Skip to main navigation Skip to search Skip to main content

Exploring a POS-based Two-stage Approach for Improving Low-Resource AMR-to-Text Generation

  • Marco Antonio Sobrevilla Cabezudo
  • , Thiago Alexandre Salgueiro Pardo

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

1 Scopus citations

Abstract

This work presents a two-stage approach for tackling low-resource AMR-to-text generation for Brazilian Portuguese. Our approach consists of (1) generating a masked surface realization in which some tokens are masked according to its Part-of-Speech class and (2) infilling the masked tokens according to the AMR graph and the previous masked surface realization. Results show a slight improvement over the baseline, mainly in BLEU (1.63) and METEOR (0.02) scores. Moreover, we evaluate the pipeline components separately, showing that the bottleneck of the pipeline is the masked surface realization. Finally, the human revision suggests that models still suffer from hallucinations, and some strategies to deal with the problems found are proposed.

Original languageEnglish
Title of host publicationGEM 2022 - 2nd Workshop on Natural Language Generation, Evaluation, and Metrics, Proceedings of the Workshop
PublisherAssociation for Computational Linguistics (ACL)
Pages531-538
Number of pages8
ISBN (Electronic)9781959429128
StatePublished - 2022
Externally publishedYes
Event2nd Workshop on Natural Language Generation, Evaluation, and Metrics, GEM 2022, as part of EMNLP 2022 - Abu Dhabi, United Arab Emirates
Duration: 7 Dec 2022 → …

Publication series

NameGEM 2022 - 2nd Workshop on Natural Language Generation, Evaluation, and Metrics, Proceedings of the Workshop

Conference

Conference2nd Workshop on Natural Language Generation, Evaluation, and Metrics, GEM 2022, as part of EMNLP 2022
Country/TerritoryUnited Arab Emirates
CityAbu Dhabi
Period7/12/22 → …

Fingerprint

Dive into the research topics of 'Exploring a POS-based Two-stage Approach for Improving Low-Resource AMR-to-Text Generation'. Together they form a unique fingerprint.

Cite this