Set Generation Networks for End-to-End Knowledge Base Population

Sui, Dianbo; Wang, Chenhao; Chen, Yubo; Liu, Kang; Zhao, Jun; Bi, Wei

doi:10.18653/v1/2021.emnlp-main.760

Cited by 8 publications

(7 citation statements)

References 34 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…4, we report Ge-nIE and the pipeline baseline F1 for schemas with 100, 400, and 857 relations. To choose a subset of n relations, we take the n most frequent relations to mimic the strategies used by previous works to reduce the schemas (Sui et al, 2021). We first observe that GenIE is always largely better than the baseline.…”

Section: F1 Analysis Of Performance As a Function Of The Number Of Re...mentioning

confidence: 98%

“…End-to-end systems that jointly perform the extraction and the disambiguation of entities and relations have been proposed to address the error propagation (Trisedya et al, 2019;Sui et al, 2021;. To mitigate the propagation of errors, these systems are endowed with the ability to leverage entity information in the relation extraction and vice-versa, which has resulted in significant performance gains.…”

Section: Closed Information Extractionmentioning

confidence: 99%

“…We compare GenIE against Set Generation Networks (SetGenNet; Sui et al, 2021) which is, to the best of our knowledge, the strongest model on Wiki-NRE and Geo-NRE. Note that the authors did not release code or the model and there is no other model from the literature trained and evaluated on REBEL for cIE. SetGenNet (Sui et al, 2021) is an end-to-end state-of-the-art model for triplet extraction. It consists of a transformer encoder (Vaswani et al, 2017) that encodes the input followed by a non-autoregressive transformer decoder (Gu et al, 2018).…”

Section: Baselinesmentioning

confidence: 99%

“…Unfortunately, most works do not have available or directly usable code. In particular, we first concentrated on SetGenNet (Sui et al, 2021) as, to the best of our knowledge, it is the strongest model on the task of cIE. However, the authors do not report a link to the code 6 in neither the arXiv nor the ACL Antology version of the paper.…”

Section: Note On End-to-end Baselinesmentioning

confidence: 99%

See 3 more Smart Citations

GenIE: Generative Information Extraction

Josifoski¹,

Cao²,

Peyrard³

et al. 2022

Proceedings of the 2022 Conference of the North American Chapter of the Association for Computational Linguistics: Human Langua

View full text Add to dashboard Cite

Structured and grounded representation of text is typically formalized by closed information extraction, the problem of extracting an exhaustive set of (subject, relation, object) triplets that are consistent with a predefined set of entities and relations from a knowledge base schema. Most existing works are pipelines prone to error accumulation, and all approaches are only applicable to unrealistically small numbers of entities and relations. We introduce GenIE (generative information extraction), the first end-toend autoregressive formulation of closed information extraction. GenIE naturally exploits the language knowledge from the pre-trained transformer by autoregressively generating relations and entities in textual form. Thanks to a new bi-level constrained generation strategy, only triplets consistent with the predefined knowledge base schema are produced. Our experiments show that GenIE is state-of-the-art on closed information extraction, generalizes from fewer training data points than baselines, and scales to a previously unmanageable number of entities and relations. With this work, closed information extraction becomes practical in realistic scenarios, providing new opportunities for downstream tasks. Finally, this work paves the way towards a unified end-to-end approach to the core tasks of information extraction.

show abstract

Section: F1 Analysis Of Performance As a Function Of The Number Of Re...mentioning

confidence: 98%

Section: Closed Information Extractionmentioning

confidence: 99%

Section: Baselinesmentioning

confidence: 99%

Section: Note On End-to-end Baselinesmentioning

confidence: 99%

See 2 more Smart Citations

GenIE: Generative Information Extraction

Josifoski¹,

Cao²,

Peyrard³

et al. 2022

Proceedings of the 2022 Conference of the North American Chapter of the Association for Computational Linguistics: Human Langua

View full text Add to dashboard Cite

show abstract

“…Event extraction aims at structuring the event information in unstructured text. It is widely used in text summarization [2], question answering [3] and knowledge base population [4]. As the sub task of event extraction, event detection is an important research field of natural language processing, which aims to identify event trigger words from a given text and classify them into the correct categories.…”

Section: Introductionmentioning

confidence: 99%

Novel Robust Span-based Model for Event Detection

Zhang

Yuan

Song

et al. 2023

J. Phys.: Conf. Ser.

View full text Add to dashboard Cite

Event detection (ED) aims to identify events and their types from the given text, which is the vital subtask of event extraction. However, poor quality training corpus can seriously affect the validity of current dominating deep learning-based techniques. In order to improve the robustness of the model, this paper converts the ED task into the named entity recognition (NER) task, and uses the span-based NER framework to identify and classify event triggers. In addition, we ameliorate the framework through data cleaning strategy, hierarchical learning rate strategy, sentence-level loss function and model ensemble strategy. The experimental results suggest that our approach achieves state-of-the-art performance on the widely used Chinese dataset. Moreover, the deep analysis has confirmed the effectiveness of the improvement strategy in resisting data pollution.

show abstract

Knowledge Graph Population with Out-of-KG Entities

Cedric

2022

The Semantic Web: ESWC 2022 Satellite Events

View full text Add to dashboard Cite

Set Generation Networks for End-to-End Knowledge Base Population

Cited by 8 publications

References 34 publications

GenIE: Generative Information Extraction

GenIE: Generative Information Extraction

Novel Robust Span-based Model for Event Detection

Knowledge Graph Population with Out-of-KG Entities

Contact Info

Product

Resources

About