| As the critical technique to achieve software reuse, the basic idea ofComponent-Based Software Development (CBSD) is creating reusable softwarecomponents, and using these components to develop new application software.The application of CBSD can cut down the cost of software developmenteffectively while improve software quality and the efficiency of softwaredevelopment. With in-depth study and practice of CBSD, the number ofcomponents is increasing, which lead to expanding the size of the componentlibrary. Provided an effective method of component retrieval for users to reusecomponents has become the critical problem which needs to be resolved quickly.Facet-based component retrieval has become extensive research and applicationof software reuse, some better methods such as XML, tree matching, ontology,and so on being applied. But, it needs to be further explored since some issuessuch as not parse the query and component matching calculation inaccurate arestill exist.In this paper, a new component description model was given on the basis offaceted classification and description. To solve the problem of slow componentretrieval caused by expansion of component library and existing encoding can'tfill requirement, a new term encoding strategy is designed. A term index iscreated based this encoding, which can preprocess the components and improvethe retrieval efficiency.By analyzing the particularity of natural language parsing in componentretrieval, as well as the existing Chinese word segmentation method, amaximum matching method of the forward verbatim is designed. Parse thequery to component terms, and find the components which contain these termsby term index.Considering the feature of facet-based component, the existing methodswhich use the three tree matching model directly to calculate the componentmatching is less accurate. A new tree matching model, contain matching is designed in this paper and the calculation way of matching cost is improvedbased on this new model. A concept of matching degree is proposed to describethe degree of the match between components and analyze component matchingin multi-angle.Finally, analyzed the recall and precision of different matching degree, andselected the better degree as the threshold. Compared with the existingcomponent retrieval methods such as space encoding and tree matching modelin the recall and precision at selected threshold, and analyzed the retrieval time,the efficiency of this method is validated. The experiments show that the methodproposed in this paper can improve the precision and efficiency of componentretrieval while keeping a higher recall. |
