A tool for the identification of chemical entities (CheNER-BioC)

The CHEMDNER task is a Named Entity Recognition (NER) challenge that aims at labeling different types of chemical names in biomedical text. We approach this challenge by proposing a hybrid approach that combines linear Conditional Random Fields (CRF) together with regular expression taggers and dictionary usage, followed by a post-processing step to tag those chemical names in a corpus of Medline abstracts. Our system performs with an F-score of 72.08 and 70.62% on the development and sample sets, respectively, for the CDI subtask. For the CEM subtask the performance increases to 72.61% and 73.68% on the development and sample sets, respectively.