Document-Level Zero-Shot Relation Extraction with Entity Side Information

TL;DR

This work tackles document-level zero-shot relation extraction (DocZSRE) by moving away from LLM-generated synthetic data and toward Entity Side Information to predict unseen relations. The proposed DocZSRE-SI framework consists of two modules: Building Entity Side Information (descriptions, hypernyms, and types) and a Zero-Shot Relation Extraction module that embeds this information, encodes relation labels, and uses a Dynamic Weighted Score to select the best unseen label. Key contributions include a robust embedding scheme that combines description, hypernym, type, role-based, and context features, and a weighted scoring mechanism with a Confidence Weightage that improves reliability. Experiments on DocRED, RE-DocRED, and MEN-Dataset show substantial macro F1 improvements over baselines and GenRDK, especially for larger unseen label sets, while highlighting challenges in inter-sentential extraction and variance across datasets, paving the way for scalable DocZSRE in low-resource languages.

Abstract

Document-Level Zero-Shot Relation Extraction (DocZSRE) aims to predict unseen relation labels in text documents without prior training on specific relations. Existing approaches rely on Large Language Models (LLMs) to generate synthetic data for unseen labels, which poses challenges for low-resource languages like Malaysian English. These challenges include the incorporation of local linguistic nuances and the risk of factual inaccuracies in LLM-generated data. This paper introduces Document-Level Zero-Shot Relation Extraction with Entity Side Information (DocZSRE-SI) to address limitations in the existing DocZSRE approach. The DocZSRE-SI framework leverages Entity Side Information, such as Entity Mention Descriptions and Entity Mention Hypernyms, to perform ZSRE without depending on LLM-generated synthetic data. The proposed low-complexity model achieves an average improvement of 11.6% in the macro F1-Score compared to baseline models and existing benchmarks. By utilizing Entity Side Information, DocZSRE-SI offers a robust and efficient alternative to error-prone, LLM-based methods, demonstrating significant advancements in handling low-resource languages and linguistic diversity in relation extraction tasks. This research provides a scalable and reliable solution for ZSRE, particularly in contexts like Malaysian English news articles, where traditional LLM-based approaches fall short.

Figures (1)

• Figure 1: High-Level architecture of DocZSRE-SI Framework