Summary of Cascading Large Language Models For Salient Event Graph Generation, by Xingwei Tan et al.
Cascading Large Language Models for Salient Event Graph Generation
by Xingwei Tan, Yuxiang Zhou, Gabriele Pergola, Yulan He
First submitted to arxiv on: 26 Jun 2024
Categories
- Main: Computation and Language (cs.CL)
- Secondary: Artificial Intelligence (cs.AI)
GrooveSquid.com Paper Summaries
GrooveSquid.com’s goal is to make artificial intelligence research accessible by summarizing AI papers in simpler terms. Each summary below covers the same AI paper, written at different levels of difficulty. The medium difficulty and low difficulty versions are original summaries written by GrooveSquid.com, while the high difficulty version is the paper’s original abstract. Feel free to learn from the version that suits you best!
| Summary difficulty | Written by | Summary |
|---|---|---|
| High | Paper authors | High Difficulty Summary Read the original abstract here |
| Medium | GrooveSquid.com (original content) | Medium Difficulty Summary The paper proposes CALLMSAE, a framework for generating event graphs from long documents, which addresses the limitations of existing methods by identifying salient events. The approach uses large language models (LLMs) to generate summaries, from which salient events are identified, and then develops an iterative code refinement prompting strategy to generate event relation graphs. The paper also presents NYT-SEG, a large-scale automatically annotated event graph dataset that can serve as distant supervision signals. Fine-tuning contextualized graph generation models on this dataset outperforms models trained on CAEVO data, achieving better results on a human-annotated test set. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary The paper creates a new way to make event graphs from long documents. This is hard because there are many tasks involved like finding events, seeing how they relate to each other, and matching unstructured text with structured graphs. The current methods don’t do a good job of distinguishing important events that help us understand stories. CALLMSAE uses large language models to find the most important events and then makes event relation graphs using an iterative process. They also created a big dataset called NYT-SEG that can be used to train models. This new approach does better than existing methods on a test set. |
Keywords
* Artificial intelligence * Fine tuning * Prompting
