Mining Patents with tmChem, GNormPlus and an Ensemble of Open Systems

The significant amount of medicinal chemistry information contained in patents make them an attractive target for text mining. The CHEMDNER task at BioCreative V focused on information extraction from patents. This manuscript describes our submissions to the CEMP (chemical named entity recognition) and GPRO (gene and related object identification) subtasks. Our CEMP submission is an ensemble of five open systems, including both versions of tmChem, our previous work on chemical named entity recognition. Their output is combined using a machine learning classification approach. Our CEMP system obtained 0.8752 precision and 0.9129 recall, for 0.8937 f-score. Our submission to the GPRO task is an extension of GNormPlus, our previous work for gene and protein named entity recognition. Our GPRO system obtained a performance of 0.8143 precision and 0.8141 recall for 0.8137 f-score. Both submissions achieved the highest performance in their respective tasks.