点击次数：

论文类型：期刊论文

发表时间：2011-10-01

发表刊物：JOURNAL OF COMPUTATIONAL AND THEORETICAL NANOSCIENCE

收录刊物：Scopus、SCIE、EI

卷号：8

期号：10

页面范围：1925-1932

ISSN号：1546-1955

关键字：Interaction Extraction; Graph Kernel; Biomedical Literature

摘要：Extracting protein protein interactions from biomedical literatures is an important task in biomedical text mining. In this paper, we propose an improved all-paths kernel based method for this task. First, we improve all-paths kernel method by adding adjacent label to node label so that it can match the contiguous label sequences of graphs. Second, we divide the dependency graph into the token subgraph and relation subgraph which can better represent sentence structure to graph kernel and efficiently reduce the weak features in feature space. We evaluate the proposed method on five publicly available PPI corpora and perform detailed comparisons with other methods. As a result, our method significantly outperforms all-paths kernel method.