点击次数：

论文类型：期刊论文

发表时间：2012-03-01

发表刊物：Journal of Computational Information Systems

收录刊物：EI、Scopus

卷号：8

期号：3

页面范围：1221-1228

ISSN号：15539105

摘要：Automatic extraction of protein-protein interactions (PPI) has become a critical task in the field of biomedical text mining due to the dynamic progress in biomedical technology. Kernel-based machine learning methods have been widely used to extract PPI automatically from biomedical literature, and different kernels focusing on different parts of sentence structure have been proposed for the PPI extraction task. In this paper, we present a method to combine useful features explored from dependency graph for PPI extraction. Based on the shortest path between proteins in the dependency graph, a new distance feature, SentenceDistance, is proposed and used in feature-based kernel, which contributes to the improvement of the performance. Further, we combine the feature-based kernel with graph and walk-weighted subsequence kernels, obtaining 0.624 F-Score and 0.872 AUC on the AIMed corpus, which is comparable with the state-of-the-art approaches. 1553-9105/Copyright ? 2012 Binary Information Press.