount of component testing information which generated from component testing process raises public awareness and increases user's demands of query day by day. The existed component testing information retrieval system mainly rely on testing management tools and stress on the testing information generated from traditional software testing which is usually for special system or particular project. These retrieval systems ignore the special characteristics of component testing information. Different users have different degree of understanding for the structural of component testing information. During the query process of component testing information, users can not express their purposes accurately and completely, so that the query results can not efficiently meet the user's query intentions. Traditional search methods based on keywords or classification concentrate on accurate matching with users'query requirements, which have not considered the importance of semantic information, and so that the query results may appear a larger number of reduplicated information or leaked information. Therefore, how to use the semantic information from component testing information to expand original query for finding the most related clusters have become an urgent problem.Aiming at the above problem, this dissertation presents a component testing information semantic retrieval method based on Ontology. Firstly, the method is to analyze the contents and features of component testing information. Then, component testing information domain Ontology (CTIO) and extract semantic information and semantic relations from it is built up. Finally, it is necessary to use the extracted semantic information and semantic relations to expand original query. The whole retrieval process includes following parts.Firstly, collecting and analyses component testing information and making use of its characteristic and semantic information to build CTIO.CTIO can be divided into two separate categories:CTIO-Knowledge Base (CTIO-KB) and CTIO-Instance Base (CTIO-IB). CTIO-KB mainly describes concepts and the relationship between the concepts. CTIO-IB mainly includes real value of entities'attributes, which have two types:text-type and numerical-type.Secondly, preprocessing users'query requirements. It is necessary to use Inverse Document Frequency (IDF) to calculate the weigh of query condition expression, and select expansion priority of condition expression by calculated weigh arranged in descending order. Then, the equivalence relationship, super class relationship and subclass relationship are used to expand class concept of query condition expression. The value of numerical attribute should use numerical similarity to expand and the value of text type should use similarity calculated by the method called text's similarity of the semantic dictionary based on statistical to expand.Thirdly, in order to guarantee that the similarity of query results and original query show in descending order, an algorithm of result sorting is proposed. Different type of condition expression should use different similarity calculation. A concept similarity computing method based on CTIO is presented for class concept similarity calculating.Finally, an experimental analysis shows that the algorithm is effective in solving component testing information retrieving. We compared with existing methods to verify the effectiveness and feasibility of our proposed algorithm. The experimental results indicate that proposed method can lead to a higher recall ratio and precision ratio, and can be applied in component testing domain effectively.