Identifying biological terms from text by support vector machine

• Published in: 2011 6th IEEE Conference on Industrial Electronics and Applications pp. 455 - 458
• Main Authors: Zhenfei Ju, Meichen Zhou, Fei Zhu
• Format: Conference Proceeding
• Language: English
• Published: IEEE 01.06.2011
• Subjects: Biological Data Mining; Biological Terms Identification; Biology; Conferences; Data mining; Hidden Markov models; Machine Learning; Support Vector Machine; Support vector machines; Testing; Training
• ISBN: 9781424487547; 1424487544
• ISSN: 2156-2318
• DOI: 10.1109/ICIEA.2011.5975627

In contemporary society, an increasing number of people are involved in the biomedical research. However there is still a large amount of biological knowledge in the various unstructured documents so that it is difficult to analyze biological data. How to identify biological terms effectively from text is one of the important problems in the area of bioinformatics. Nowadays the precision of the best biological terms identification system has reached more than 80%, but is lower than the one of general system. Here we aim to recognize the name of the specified type from biological data set. We choose support vector machine (SVM) to do the work. With the help of GENIA corpus which is a collection of Medline abstracts, we get the precision rate= 84% and recall rate=81% in total for the two categories classification problem. When meeting the multiple categories classification problem, SVM can identify biological terms accurately, but the recall rate is very low. The increasing number of test data will not result in a decrease of precision, and the recall rate will increase.