A fast document classification algorithm for gene symbol disambiguation in the BITOLA literature-based discovery support system.

Gene symbol disambiguation is an important problem for biomedical text mining systems. When detecting gene symbols in MEDLINE citations one of the biggest challenges is the fact that many gene symbols also denote other, more general biomedical concepts (e.g. CT, MR). Our approach to this problem is first to classify the citations into genetic and non-genetic domains and then to detect gene symbols only in the genetic domain. We used ontological information provided by Medical Subject Headings (MeSH) for this classification task. The proposed algorithm is fast and is able to process the full MEDLINE distribution in a few hours. It achieves predictive accuracy of 0.91. The algorithm is currently implemented in the BITOLA literature-based discovery support system (http://www.mf.uni-lj.si/bitola/).