Context-aware information retrieval systems

Abstract

Information Retrieval (IR) became indispensable to our modern knowledge-based society. Modern information environments are becoming large and complex as well as ubiquitous, because the amount of available heterogeneous information grows exponentially each year. Almost every aspect of our lives and every profession are affected by the information available on the Internet. Indeed, we live in a search society - belief that (almost) everything is known, we just have to find the information. We search for everything the good book, the new movie in cinema, the best car, the most comfortable home, the best vacation plans, even the best search engines. Full-text search on the World Wide Web (WWW) is perhaps the most widely used IR application, this application is concerned with the processing, indexing and retrieval of huge amount of textual documents. This dissertation investigates whether the inclusion of a contextual dimension (i.e. “Content” of queries and documents and “User” in our case) in the IR process can improve the effectiveness of an IR System by better indexing the documents and computing the mappings between them and the queries more accurately. Thus, we propose a semantically enriched context-aware Information Retrieval System, based on Folksonomies or social tags so as to provide more relevant search results and cope with the traditional Information Retrieval issues that does not satisfy our modern society needs. In this regard, two effective novel indexing and query-document mapping methods are proposed and evaluated. The first method focuses on the semantic aspects of documents and queries with a semantically enriched stemming algorithm based on the well-known Porter Algorithm. The second method focuses on the trustworthiness of the social environment of a user by exploring social-bookmarking as a new indexing technique through the use of Folksonomies as a new alternative to Ontologies in knowledge representation for IR purposes.