Optimizing and evaluating the MEDPILOT search engine. Boosting medical information retrieval by using a morpheme thesaurus

Autor: Markó, Kornél, Dzeyk, Waldemar
Jazyk: angličtina
Rok vydání: 2008
Předmět:
Zdroj: Journal of the European Association for Health Information and Libraries, 4(1):14-19
Popis: This article describes the implementation and evaluation of a computational linguistic approach to improve the quality of the MEDPILOT medical search engine, maintained by the German National Library of Medicine. At the core of the system lies a new type of multilingual dictionary, in which entries are equivalence classes of morphemes, i.e. semantically minimal units. Documents, as well as user queries, are mapped to these language-independent (conceptual) classes on which retrieval operations are performed. Early results of the evaluation have shown that the used language technology has many advantages in medical information retrieval. In combination with up-to-date search software of the linguistic approach leads to more and better results (i.e. relevant hits) for phenomena such as synonyms, translations and linguistic variants (inflection, derivation, word-composition, etc.). Additionally, a normalization of laymen and expert queries can be achieved. The formal structure of user queries as well as the information needs of MEDPILOT users were analyzed in detail. Results and consequences for the development of user centered design and usability improvements of the search engine will be discussed.
Databáze: OpenAIRE