Zero-Shot Relation Extraction via Reading Comprehension

2017/06/13 by Omer Levy, Minjoon Seo, Levy, Omer +5 · 67 citations
Computer Science · #Artificial Intelligence (cs.AI) #Computation and Language (cs.CL) #FOS: Computer and information sciences #Machine Learning (cs.LG) #cs.AI #cs.CL #cs.LG

paper · pdf · doi:10.48550/arxiv.1706.04115

CoNLL 2017

arxiv created 2017/06/13 · arxiv updated 2017/06/14

Abstract

We show that relation extraction can be reduced to answering simple reading comprehension questions, by associating one or more natural-language questions with each relation slot. This reduction has several advantages: we can (1) learn relation-extraction models by extending recent neural reading-comprehension techniques, (2) build very large training sets for those models by combining relation-specific crowd-sourced questions with distant supervision, and even (3) do zero-shot learning by extracting new relation types that are only specified at test-time, for which we have no labeled training examples. Experiments on a Wikipedia slot-filling task demonstrate that the approach can generalize to new questions for known relation types with high accuracy, and that zero-shot generalization to unseen relation types is possible, at lower accuracy levels, setting the bar for future work on this task.

Cited by

• Investigating Model Editing for Unlearning in Large Language Models
• An Information-Theoretic Framework for Robust Large Language Model Editing
• EtCon: Edit-then-Consolidate for Reliable Knowledge Editing
• EvoEdit: Lifelong Free-Text Knowledge Editing through Latent Perturbation Augmentation and Knowledge-driven Parameter Fusion
• Lightweight Model Editing for LLMs to Correct Deprecated API Recommendations
• Representation Interventions Enable Lifelong Unstructured Knowledge Control
• Catastrophic Forgetting in Kolmogorov-Arnold Networks
• Can Fine-Tuning Erase Your Edits? On the Fragile Coexistence of Knowledge Editing and Adaptation
• Balancing Knowledge Updates: Toward Unified Modular Editing in LLMs
• ForgetBench: Benchmarking Forgetting Dynamics of Long-Term Parametric Memory in Language Models
• Metis: Memory Foundation Model
• A Survey on Unlearning in Large Language Models
• Edit Less, Achieve More: Dynamic Sparse Neuron Masking for Lifelong Knowledge Editing in LLMs
• Measuring the Effect of Disfluency in Multilingual Knowledge Probing Benchmarks
• Understanding Parametric Knowledge Injection in Retrieval-Augmented Generation
• ZeroGR: A Generalizable and Scalable Framework for Zero-Shot Generative Retrieval
• EvoEdit: Evolving Null-space Alignment for Robust and Efficient Knowledge Editing
• REPAIR: Robust Editing via Progressive Adaptive Intervention and Reintegration
• Autoencoding-Free Context Compression for LLMs via Contextual Semantic Anchors
• Energy-Regularized Sequential Model Editing on Hyperspheres
• KnowledgeSmith: Uncovering Knowledge Updating in LLMs with Model Editing and Unlearning
• Is Model Editing Built on Sand? Revealing Its Illusory Success and Fragile Foundation
• SUIT: Knowledge Editing with Subspace-Aware Key-Value Mappings
• Fine-tuning Done Right in Model Editing
• Layerwise Importance Analysis of Feed-Forward Networks in Transformer-based Language Models
• From Injection to Defense: Constructing Edit-Based Fingerprints for Large Language Models
• A Dual-Axis Taxonomy of Knowledge Editing for LLMs: From Mechanisms to Functions
• Comparing Knowledge Injection Methods for LLMs in a Low-Resource Regime
• EMSEdit: Efficient Multi-Step Meta-Learning-based Model Editing
• Aligning Language Models with Real-time Knowledge Editing
• Latent Knowledge Scalpel: Precise and Massive Knowledge Editing for Large Language Models
• A Survey on Generative Model Unlearning: Fundamentals, Taxonomy, Evaluation, and Future Direction
• NeuralDB: Scaling Knowledge Editing in LLMs to 100,000 Facts with Neural KV Database
• PropMEND: Hypernetworks for Knowledge Propagation in LLMs
• An Exploration of Knowledge Editing for Arabic
• Know-MRI: A Knowledge Mechanisms Revealer&Interpreter for Large Language Models
• Knowledge Fusion via Bidirectional Information Aggregation
• Towards a Principled Evaluation of Knowledge Editors
• LoRA-Augmented Generation (LAG) for Knowledge-Intensive Language Tasks
• QueueEDIT: Structural Self-Correction for Sequential Model Editing in LLMs
• MEMOIR: Lifelong Model Editing with Minimal Overwrite and Informed Retention for LLMs
• Instruction Tuning with and without Context: Behavioral Shifts and Downstream Impact
• Model Merging for Knowledge Editing
• MobiEdit: Resource-efficient Knowledge Editing for Personalized On-device LLMs
• Osiris: A Lightweight Open-Source Hallucination Detection System
• Beyond Memorization: A Rigorous Evaluation Framework for Medical Knowledge Editing
• Zero-Shot Open-Schema Entity Structure Discovery
• KARE-RAG: Knowledge-Aware Refinement and Enhancement for RAG
• ImpRAG: Retrieval-Augmented Generation with Implicit Queries
• One for All: Update Parameterized Knowledge Across Multiple Models
• COMPKE: Complex Question Answering under Knowledge Editing
• GuideX: Guided Synthetic Data Generation for Zero-Shot Information Extraction
• Context-Robust Knowledge Editing for Language Models
• InComeS: Integrating Compression and Selection Mechanisms into LLMs for Efficient Model Editing
• DocMEdit: Towards Document-Level Model Editing
• Benchmarking and Rethinking Knowledge Editing for Large Language Models
• LyapLock: Bounded Knowledge Preservation in Sequential Large Language Model Editing
• UltraEdit: Training-, Subject-, and Memory-Free Lifelong Editing in Language Models
• EAMET: Robust Massive Model Editing via Embedding Alignment Optimization
• PT-MoE: An Efficient Finetuning Framework for Integrating Mixture-of-Experts into Prompt Tuning
• On the Superimposed Noise Accumulation Problem in Sequential Knowledge Editing of Large Language Models
• Memorization and Knowledge Injection in Gated LLMs
• Reversible Lifelong Model Editing via Semantic Routing-Based LoRA
• Reverse-Engineering Model Editing on Language Models
• How Effective are Generative Large Language Models in Performing Requirements Classification?
• GROM: Gradient-Free Rapid One-Shot Machine Unlearning
• How new data permeates LLM knowledge and how to dilute it
• Efficient Tuning of Large Language Models for Knowledge-Grounded Dialogue Generation

Related

• ZS-BERT: Towards Zero-Shot Relation Extraction with Attribute Representation Learning
• RelationPrompt: Leveraging Prompts to Generate Synthetic Data for Zero-Shot Relation Triplet Extraction
• Grasping the Essentials: Tailoring Large Language Models for Zero-Shot Relation Extraction
• Weakly-Supervised Questions for Zero-Shot Relation Extraction
• FewRel: A Large-Scale Supervised Few-Shot Relation Classification Dataset with State-of-the-Art Evaluation