基于树核函数的蛋白质相互作用关系抽取研究
首发时间:2012-01-18
摘要:本文分析了目前关系抽取中结构化信息表达形式所存在的问题,提出了一种最短依存路径指导的成分句法树(SDP-CPT)裁剪策略,即依据两个蛋白质之间的最短依存路径,对句法树进行有针对性的裁剪,探究更有效的蛋白质关系实例结构化表达形式。在多个语料库上的实验表明,本文提出的PPI抽取方法其性能显著高于其它几种结构化信息表达形式。特别是,在AIMed语料上的PPI抽取取得了58.1的F1值,是目前基于单一核函数的PPI抽取系统的最好水平。
For information in English, please click here
Research on Tree Kernel-Based Protein-Protein Interaction Extraction
Abstract:This paper conducts a comprehensive investigation into the disadvantages of various kinds of existing structured information reprepsentation in relation extraction, and then proposes a pruning strategy for shortest dependency path-directed constitute parse tree (SDP-CPT). Based on the short dependency path between two proteins, the syntactic parse tree is pruned in order to form a more effecitive structure for PPI extraction. The experiments on multiple corpora show that our method significantly outperforms other kinds of structured information representation. particularly, it achieves a promising performance of 58.1 in F-measure, being the state-of-the-art single kernel-based one in current PPI extraction systems.
Keywords: PPI Extraction;Dependency Information;Tree Kernel Function
论文图表:
引用
No.****
同行评议
共计0人参与
勘误表
基于树核函数的蛋白质相互作用关系抽取研究
评论
全部评论0/1000