Associative conceptual space-based information retrieval systems
In this `Information Era' with the availability of large collections of books, articles, journals, CD-ROMs, video films and so on, there exists an increasing need for intelligent information retrieval systems that enable users to find the information desired easily. Many attempts have been made to construct such retrieval systems, including the electronic ones used in libraries and including the search engines for the World Wide Web. In many cases, however, the so-called `precision' and `recall' of these systems leave much to be desired. In this paper, a new AI-based retrieval system is proposed, inspired by, among other things, the WEBSOM-algorithm. However, contrary to that approach where domain knowledge is extracted from the full text of all books, we propose a system where certain specific meta-information is automatically assembled using only the index of every document. This knowledge extraction process results into a new type of concept space, the so-called Associative Conceptual Space where the `concepts' as found in all documents are clustered using a Hebbian-type of learning algorithm. Then, each document can be characterised by comparing the concepts as occurring in it to those present in the associative conceptual space. Applying these characterisations, all documents can be clustered such that semantically similar documents lie close together on a Self-Organising Map. This map can easily be inspected by its user.
Schuemie, M.J., & van den Berg, J.. (1998). Associative conceptual space-based information retrieval systems. Retrieved from http://hdl.handle.net/1765/754