Data Filtering in Content Industries: The Case of Recommender Systems in the Scientific and Technical Information (STI) Sphere

Information retrieval is an activity whose aim is to enable users to obtain relevant documents, so that it meets their information needs. The corpus is made up of documents that are analyzed during the indexing phase, producing document representations that can be interpreted by a computer system. Faced with the heterogeneity of items represented by documents in our context of Scientific and Technical Information (STI). Online catalogs reflect the diversity of users and their needs, and recommendation models need to adapt to best meet their users' expectations. Given the growth of information on the Web, and the diversity of its users and their needs, the “one-size-fits-all” approach to the Web was called into question. Adaptive hypermedia systems were one solution to this problem. Thus, a recommendation system is a system for filtering and personalizing information, aiming to predict and present choices and preferences likely to interest the user, around objects (books, music, films etc.). In our research, we aim to design a recommendation system model to optimize the selection and access of research users to document resources on online catalogs and Scientific and Technical Information (STI) platforms based on machine learning techniques using a constructivist engineering method which proceeds to the elaboration of a model associated to a prototype. Keywords: Modeling; information personalization; recommendation systems; scientific and technical information; information retrieval; online catalog.