EntGPT: Entity Linking with Generative Large Language Models

2024-02-09

AI Summary

Entity linking (EL) aims to map textual mentions to corresponding knowledge base entries, yet conventional discriminative models suffer from poor cross-domain generalization and training complexity. This paper proposes EntGPT, a novel generative large language model (LLM)-based EL framework comprising two variants: zero-shot hard-prompting (EntGPT-P) and instruction-tuning (EntGPT-I). EntGPT-P employs a three-step structured hard prompt to enforce strong contextual control; EntGPT-I reformulates EL as an instruction-following task and introduces lightweight fine-tuning—the first such adaptation for EL. Evaluated on ten EL benchmarks, EntGPT achieves zero-shot state-of-the-art performance, with up to 36% absolute micro-F1 gain; under supervised settings, it yields an average 2.1% improvement and significantly enhances cross-task transferability—outperforming baselines on six QA tasks. The framework is compatible with both open- and closed-weight LLMs and generalizes across domains without domain-specific adaptation.

Technology Category

• Natural Language Processing: Generation
• Machine Learning: Large Multimodal Models (LMMs)
• Data Mining & Knowledge Management: Linked Open Data, Knowledge Graphs & KB Completion

Application Category

• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Search and Retrieval-Augmented AI: Multilingual and cross-lingual Web search

Abstract

Entity Linking in natural language processing seeks to match text entities to their corresponding entries in a dictionary or knowledge base. Traditional approaches rely on contextual models, which can be complex, hard to train, and have limited transferability across different domains. Generative large language models like GPT offer a promising alternative but often underperform with naive prompts. In this study, we introduce EntGPT, employing advanced prompt engineering to enhance EL tasks. Our three-step hard-prompting method (EntGPT-P) significantly boosts the micro-F_1 score by up to 36% over vanilla prompts, achieving competitive performance across 10 datasets without supervised fine-tuning. Additionally, our instruction tuning method (EntGPT-I) improves micro-F_1 scores by 2.1% on average in supervised EL tasks and outperforms several baseline models in six Question Answering tasks. Our methods are compatible with both open-source and proprietary LLMs. All data and code are available on GitHub at https://github.com/yifding/In_Context_EL.

Problem

• Enhancing Entity Linking using advanced prompt engineering
• Improving performance of generative LLMs in EL tasks
• Achieving competitive results without supervised fine-tuning

Innovation

• Advanced prompt engineering for EL tasks
• Three-step hard-prompting boosts performance
• Instruction tuning enhances supervised EL tasks