一种基于用户需求协作修正的构件检索方法

来源期刊：中南大学学报(自然科学版)2011年第2期

论文作者：钟鸣 张尧学 周悦芝 田鹏伟 翁林开

文章页码：398 - 403

关键词：软件复用；构件检索；向量；协作修正

Key words：software reuse; component retrieval; vector; collaborative revision

摘    要：针对当前主流的构件检索方法只从构件分类和索引的角度出发研究提高构件检索效率，却忽略用户理解与构件描述之间差异的现状，提出一种基于用户需求协作修正的构件检索方法。该方法采用向量对构件功能和用户需求进行统一描述，并基于向量内积相似度实现构件检索；并在此基础上引入反馈机制，根据用户的历史检索记录创建和维护用户需求协作修正模型，通过该模型来对新的用户需求进行协作修正，以此缩小用户理解与构件分类描述之间的差异，提高构件检索效率。在构件库原型系统中分别以平均检索时间和平均目标位置为标准对该方法进行性能评测。研究结果表明：在基于向量的构件检索方法中，对用户需求进行协作修正能有效提高构件检索的准确度，缩短构件检索时间。

Abstract: Current methods of component retrieval focus mainly on improving the retrieval efficiency by using advanced classification and index, but ignores the discrepancy between user understanding and component description, thus a novel method for component retrieval based on collaborative revision of user requirement was proposed, which describes both components and user requirements in the format of vectors, and calculates their inner product to get the matching degree of each component. Furthermore, a model of collaborative revision was presented to collaboratively revise user’s new requirement according to the feed-back of user retrieval history, leading to minifying the discrepancy between user understanding and component description and improving the efficiency of component retrieval. Several experiments were made on the component repository prototype by the metrics of average retrieval time and average target position. The results show that, using the user requirement of collaborative revision, the vector based component retrieval method has higher retrieval accuracy and takes shorter time to find the target component.