Abstract

In the gene normalization task, a rule-based approach has certain advantages including the fact that no gold standard is likely to contain all the genes that need to be considered. We have developed a rule-based algorithm that includes pattern matching for gene symbols and an approximate term searching technique for gene names. The algorithm performs confidence estimation by appropriately weighting measures of uniqueness, inverse distance, and coverage. An F-measure of 0.753 has been achieved, using nominal confidence-measure weights.



Close Window