Improving Information Extraction by Acquiring External Evidence with\n Reinforcement Learning

2016/03/25 by Karthik Narasimhan, Adam Yala, Narasimhan, Karthik +3 · 5 citations
Computer Science · Decision Sciences · #Computation and Language (cs.CL) #Data Quality and Management #FOS: Computer and information sciences #Topic Modeling #Web Data Mining and Analysis

paper · pdf · doi:10.48550/arxiv.1603.07954

openalex publication_date 2016/03/25 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/28

Abstract

Most successful information extraction systems operate with access to a large\ncollection of documents. In this work, we explore the task of acquiring and\nincorporating external evidence to improve extraction accuracy in domains where\nthe amount of training data is scarce. This process entails issuing search\nqueries, extraction from new sources and reconciliation of extracted values,\nwhich are repeated until sufficient evidence is collected. We approach the\nproblem using a reinforcement learning framework where our model learns to\nselect optimal actions based on contextual information. We employ a deep\nQ-network, trained to optimize a reward function that reflects extraction\naccuracy while penalizing extra effort. Our experiments on two databases -- of\nshooting incidents, and food adulteration cases -- demonstrate that our system\nsignificantly outperforms traditional extractors and a competitive\nmeta-classifier baseline.\n

Citations

• Deep Reinforcement Learning with a Natural Language Action Space
• Language Understanding for Text-based Games Using Deep Reinforcement\n Learning
• Compositional Vector Space Models for Knowledge Base Completion

Cited by

• Learning how to Active Learn: A Deep Reinforcement Learning Approach
• A Coarse to Fine Question Answering System based on Reinforcement Learning
• A Hierarchical Framework for Relation Extraction with Reinforcement Learning
• NLPGym -- A toolkit for evaluating RL agents on Natural Language Processing Tasks
• Event Extraction with Generative Adversarial Imitation Learning

Related

• Learning to Extract Rational Evidence via Reinforcement Learning for Retrieval-Augmented Generation
• Knowledge-guided Open Attribute Value Extraction with Reinforcement Learning
• Adaptive Reinforcement Learning Planning: Harnessing Large Language Models for Complex Information Extraction
• Extract-0: A Specialized Language Model for Document Information Extraction
• Towards Information-Seeking Agents