Automated Metadata Generation and its Application to Biological Association Extraction

Text mining methods are used in this paper to extract associations among biological objects. Transitive association methods using metadata (MeSH terms) have the potential to discover implicit associations without relying on explicit co-occurrence of objects of interest. To avoid costly manual metadata assignment and deal with missing metadata, automated metadata generation methods are described in the paper. The association knowledge extracted using automatically generated metadata is found to be as good as that that using manually assigned metadata, in terms of precision.