Knowledge Based Information Retrieval: A Semiotic Approach

The overall objective of this study is to analyze the document retrieval process and the main information retrieval (IR) concepts from the point of view of semiotics and design retrieval mechanisms based on the findings of the semiotic analysis of the retrieval situation. Semiotics is a discipline which studies 'sign systems' and how signs are exchanged in communication. The semiotic view of IR interaction presented in this dissertation views document retrieval as a kind of human communication process taking place in a social and cultural realm. The most important result of the semiotic model developed is the explication of the distinction between the knowledge production and transfer functions of document retrieval. The consequence of this finding is the conceptualization of the retrieval process as a dynamic and complex interplay between knowledge production and transfer tasks. It is hypothesised that, in the case of knowledge production, users of retrieval systems are interested in exploring new areas of the document collection which are not a priori known. Two knowledge based systems are developed based on the Okapi probabilistic retrieval system. The purpose of the retrieval systems designed is posited, in general terms, as to suggest the users new search areas of potential interest. This is achieved by treating the Inspec thesaurus as a semantic network, and applying a heuristic spreading activation technique to generate clusters of terms linked in the Inspec thesaurus. Each cluster or batch of terms is conceived as representing a part of the general search area defined by the initial user search terms. The main design objective here is to enable the user to identify new search areas from the term information contained in the batches. Two evaluation experiments were carried out with real users who had real information needs to test whether the batches were actually effective in defining search areas related to the original user queries and whether they were useful in pointing new areas which were potentially relevant to the users. A number of hypotheses related to the retrieval effectiveness of the knowledge based systems designed were also tested in the experiments. The main findings of the experiments indicate that: • the batches were useful in representing search domains relevant to the users' queries • in many cases the batches represented new ideas or new search domains to the users • the knowledge based systems had similar retrieval effectiveness in terms of precision as the Okapi syste