COMET: Commonsense Transformers for Automatic Knowledge Graph\n Construction

2019/06/12 by Antoine Bosselut, Hannah Rashkin, Bosselut, Antoine +9 · 14 citations
Biochemistry, Genetics and Molecular Biology · Computer Science · #Artificial Intelligence (cs.AI) #Biomedical Text Mining and Ontologies #Computation and Language (cs.CL) #FOS: Computer and information sciences #Natural Language Processing Techniques #Topic Modeling

paper · pdf · doi:10.48550/arxiv.1906.05317

openalex publication_date 2019/06/12 · openalex created_date 2022/07/28 · openalex updated_date 2026/07/28

Abstract

We present the first comprehensive study on automatic knowledge base\nconstruction for two prevalent commonsense knowledge graphs: ATOMIC (Sap et\nal., 2019) and ConceptNet (Speer et al., 2017). Contrary to many conventional\nKBs that store knowledge with canonical templates, commonsense KBs only store\nloosely structured open-text descriptions of knowledge. We posit that an\nimportant step toward automatic commonsense completion is the development of\ngenerative models of commonsense knowledge, and propose COMmonsEnse\nTransformers (COMET) that learn to generate rich and diverse commonsense\ndescriptions in natural language. Despite the challenges of commonsense\nmodeling, our investigation reveals promising results when implicit knowledge\nfrom deep pre-trained language models is transferred to generate explicit\nknowledge in commonsense knowledge graphs. Empirical results demonstrate that\nCOMET is able to generate novel knowledge that humans rate as high quality,\nwith up to 77.5% (ATOMIC) and 91.7% (ConceptNet) precision at top 1, which\napproaches human performance for these resources. Our findings suggest that\nusing generative commonsense models for automatic commonsense KB completion\ncould soon be a plausible alternative to extractive methods.\n

Citations

• YAGO2: A spatially and temporally enhanced knowledge base from Wikipedia
• Sequence to Sequence Learning with Neural Networks

Cited by

• Social Story Frames: Contextual Reasoning about Narrative Intent and Reception
• VisKnow: Constructing Visual Knowledge Base for Object Understanding
• YOLOA: Real-Time Affordance Detection via LLM Adapter
• OpenGloss: A Synthetic Encyclopedic Dictionary and Semantic Knowledge Graph
• Acquiring Common Chinese Emotional Events Using Large Language Model
• ReMindRAG: Low-Cost LLM-Guided Knowledge Graph Traversal for Efficient RAG
• Holistic Order Prediction in Natural Scenes
• DITING: A Multi-Agent Evaluation Framework for Benchmarking Web Novel Translation
• Who's Laughing Now? An Overview of Computational Humour Generation and Explanation
• MIXRAG : Mixture-of-Experts Retrieval-Augmented Generation for Textual Graph Understanding and Question Answering
• TComQA: Extracting Temporal Commonsense from Text
• ColorGPT: Leveraging Large Language Models for Multimodal Color Recommendation
• Calibrate Before Use: Improving Few-Shot Performance of Language Models
• Knowledge-Aware Self-Correction in Language Models via Structured Memory Graphs

Related

• KG-BART: Knowledge Graph-Augmented BART for Generative Commonsense Reasoning
• Symbolic Knowledge Distillation: from General Language Models to Commonsense Models
• On the Role of Conceptualization in Commonsense Knowledge Graph Construction
• Commonsense Knowledge Base Completion with Structural and Semantic\n Context
• Commonsense Knowledge in Word Associations and ConceptNet