Research on Entity Relation Extraction for Military Field

Research on Entity Relation Extraction for Military Field Chen Liang12, Hongying Zan1,Yajun Liu1,Yunfang Wu3 1Natural Language Processing Laboratory, School of Information and Engineering, Zhengzhou University, Zhengzhou, China 2Key Laboratory of Universal Wireless Communications, Ministry of Education, Beijing University of Posts and Telecommunications, Beijing, China 3Key Laboratory of Computational Linguistics, Peking University, Beijing, China liangchenee@163.com, iehyzan@zzu.edu.cn, liuyajun_gz@163.com, wuyf@pku.edu.cn Abstract The purpose of entity relation extraction is to obtain the relation between named entities from the text. However，most of the researches fo- cus on the tourism, music, medical and other fields, seldom relate with entity relation extrac- tion in the military field. This paper preprocess- es military data sets, defines entity relations and completes classification. Based on dependen- cy syntax relations, we extract the core verb features and dependency triples fea- tures in the sentences, using SVM as the classifier. The experiment results show that the features extracted in this paper can effectively improve the performance of relation detection and relation extraction. The F1 value for rela- tion extraction can achieve 89%. 1 Introduction As an important technology in natural language processing, entity relation extraction refers to the relation between named entities obtained from the corpus expressed by natural language. For example, the relation between people and articles (ART), the relation between people and organizations (ORG- AFF) are entity relations. Entity relation extraction is a more in-depth study based on named entity recognition(NER). It is widely used in question answering and machine translation. Nowadays, with the rapid development of the Internet, the re- lation extraction technology not only has profound theoretical significance, but also has broad applica- tion prospects. The study of relation extraction was originally conducted on the main line of message understand- ing conference(MUC). Later, automatic content extraction(ACE) replaced it. A large number of excellent extraction methods have been proposed in the evaluation, which has effectively promoted the development of relation extraction research. Currently, the methods for entity relation extrac- tion mainly include machine learning methods, hybrid extraction methods, and deep learning methods. 2 Related Works Entity relation extraction is a key issue in infor- mation extraction. Researchers have made many breakthroughs in this field currently. In the ma- chine learning based relation extraction methods, some studies use unsupervised methods to extract relations through clustering. YanY et al., (2009) presented a clustering methods based on pattern combination. Experiments on two different types of texts in Wikipedia had achieved good results. Shi et al., (2016) constructed a corpus for relational discriminant research based on ACE2005 task Chinese news text corpus, combining with the word vector trained by Sohu news data and using the convolutional neural network model for rela- tion discrimination. The F1 value can reach 81.78%. The unsupervised methods have shown its good cross-domain versatility in the above research, PACLIC 32 81 32nd Pacific Asia Conference on Language, Information and Computation Hong Kong, 1-3 December 2018 Copyright 2018 by the authors but its ability to describe the types of relations and the extraction result in special fields are poor. The …

Content truncated to 500 words