Indexing and Comparison of Multi-Dimensional Entities in a Recommender System based on Ontological Approach

Abstract

Abstract: The paper describes an application of indexing—the technology currently widely used for processing and comparing textual information—to multi-dimensional entities of knowledge domains. We propose a model for building a frame-based ontology, which contains a domain conceptual model as well as a controlled vocabulary of “base terms” used for indexing. Further, the ontology constitutes the structure for the knowledge base of the recommender system developed by us, whose task is to support human-computer interaction in web applications. The system automatically represents the interaction task being solved as a structured set of base terms, and compares it with the pre-indexed design guidelines representing practical knowledge of the domain. The interaction task context is defined by input data: 1) semi-structured attributes of target users and 2) natural-language requirements for a particular web application. The former are processed mostly via production model rules stored in the knowledge base, while the requirement text is mined for base terms from the controlled vocabulary. As a result of the comparison, the system provides a set of guidelines relevant for a particular interaction task context, seeking to save work effort of interface designers. Also, the proposed approach for indexing multi-dimensional entities can be applied in various recommender and knowledge-based systems.