Surface pattern-enhanced relation extraction with global constraints
- PDF / 2,270,713 Bytes
- 32 Pages / 439.37 x 666.142 pts Page_size
- 15 Downloads / 208 Views
Surface pattern-enhanced relation extraction with global constraints Haiyun Jiang1 Wei Wang1
· JunTao Liu1 · Sheng Zhang2 · Deqing Yang3 · Yanghua Xiao1 ·
Received: 23 June 2019 / Revised: 29 June 2020 / Accepted: 31 July 2020 © Springer-Verlag London Ltd., part of Springer Nature 2020
Abstract Relation extraction is one of the most important tasks in information extraction. The traditional works either use sentences or surface patterns (i.e., the shortest dependency paths of sentences) to build extraction models. Intuitively, the integration of these two kinds of methods will further obtain more robust and effective extraction models, which is, however, ignored in most of the existing works. In this paper, we aim to learn the embeddings of surface patterns to further augment the sentence-based models. To achieve this purpose, we propose a novel pattern embedding learning framework with the weighted multi-dimensional attention mechanism. To suppress noise in the training dataset, we mine the global statistics between patterns and relations and introduce two kinds of prior knowledge to guide the pattern embedding learning. Based on the learned embeddings, we present two augmentation strategies to improve the existing relation extraction models. We conduct extensive experiments on two popular datasets (i.e., NYT and KnowledgeNet) and observe promising performance improvements.
B
Yanghua Xiao [email protected] Haiyun Jiang [email protected] JunTao Liu [email protected] Sheng Zhang [email protected] Deqing Yang [email protected] Wei Wang [email protected]
1
Shanghai Key Laboratory of Data Science, School of Computer Science, Fudan University, Shanghai, China
2
Ping An Health Technology, Shanghai, China
3
School of Data Science, Fudan University, Shanghai, China
123
H. Jiang et al.
Keywords Relation classification · Surface pattern · Deep learning · Self-attention · Global constraints
1 Introduction Relation extraction (RE) is a fundamental task for many AI applications, e.g., information retrieval [4], text analysis [63] and knowledge acquisition [16]. The task of RE aims to identify and categorize the relations between a pair of entity mentions in texts, where the relation set is predefined. For example, a RE system will identify the relation person-nationality for the entity pair John Steinbeck, United States. More often than not, an entity pair might have multiple semantic relations that also need to be identified by a RE model. For example, the entity pair Jobs, Apple has two semantic relations the-CEO-of and founder-of. In this paper, the input is an entity pair of interest with its labeled sentences or surface patterns and the outputs are the predicted relations. In general, the existing works for RE can be classified into two categories: sentence-based and surface pattern-based methods. Sentence-based methods The sentence-based methods take all the information in sentences to infer the target relations given an entity pair. In the early years, researchers con
Data Loading...
