Structural Biology · arXiv:2307.01972
Open-Domain Hierarchical Event Schema Induction by Incremental Prompting and Verification
Authors: Sha Li, Ruining Zhao, Manling Li, Heng Ji, Chris Callison-Burch et al. (arXiv:2307.01972) · arXiv: 2307.01972
Core Methodological Innovation
Featured in Nujan's Structural Biology Literature Digest. Open side-by-side with interactive equation, methodology, and citation breakdowns.
Key Quantitative & Theoretical Takeaway: Open arXiv:2307.01972 in Nujan to inspect every mathematical symbol, experimental baseline, and citation directly alongside the original PDF.
Abstract
Event schemas are a form of world knowledge about the typical progression of events. Recent methods for event schema induction use information extraction systems to construct a large number of event graph instances from documents, and then learn to generalize the schema from such instances. In contrast, we propose to treat event schemas as a form of commonsense knowledge that can be derived from large language models (LLMs). This new paradigm greatly simplifies the schema induction process and allows us to handle both hierarchical relations and temporal relations between events in a straightforward way. Since event schemas have complex graph structures, we design an incremental prompting and verification method to break down the construction of a complex event graph into three stages: event skeleton construction, event expansion, and event-event relation verification. Compared to directly using LLMs to generate a linearized graph, our method can generate large and complex schemas with 7.2% F1 improvement in temporal relations and 31.0% F1 improvement in hierarchical relations. In addition, compared to the previous state-of-the-art closed-domain schema induction model, human assessors were able to cover $\sim$10% more events when translating the schemas into coherent stories and rated our schemas 1.3 points higher (on a 5-point scale) in terms of readability.
Step-by-Step Equation & Methodology Breakdown
What is Open-Domain Hierarchical Event Schema Induction by Incremental Prompting and Verification (arXiv:2307.01972) about?
Event schemas are a form of world knowledge about the typical progression of events. Recent methods for event schema induction use information extraction systems to construct a large number of event graph instances from documents, and then learn to generalize the schema from such instances. In contrast, we propose to treat event schemas as a form of commonsense knowledge that can be derived from large language models (LLMs). This new paradigm greatly simplifies the schema induction process and allows us to handle both hierarchical relations and temporal relations between events in a straightforward way. Since event schemas have complex graph structures, we design an incremental prompting and verification method to break down the construction of a complex event graph into three stages: event skeleton construction, event expansion, and event-event relation verification. Compared to directly using LLMs to generate a linearized graph, our method can generate large and complex schemas with 7.2% F1 improvement in temporal relations and 31.0% F1 improvement in hierarchical relations. In addition, compared to the previous state-of-the-art closed-domain schema induction model, human assessors were able to cover $\sim$10% more events when translating the schemas into coherent stories and rated our schemas 1.3 points higher (on a 5-point scale) in terms of readability.
How can I read arXiv:2307.01972 with interactive explanations?
Visit https://nujan.app/abs/2307.01972 to open Open-Domain Hierarchical Event Schema Induction by Incremental Prompting and Verification in Nujan's split-screen research workspace with instant equation and methodology breakdowns.