摘要
提出一种基于神经网络的方法实现细菌和栖息地的关系抽取,充分利用神经网络的特性实现对隐含的深层特征的自动学习,以避免传统人工特征设计的复杂性和冗余性.该方法利用单词以及实体属性的分布式向量丰富句法和语义信息,使用两个不同神经网络模型从不同角度进行关系抽取,并融合文档级别的分类结果,在生物医学自然语言处理Bio NLP-ST 2016共享任务的BB-event语料上进行实验,取得了不错的F1值,表明该方法在微生物生长环境关系抽取上具有良好的性能.
Proposed in this paper is a neural network-based method for extracting the relationship between bacteria and their habitats. In this method,implicit senior features are learnt automatically to avoid the complexity and redundancy of the traditional artificial design of features,and,distributed vector representation with rich syntactic and semantic knowledge for words and entities,two different neural network models,as well as integrated document-level extraction results,are comprehensively employed to make an evaluation on the BB-event corpus from Bio NLP-ST 2016. Experimental results show that the proposed method achieves preferable F1 score,which means that it is effective in bacteria biotope extraction.
作者
王健
李虹磊
林鸿飞
杨志豪
张绍武
WANG Jian LI Hong-lei LIN Hong-fei YANG Zhi-hao ZHANG Shao-wu(School of Computer Science and Technology, Dalian University of Technology, Dalian 116024, Liaoning, China)
出处
《华南理工大学学报(自然科学版)》
EI
CAS
CSCD
北大核心
2017年第3期76-81,共6页
Journal of South China University of Technology(Natural Science Edition)
基金
国家自然科学基金资助项目(61572098
61572102
61562080)
国家重点研发计划项目(2016YFB1001103)~~
关键词
微生物生长环境关系抽取
卷积神经网络
长短时记忆神经网络
分布式向量
bacteria biotope extraction
convolutional neural network
long short-term memory neural network
distributed vector representation