DSREFC: Improving Distantly-supervised Neural Relation Extraction Using Feature Combination

Distant supervisory relationship extraction can automatically align the expected entity pairs, and automatically obtain a large number of annotation data, thus saving a lot of labor costs. However, the automatic acquisition of annotated data will lead to the introduction of noise data, making little effect of relation extraction task. To solve this problem, we propose the relation extraction model DSREFC, which integrates semantic features and syntactic features into the representation and uses attention mechanism to obtain bag representation. The DSREFC model has three characteristics: 1) The BERT+Bi-LSTM is used as the text representation extractor to extract the semantic information of the text. 2) the grammatical information is extracted with the GCN network used in the text, and combine theBi-LSTM output with the GCN output to obtain a distributed representation of each token. 3) The two-step attention mechanism is used to remove the influence of the noise data by giving the noise data a lower weight value. Attention is used to obtain the sentence representation for each token and attention is used to obtain the packet representation for each sentence in the packet. Experiments show that the DSREFC model combining semantic features and grammatical features can significantly improve the effect of relation extraction.

[1]  Jun Zhao,et al.  Distant Supervision for Relation Extraction via Piecewise Convolutional Neural Networks , 2015, EMNLP.

[2]  Ming-Wei Chang,et al.  BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding , 2019, NAACL.

[3]  Max Welling,et al.  Semi-Supervised Classification with Graph Convolutional Networks , 2016, ICLR.

[4]  Zhiyuan Liu,et al.  Neural Relation Extraction with Selective Attention over Instances , 2016, ACL.

[5]  Yang Xiang,et al.  Distant Supervision for Relation Extraction via Group Selection , 2015, ICONIP.

[6]  Partha Talukdar,et al.  CESI: Canonicalizing Open Knowledge Bases using Embeddings and Side Information , 2018, WWW.

[7]  Chiranjib Bhattacharyya,et al.  RESIDE: Improving Distantly-Supervised Neural Relation Extraction using Side Information , 2018, EMNLP.

[8]  Diego Marcheggiani,et al.  Encoding Sentences with Graph Convolutional Networks for Semantic Role Labeling , 2017, EMNLP.

[9]  Luke S. Zettlemoyer,et al.  Knowledge-Based Weak Supervision for Information Extraction of Overlapping Relations , 2011, ACL.

[10]  Leila Safari,et al.  A Study of Recent Contributions on Information Extraction , 2018, ArXiv.

[11]  Andrew McCallum,et al.  Modeling Relations and Their Mentions without Labeled Text , 2010, ECML/PKDD.

[12]  Noah A. Smith,et al.  Proceedings of the 54th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) , 2016, ACL 2016.

[13]  Daniel Jurafsky,et al.  Distant supervision for relation extraction without labeled data , 2009, ACL.

[14]  Quoc V. Le,et al.  Sequence to Sequence Learning with Neural Networks , 2014, NIPS.

[15]  Sharmistha Jat,et al.  Improving Distantly Supervised Relation Extraction using Word and Entity Based Attention , 2018, AKBC@NIPS.

[16]  Josiah Poon,et al.  A comparative study of the current technologies and approaches of relation extraction in biomedical literature using text mining , 2017, 2017 4th IEEE International Conference on Engineering Technologies and Applied Sciences (ICETAS).

[17]  Mihai Surdeanu,et al.  The Stanford CoreNLP Natural Language Processing Toolkit , 2014, ACL.

[18]  Oren Etzioni,et al.  Open Language Learning for Information Extraction , 2012, EMNLP.

[19]  Christopher D. Manning,et al.  Incorporating Non-local Information into Information Extraction Systems by Gibbs Sampling , 2005, ACL.

[20]  Jun Zhao,et al.  Distant Supervision for Relation Extraction with Sentence-Level Attention and Entity Descriptions , 2017, AAAI.

[21]  Yawei Zhao,et al.  Knowledge Graph Completion via Complete Attention between Knowledge Graph and Entity Descriptions , 2019, CSAE.