Mining Knowledge from Corpora: an Application to Retrieval and Indexing

The present work aims at discovering new associations between medical concepts to be exploited as input in retrieval and indexing. Material and Methods: Association rules method is applied to documents. The process is carried out on three major document categories referring to e-health information consumers: health professionals, students and lay people. Association rules evaluation is founded on statistical measures combined with domain knowledge. Results: Association rules represent existing relations between medical concepts (60.62%) and new knowledge (54.21%). Based on observations, 463 expert rules are defined by medical librarians for retrieval and indexing. Conclusions: Association rules bear out existing relations, produce new knowledge and support users and indexers in document retrieval and indexing. Keywords. Data Analysis; Indexing; Terminology; Data Mining; MeSH.

• Type: Conference
• Year: 2008
• Where: MIE
• Authors: Lina Fatima Soualmia, Badisse Dahamna, Stéfan Jacques Darmoni
• Keywords: Association Rules, Association Rules Method, Healthcare, Medical Concepts, MIE 2008