In many fields of molecular biology, especially in signal transduction, thousands of new results are published each year, and relationships between the information in different articles is often not immediately apparent, even if a research manages to read all relevant, published articles. Medline in NCBI contains over 10 million abstracts, and approximately 40,000 new abstracts are added each month. Although there are growing numbers of sequence database and other hand-constructed databases, most new information is unstructured text Medline and full text journals. Biological literature mining can be useful in accomplishing the following tasks: identification of the names of biological entities, identification of various among biological entities, and identification of the status of biological discoveries stated in literature and web pages.
Currently, our groups work on biological literature mining which including biological named entity recognition and biological relation extraction. We have designed two named entity recognition system which one base on dictionary-based method and the other machine learning method. In relation extraction, we have implemented a pattern-based system aim for extracting Gene-Disease and Protein-Protein relation. We believe that these system can facilitate the biologist’s paper reading efforts.

Demo site URL: http://bioinformatics.iis.sinica.edu.tw/BioLiteratureMining/