Journals Proceedings

Abstract

Software requirements that exist in natural language can easily be understood by various stakeholders. However, when it comes to extracting common requirements from the natural language requirement documents for reuse, manual extraction process can be arduous, expensive, and very error-prone on the results. In this paper, we describe a process of identifying similar requirement documents for reuse in Software Product Lines. Online product reviews were extracted and used as the input mainly due to the scarcity of publicly available requirement documents. Latent Semantic Analysis technique from Information Retrieval was used to identify similar requirement documents and filter out the unrelated ones after the text has been pre-processed. Similar documents were then clustered together by using K-means and Hierarchical Agglomerative Clustering algorithm. As a result, the output from the clustering process will be used to recommend group of related requirement documents to be used in requirement engineering phase for a SPL development