Chinese Relation Extraction with Multi-Grained Information and External Linguistic Knowledge

Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics, pages 4377–4386 Florence, Italy, July 28 - August 2, 2019. c⃝2019 Association for Computational Linguistics 4377 Chinese Relation Extraction with Multi-Grained Information and External Linguistic Knowledge Ziran Li1,2∗, Ning Ding1,2∗, Zhiyuan Liu2,3,4, Hai-Tao Zheng1,2† , Ying Shen5 1Tsinghua Shenzhen International Graduate School, Tsinghua University 2Department of Computer Science and Technology, Tsinghua University, Beijing, China 3Institute for Artificial Intelligence, Tsinghua University, Beijing, China 4State Key Lab on Intelligent Technology and Systems, Tsinghua University, Beijing, China 5School of Electronics and Computer Engineering, Peking University Shenzhen Graduate School {lizr18,dingn18}@mails.tsinghua.edu.cn Abstract Chinese relation extraction is conducted using neural networks with either character-based or word-based inputs, and most existing meth- ods typically suffer from segmentation errors and ambiguity of polysemy. To address the is- sues, we propose a multi-grained lattice frame- work (MG lattice) for Chinese relation ex- traction to take advantage of multi-grained language information and external linguistic knowledge. In this framework, (1) we incorpo- rate word-level information into character se- quence inputs so that segmentation errors can be avoided. (2) We also model multiple senses of polysemous words with the help of external linguistic knowledge, so as to alleviate poly- semy ambiguity. Experiments on three real- world datasets in distinct domains show con- sistent and significant superiority and robust- ness of our model, as compared with other baselines. The source code of this paper can be obtained from https://github.com/ thunlp/Chinese_NRE. 1 Introduction Relation extraction (RE) has a pivotal role in infor- mation extraction (IE), aiming to extract seman- tic relations between entity pairs in natural lan- guage sentences. In downstream applications, this technology is a key module for constructing large- scale knowledge graphs. Recent developments in deep learning have heightened the interest for neu- ral relation extractions (NRE), which attempt to use neural networks to automatically learn seman- tic features (Liu et al., 2013; Zeng et al., 2014, 2015; Lin et al., 2016; Zhou et al., 2016; Jiang et al., 2016). ∗ indicates equal contribution † Corresponding author: Hai-Tao Zheng. ( E-mail: zheng.haitao@sz.tsinghua.edu.cn ) Figure 1: An example of segmentation ambiguity and polysemy ambiguity in Chinese RE. Although it is not necessary for NRE to per- form feature engineering, they ignore the fact that different language granularity of input will have a significant impact on the model, especially for Chinese RE. Conventionally, according to the difference in granularity, most existing methods for Chinese RE can be divided into two types: character-based RE and word-based RE. For the character-based RE, it regards each in- put sentence as a character sequence. The short- coming of this kind of method is that it can- not fully exploit word-level information, capturing fewer features than the word-based methods. For the word-based RE, word segmentation should be first performed. Then, a word sequence is derived and fed into the neural network model. However, the performance of the word-based models could be significantly impacted by the quality of seg- mentation. For example, as shown in Fig 1, the …

Content truncated to 500 words