论文信息 - Named Entity Recognition with Combinations of Conditional Random Fields

Named Entity Recognition with Combinations of Conditional Random Fields

The Gene Mention task is a Named Entity Recognition (NER) task for labeling gene and gene product names in biomedical text. To deal with acceptable alternatives additionally to the gold standard, we use combinations of Conditional Random Fields (CRF) together with a normalizing tagger. This process is followed by a postprocessing step including an acronym disambiguation based on Latent Semantic Analysis (LSA). For robust model selection we apply 50-fold Bootstrapping to obtain an average F-Score of 84.58 % on the trainingset and 86.33 % on the test set.

[1] Andrew McCallum,et al. Conditional Random Fields: Probabilistic Models for Segmenting and Labeling Sequence Data , 2001, ICML.

[2] M. Kenward,et al. An Introduction to the Bootstrap , 2007 .

[3] Burr Settles,et al. Biomedical Named Entity Recognition using Conditional Random Fields and Rich Feature Sets , 2004, NLPBA/BioNLP.

[4] R. Zimmer,et al. ProMiner: Organism-specific protein name detection using approximate string matching , 2004 .

[5] Alfonso Valencia,et al. Overview of BioCreAtIvE: critical assessment of information extraction for biology , 2005, BMC Bioinformatics.

[6] Fernando Pereira,et al. Identifying gene and protein mentions in text using conditional random fields , 2005, BMC Bioinformatics.

[7] Martin Hofmann-Apitius,et al. Biomedical and Chemical Named Entity Recognition with Conditional Random Fields: The Advantage of Dictionary Features , 2006, SMBM.

[8] Sophia Ananiadou,et al. Developing a Robust Part-of-Speech Tagger for Biomedical Text , 2005, Panhellenic Conference on Informatics.

[9] Burr Settles,et al. ABNER: an open source tool for automatically tagging genes, proteins and other entity names in text , 2005 .