2013-2014 Seminar Schedule

Computational Biology Branch (CBB) is the research branch of the National Center for Biotechnology Information (NCBI), National Library of Medicine (NLM), National Institutes of Health (NIH).

We hold weekly seminars by CBB members each Tuesday at 11 AM in the Building 38A B2 NCBI Library. Visitors' presentations usually take place in the same room, but are scheduled on a different day of the week. If scheduling a visitor's presentation, please don't forget to book a seminar room.

To schedule a seminar, please click on the appropriate date in the calendar.

Contact Ivan Ovcharenko with questions or if you need help scheduling a seminar.

Upcoming Seminars

Robert LeamanMay 7, 2015 at 11:00
tmChem: High performance chemical named entity recognition
“Chemicals and drugs are among the most frequently searched topics in PubMed. Text mining approaches for extracting chemical information are useful in many important areas of healthcare and biomedical research, including both pharmacovigilance and drug discovery. While named entity recognition (NER) - the task of locating mentions of entities such as chemicals - is a fundamental step in text mining systems, chemical mentions pose a particular challenge due to the great variety of named chemical structures and the various systematic and semi-systematic methods for specifying them. The performance of chemical NER systems has typically lagged behind other important entity types, such as gene products and diseases. This seminar describes tmChem, a chemical named entity recognition system prepared by extending two existing named entity recognition systems - BANNER and tmVar - to handle chemicals. Their output is combined in an ensemble approach to take advantage of their differences, which include different feature sets and tokenizations. We evaluated tmChem by participating in the CHEMDNER shared task, where its results (0.874 f-measure) were the highest of the participating teams and also higher than the performance typically reported for NER of gene products or diseases. The tmChem source code, trained models, and results on PubMed are publicly available. “

Di HuangMay 12, 2015 at 11:00
Dongyeon ChoMay 14, 2015 at 11:00