Quality retrieval of the empirical literature: a structured approach

The extent and growth of medical knowledge is tremendous. As of December 1990, the MEDLINE database had 6.5 million records in its collection, with approximately a third added during the last five years. Despite numerous innovations, retrieval technology has not kept pace with the exploding numbers of authors, articles, journals, books and conferences. A major limiting factor is that the basic mechanisms of retrieval systems almost uniformly rely on keyword representation and searching. These keywords are either assigned to texts perhaps with the assistance of special vocabularies, or appear naturally in these texts. In particular no structural or role information is preserved to connect these keywords to each other. This paper presents an alternative approach wherein role preserving, structured representations are used. This approach has the potential to increase retrieval quality.