Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Text2NKG: Fine-Grained N-ary Relation Extraction for N-ary relational Knowledge Graph Construction

Haoran Luo, Haihong E, Yuhao Yang, Tianyu Yao, Yikai Guo, Zichen Tang, Wentai Zhang, Kaiyang Wan, Shiyao Peng, Meina Song, Wei Lin, Yifan Zhu, Luu Anh Tuan

TL;DR

This work introduces a span-tuple classification approach with hetero-ordered merging and output merging to accomplish fine-grained n-ary relation extraction in different arity, and supports four typical NKG schemas: hyper-relational schema, event-based schema, role-based schema, and hypergraph-based schema, with high flexibility and practicality.

Abstract

Beyond traditional binary relational facts, n-ary relational knowledge graphs (NKGs) are comprised of n-ary relational facts containing more than two entities, which are closer to real-world facts with broader applications. However, the construction of NKGs remains at a coarse-grained level, which is always in a single schema, ignoring the order and variable arity of entities. To address these restrictions, we propose Text2NKG, a novel fine-grained n-ary relation extraction framework for n-ary relational knowledge graph construction. We introduce a span-tuple classification approach with hetero-ordered merging and output merging to accomplish fine-grained n-ary relation extraction in different arity. Furthermore, Text2NKG supports four typical NKG schemas: hyper-relational schema, event-based schema, role-based schema, and hypergraph-based schema, with high flexibility and practicality. The experimental results demonstrate that Text2NKG achieves state-of-the-art performance in F1 scores on the fine-grained n-ary relation extraction benchmark. Our code and datasets are publicly available.

Text2NKG: Fine-Grained N-ary Relation Extraction for N-ary relational Knowledge Graph Construction

TL;DR

This work introduces a span-tuple classification approach with hetero-ordered merging and output merging to accomplish fine-grained n-ary relation extraction in different arity, and supports four typical NKG schemas: hyper-relational schema, event-based schema, role-based schema, and hypergraph-based schema, with high flexibility and practicality.

Abstract

Beyond traditional binary relational facts, n-ary relational knowledge graphs (NKGs) are comprised of n-ary relational facts containing more than two entities, which are closer to real-world facts with broader applications. However, the construction of NKGs remains at a coarse-grained level, which is always in a single schema, ignoring the order and variable arity of entities. To address these restrictions, we propose Text2NKG, a novel fine-grained n-ary relation extraction framework for n-ary relational knowledge graph construction. We introduce a span-tuple classification approach with hetero-ordered merging and output merging to accomplish fine-grained n-ary relation extraction in different arity. Furthermore, Text2NKG supports four typical NKG schemas: hyper-relational schema, event-based schema, role-based schema, and hypergraph-based schema, with high flexibility and practicality. The experimental results demonstrate that Text2NKG achieves state-of-the-art performance in F1 scores on the fine-grained n-ary relation extraction benchmark. Our code and datasets are publicly available.
Paper Structure (32 sections, 12 equations, 5 figures, 4 tables)

This paper contains 32 sections, 12 equations, 5 figures, 4 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. N-ary relational Knowledge Graph
  4. N-ary Relation Extraction
  5. Preliminaries
  6. Methodology
  7. Overview of Text2NKG
  8. Span-tuple Multi-label Classification
  9. Training Strategy
  10. Hetero-ordered Merging
  11. Output Merging
  12. Experiments
  13. Experimental Setup
  14. Main Results (RQ1)
  15. Results on Various NKG Schemas (RQ2)
  16. ...and 17 more sections

Figures (5)

  • Figure 1: An example of NKG construction.
  • Figure 2: Taking a real-world textual fact as an example, we can extract a four-arity structured span-tuple for entities (Einstein, University of Zurich, Doctorate, Physics) with an answer label-list for relations accordingly as a 4-ary relational fact from the sentence through n-ary relation extraction.
  • Figure 3: An overview of Text2NKG extracting n-ary relation facts from a natural language sentence in hyper-relational NKG schema for an example.
  • Figure 4: (a) Precision, Recall, and $F_1$ changes in the dev set during the training of Text2NKG. (b) The changes of the number of true facts, the number of predicted facts, and the number of predicted accurate facts during the training of Text2NKG. (c) Precision, Recall, and $F_1$ results on different null-label hyperparameter ($\alpha$) settings.
  • Figure 5: (a) The changes of the number of extracted n-ary RE in different arity, where "pred_n" represents the number of extracted n-ary facts with different arities by Text2NKG, and "ans_n" represents the ground truth. (b) Case study of Text2NKG's n-ary relation extraction in four schemas on HyperRED.