基于树核的蛋白质相互作用关系提取的研究  

Research on tree kernel-based protein-protein interaction extraction

作　　者：刘念[1] 马长林[1] 张勇[1] 王梦[1] 

机构地区：[1]华中师范大学计算机学院,湖北武汉430079

出　　处：《华中科技大学学报（自然科学版）》2013年第S2期232-236,共5页Journal of Huazhong University of Science and Technology(Natural Science Edition)

基　　金：国家自然科学基金资助项目(61202470)

摘　　要：蛋白质相互作用关系对理解生物过程具有非常重要的意义,为了解决同位语依存关系带来的噪音干扰,提出了一个改进的基于树核的PPI提取方法,通过定义一些相关的处理规则来有效优化两个蛋白质之间的最短依存路径,在此基础上,用有效优化路径来指导成分句法树的裁剪,使得用于PPI提取的成分树更加精确和简洁.实验结果表明:用有效优化路径指导的成分句法树在五个常用的语料库上都取得了较好的效果.Protein-protein interaction(PPI)plays an important role in understanding biological processes.In order to resolve the noise interference brought by appositive dependency,an improved tree kernel-based PPI extraction method was proposed.Some relevant processing rules were defined to effectively optimize the shortest dependency path between two proteins.On this basis of these rules, an effectively optimized path was used to direct the cutting of constituent parse tree,which made the constituent parse tree for PPI extraction more precise and concise.The experimental results show that our effectively optimized path-directed constituent parse tree achieves better results on five commonly used corpora.

关 键 词：蛋白质 关系提取 提取方法 同位语依存关系 树核 

分 类 号：TP391.1[自动化与计算机技术—计算机应用技术]