Boosting performance of gene mention tagging system by hybrid methods

► We have adopted three hybrid methods to complete the gene mention tagging task. ► The best hybrid method achieves of 88.42% F-score on the BioCreative II test corpus. ► The results of the hybrid methods constantly outperform those of single classifiers. ► The stacking algorithm is more effective than voting and simple set operations. ► Hybrid methods' performance relies on single models' property, number and diversity.