Unified Structure Generation for Universal Information Extraction

By Yaojie Lu, Qing Liu, Dai Dai, Xinyan Xiao, Hongyu Lin, Xianpei Han, Le Sun, Hua Wu. Annual Meeting of the Association for Computational Linguistics, 2022

620 citations recorded by the source corpus.

Information extraction suffers from its varying targets, heterogeneous structures, and demand-specific schemas. In this paper, we propose a unified text-to-structure generation framework, namely UIE, which can universally model different IE tasks, adaptively generate targeted structures, and collaboratively learn general IE abilities from different knowledge sources. Specifically, UIE uniformly encodes different extraction structures via a structured extraction language, adaptively generates target extractions via a schema-based prompt mechanism – structural schema instructor, and captures the common IE abilities via a large-scale pretrained text-to-structure model. Experiments show that UIE achieved the state-of-the-art performance on 4 IE tasks, 13 datasets, and on all supervised, low-resource, and few-shot settings for a wide range of entity, relation, event and sentiment extraction tasks and their unification. These results verified the effectiveness, universality, and transferability of UIE.

Subfields: Artificial Intelligence, Information Systems

View Paper: https://www.semanticscholar.org/paper/5eb9dad01d9b61947d7a9aae6dfff400d20555b9

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

1. Lichao Sun
2. Philip S. Yu
3. Wenjuan Han
4. Jingren Zhou
5. Xiaobin Wang
6. Zheng Yuan
7. Hongyi Yuan
8. Yong Jiang

In Information Systems

1. Philip S. Yu
2. Jingren Zhou
3. Meishan Zhang
4. Wei Wang
5. Lichao Sun
6. Yongfeng Zhang
7. Jiawei Han
8. B. Liu

Citations in the corpus

• A Comprehensive Survey of AI-Generated Content (AIGC): A History of Generative AI from GAN to ChatGPT (2023, background): “[173] propose a unified text-to-graph framework which incorporates most information extraction tasks.”
• Large language models for generative information extraction: a survey (2023, neutral): “It is also a fundamental task in other NLP tasks, thus attracting significant attention from researchers to explore new possibilities in the era of LLMs [47, 90, 91, 92, 93, 94, 95, 108, 122, 124, 125, 126, 164, 165, 172, 173, 177, 179, 182].”
• Zero-Shot Information Extraction via Chatting with ChatGPT (2023, background): “Only when both the border and the type of the predicted entity and the true entity are the same will we regard it as a correct prediction.”
• How Abilities in Large Language Models are Affected by Supervised Fine-tuning Data Composition (2023, background): “Among the tasks, LLMs especially emerge with three out-standing abilities in reasoning (Cobbe et al., 2021; Wei et al., 2022), coding (Chen et al., 2021), and aligning general human intentions (Ouyang et al., 2022a), which have drawn much attention from the LLM research community.”