Relation Extraction with Type-aware Map Memories of Word Dependencies

Findings of the Association for Computational Linguistics: ACL-IJCNLP 2021, pages 2501–2512 August 1–6, 2021. ©2021 Association for Computational Linguistics 2501 Relation Extraction with Type-aware Map Memories of Word Dependencies Guimin Chen♦∗, Yuanhe Tian♥∗, Yan Song♠♥†, Xiang Wan♥ ♦QTrade ♥University of Washington ♠The Chinese University of Hong Kong (Shenzhen) ♥Shenzhen Research Institute of Big Data ♦chenguimin@foxmail.com ♥yhtian@uw.edu ♠songyan@cuhk.edu.cn ♥wanxiang@sribd.cn Abstract Relation extraction is an important task in in- formation extraction and retrieval that aims to extract relations among the given entities from running texts. To achieve a good performance for this task, previous studies have shown that a good modeling of the contextual informa- tion is required, where the dependency tree of the input sentence can be a beneficial source among different types of contextual informa- tion. However, most of these studies focus on the dependency connections between words with limited attention paid to exploiting depen- dency types. In addition, they often treat dif- ferent dependency connections equally in mod- eling so that suffer from the noise (inaccurate dependency parses) in the auto-generated de- pendency tree. In this paper, we propose a neural approach for relation extraction, with type-aware map memories (TaMM) for encod- ing dependency types obtained from an off-the- shelf dependency parser for the input sentence. Specifically, for each word in an entity, TaMM maps all associated words along with the de- pendencies among them to memory slots and then assigns a weight to each slot according to its contribution to relation extraction. Our ap- proach not only leverages dependency connec- tions and types between words, but also distin- guishes reliable dependency information from noisy ones and appropriately model them. The effectiveness of our approach is demonstrated by the experiments on two English benchmark datasets, where our approach achieves state-of- the-art performance on both datasets.1 1 Introduction Relation extraction is an important natural lan- guage processing (NLP) task that facilitates in- formation extraction, whose results is beneficial *Equal contribution. †Corresponding author. 1The code and models involved in this paper are released at https://github.com/cuhksz-nlp/RE-TaMM. Figure 1: An illustration of an example sentence (in- cluding the entity terms “bone marrow” and “stem cells”) with its dependency parsing result. to downstream tasks such as schema induction (Nimishakavi et al., 2016), knowledge graph con- struction (Yu et al., 2017), and question answer- ing (Xu et al., 2016). Normally, relation extrac- tion aims to predict the relation between each pair of entities in a given sentence. For example, in the sentence “the [bone marrow]e1 produces [stem cells]e2” with the entity terms “bone marrow” and “stem cells”, the relation between the two entities is “Product-Producer“. Therefore, the ability of modeling the context from the input is of great im- portance to guarantee the performance of relation extraction. To this end, approaches based on neural networks have achieved promising success for the task in the past decade (Socher et al., 2012; Zeng et al., 2014; Zhang and Wang, 2015; Xu et al., 2015; dos Santos et al., 2015; Zhang et al., 2015; Wang et al., 2016; …

Content truncated to 500 words