BioLMiner System: Interaction Normalization Task and Interaction Pair Task in the BioCreative II.5 Challenge
Issue No. 03 - July-September (2010 vol. 7)
DOI Bookmark: http://doi.ieeecomputersociety.org/10.1109/TCBB.2010.47
Yifei Chen , Vrije Universiteit Brussel, Brussels
Feng Liu , Vrije Universiteit Brussel, Brussels
Bernard Manderick , Vrije Universiteit Brussel, Brussels
This paper describes a Biological Literature Miner (BioLMiner) system and its implementation. BioLMiner is a text mining system for biological literature, whose purpose is to extract useful information from biological literature, including gene and protein names, normalized gene and protein names, and protein-protein interaction pairs. BioLMiner has three main subsystems in a pipeline structure: a gene mention recognizer (GMRer), a gene normalizer (GNer), and a protein-protein interaction pair extractor (PPIEor). All these subsystems are developed based on the machine learning techniques including support vector machines (SVMs) and conditional random fields (CRFs) together with carefully designed informative features. At the same time, BioLMiner makes use of some biological specific resources and existing natural language processing tools. In order to evaluate and compare BioLMiner, it is adapted to participate in two tasks of the BioCreative II.5 challenge: interaction normalization task (INT) using GNer and interaction pair task (IPT) using PPIEor. Our system is among the highest performing systems on the two tasks from which it can be seen that GMRer provides a good support for the INT and IPT although its performance is not evaluated, and the methods developed in GNer and PPIEor are extended well to the BioCreative II.5 tasks.
Text mining, machine learning, interaction protein normalization, interaction pair extraction, BioCreactive II.5 challenge.
Yifei Chen, Feng Liu, Bernard Manderick, "BioLMiner System: Interaction Normalization Task and Interaction Pair Task in the BioCreative II.5 Challenge", IEEE/ACM Transactions on Computational Biology and Bioinformatics, vol. 7, no. , pp. 428-441, July-September 2010, doi:10.1109/TCBB.2010.47