Three Sentences Are All You Need: Local Path Enhanced Document Relation Extraction

@article{Huang2021ThreeSA,
  title={Three Sentences Are All You Need: Local Path Enhanced Document Relation Extraction},
  author={Quzhe Huang and Shengqi Zhu and Yansong Feng and Yuan Ye and Yuxuan Lai and Dongyan Zhao},
  journal={ArXiv},
  year={2021},
  volume={abs/2106.01793}
}
Document-level Relation Extraction (RE) is a more challenging task than sentence RE as it often requires reasoning over multiple sentences. Yet, human annotators usually use a small number of sentences to identify the relationship between a given entity pair. In this paper, we present an embarrassingly simple but effective method to heuristically select evidence sentences for document-level RE, which can be easily combined with BiLSTM to achieve good performance on benchmark datasets, even… 

Figures and Tables from this paper

Document-Level Relation Extraction with Sentences Importance Estimation and Focusing
TLDR
A Sentence Importance Estimation and Focusing (SIEF) framework for DocRE is proposed, where a sentence importance score and a sentence focusing loss are designed, encouraging DocRE models to focus on evidence sentences.
Modular Self-Supervision for Document-Level Relation Extraction
TLDR
This paper proposes decomposing document-level relation extraction into relation detection and argument resolution, taking inspiration from Davidsonian semantics, which enables it to incorporate explicit discourse modeling and leverage modular self-supervision for each sub-problem, which is less noise-prone and can be further refined end-to-end via variational EM.
Eider: Empowering Document-level Relation Extraction with Efficient Evidence Extraction and Inference-stage Fusion
TLDR
An evidence-enhanced framework, Eider, that empowers DocRE by efficiently extracting evidence and effectively fusing the extracted evidence in inference, and design a simple yet effective inference process that makes RE predictions on both extracted evidence and the full document, then fuses the predictions through a blending layer.
Eider: Evidence-enhanced Document-level Relation Extraction
TLDR
A novel 011 DocRE framework called EIDER is proposed that automat012 ically extracts and makes use of evidence and achieves state-of-the-art performance on 024 the DocRED, CDR, and GDA datasets.
Does Recommend-Revise Produce Reliable Annotations? An Analysis on Missing Instances in DocRED
TLDR
The underlying reason for the problems with the recommend-revise scheme is figured out: the scheme actually discourages annotators from supplementing adequate instances in the revision phase, which results in false negative samples and an obvious bias towards popular entities and relations.
What Do You Mean by Relation Extraction? A Survey on Datasets and Study on Scientific Relation Classification
TLDR
A comprehensive survey of RE datasets is provided, and the task definition and its adoption by the community are revisited, finding that cross-dataset and cross-domain setups are particularly lacking.
Enhancing Document-level Relation Extraction by Entity Knowledge Injection
TLDR
This paper introduces coreference distillation to inject coreference knowledge, endowing an RE model with the more general capability of coreference reasoning and employs representation reconciliation to inject factual knowledge and aggregate KG representations and document representations into a unified space.
Extracting entity relations for “problem-solving” knowledge graph of scientific domains using word analogy
TLDR
This paper presented an experiment with artificial intelligence papers from the Web of Science and achieved good performance, and used computer vision as an example to demonstrate the application of the extracted relations in constructing domain knowledge graphs and revealing historical research trends.

References

SHOWING 1-10 OF 20 REFERENCES
DocRED: A Large-Scale Document-Level Relation Extraction Dataset
TLDR
Empirical results show that DocRED is challenging for existing RE methods, which indicates that document-level RE remains an open problem and requires further efforts.
Neural Relation Extraction with Selective Attention over Instances
TLDR
A sentence-level attention-based model for relation extraction that employs convolutional neural networks to embed the semantics of sentences and dynamically reduce the weights of those noisy instances.
Cross-Sentence N-ary Relation Extraction with Graph LSTMs
TLDR
A general relation extraction framework based on graph long short-term memory networks (graph LSTMs) that can be easily extended to cross-sentence n-ary relation extraction is explored, demonstrating its effectiveness with both conventional supervised learning and distant supervision.
Double Graph Based Reasoning for Document-level Relation Extraction
TLDR
This paper proposes Graph Aggregation-and-Inference Network (GAIN) featuring double graphs, based on which GAIN first constructs a heterogeneous mention-level graph (hMG) to model complex interaction among different mentions across the document and proposes a novel path reasoning mechanism to infer relations between entities.
Distant Supervision for Relation Extraction beyond the Sentence Boundary
TLDR
This paper proposes the first approach for applying distant supervision to cross-sentence relation extraction with a graph representation that can incorporate both standard dependencies and discourse relations, thus providing a unifying way to model relations within and across sentences.
Reasoning with Latent Structure Refinement for Document-Level Relation Extraction
TLDR
This work proposes a novel model that empowers the relational reasoning across sentences by automatically inducing the latent document-level graph and develops a refinement strategy, which enables the model to incrementally aggregate relevant information for multi-hop reasoning.
Graph Convolution over Pruned Dependency Trees Improves Relation Extraction
TLDR
An extension of graph convolutional networks that is tailored for relation extraction, which pools information over arbitrary dependency structures efficiently in parallel is proposed, and a novel pruning strategy is applied to the input trees by keeping words immediately around the shortest path between the two entities among which a relation might hold.
Fact distribution in Information Extraction
TLDR
This paper compares three IE evaluation corpora, from the Message Understanding Conferences, and finds that a significant proportion of the facts mentioned therein are not described within a single sentence.
RENET: A Deep Learning Approach for Extracting Gene-Disease Associations from Literature
TLDR
A deep learning approach is designed and implemented, named RENET, which considers the correlation between the sentences in an article to extract gene-disease associations and has significantly improved the precision and recall rate.
...
...