Semantic-based Component Retrieval System Realization

Component-Based Software Development (CBSD) technology promotes assembling applications of certain domain by means of reusing software components rather than doing everything from the very beginning. CBSD takes advantage of existing assets to reduce efforts and raise productivity. Software components are fully tested so that they have high quality. However, when there are a great number of components in repository, it' s hard to retrieve components which meet the specific reuse need. Thus the description and retrieval of reusable components are always hot issues in CBSD.Traditional keyword-based, signature method or facet-based component retrieval have low precision and recall due to the lack of domain knowledge. Many researchers propose semantic-based component retrieval method. Semantic-based retrieval method takes the semantic of user query and the knowledge of certain domain into account during every step of retrieval process, which results in higher precision and recall. There' s one major problem in semantic-based component retrieval process is the discrepancy between nature language which users are familiar with and formal semantic descriptions of components.In this paper, an interactive query generation method is proposed, which allows consumer use natural language. The method uses concept identification process to map users' nature language to ontology descriptions of components, then extends initial queries by semantic constraints and revises queries according to retrieved results. Concept identification makes full use of user query context (mainly composed of predicate, object, attribute, and adverbial). Concept identification and query revision use conversational process to interact with users. This method extracts implicit semantics in user queries so that it. leads to improve recall and precision rates, reduce retrieval effort in component retrieval.This paper researches on each step during component-based component retrieval, and implements the whole component retrieval system. We demonstrate a real query process and analyze the precision, recall and time performance of this system in order to verify our theory.