Ontology-based recommender system for COTS components

Abstract

Commercial Off-The-Shelf (COTS) components are coarse-grained software components that satisfy high-level requirements by integrating several services and offering several interfaces. They are usually used to build larger systems. The paper proposes an ontology-based recommender system for COTS components, that contributes to COTS-based development by improving COTS components identification. It combines into a single framework information retrieval technologies and knowledge about COTS components and users in order to provide the most relevant COTS components meeting users needs. The recommender system is based on (1) an ontology of COTS components, named ONTOCOTS, that describes COTS components and unifies their heterogeneous descriptions available on the Web, and (2) a user model that represents user preferences and interest domains. The proposed recommender system is broken down on two main processes. The first one is responsible for extracting information about COTS components from COTS repositories and representing it as ONTOCOTS instances. The second one is the recommendation process during which the user query is expanded using the linguistic ontology WordNet, and is used along with the user profile and the domain ontology ODP (Open Directory Project) to generate a formal query. Results list is ranked according to the satisfaction degree of user requirements and preferences. Experimentations show an amelioration in recommendations relevance by placing the relevant COTS components at the top of the recommendation list.