Publication Details

IDIAPers @ Causal News Corpus 2022: Efficient Causal Relation Identification Through a Prompt-based Few-shot Approach

BURDISSO Sergio, FAJČÍK Martin, SMRŽ Pavel and MOTLÍČEK Petr et al. IDIAPers @ Causal News Corpus 2022: Efficient Causal Relation Identification Through a Prompt-based Few-shot Approach. In: Proceedings of the 5th Workshop on Challenges and Applications of Automated Extraction of Socio-political Events from Text (CASE 2022). Abu Dhabi: Association for Computational Linguistics, 2022, pp. 61-69. ISBN 978-1-959429-05-0. Available from: https://aclanthology.org/2022.case-1.9/
Czech title
IDIAPers @ Causal News Corpus 2022: Efektivní identifikace kauzálních vztahů prostřednictvím příkazů založených na "few-shot" učení
Type
conference paper
Language
english
Authors
Burdisso Sergio (IDIAP)
Fajčík Martin, Ing., Ph.D. (DCGM FIT BUT)
Smrž Pavel, doc. RNDr., Ph.D. (DCGM FIT BUT)
and others
URL
Keywords

few-shot learning, classifier, causal relation identification, causal event identification, ensembling

Abstract

In this paper, we describe our participation in the subtask 1 of CASE-2022, Event Causality Identification with Casual News Corpus. We address the Causal Relation Identification (CRI) task by exploiting a set of simple yet complementary techniques for fine-tuning language models (LMs) on a small number of annotated examples (i.e., a few-shot configuration). We follow a prompt-based prediction approach for fine-tuning LMs in which the CRI task is treated as a masked language modeling problem (MLM). This approach allows LMs natively pre-trained on MLM problems to directly generate textual responses to CRI-specific prompts. We compare the performance of this method against ensemble techniques trained on the entire dataset. Our best-performing submission was fine-tuned with only 256 instances per class, 15.7% of the all available data, and yet obtained the second-best precision (0.82), third-best accuracy (0.82), and an F1-score (0.85) very close to what was reported by the winner team (0.86).

Published
2022
Pages
61-69
Proceedings
Proceedings of the 5th Workshop on Challenges and Applications of Automated Extraction of Socio-political Events from Text (CASE 2022)
Conference
Conference on Empirical Methods in Natural Language Processing, Abu Dhabi, AE
ISBN
978-1-959429-05-0
Publisher
Association for Computational Linguistics
Place
Abu Dhabi, AE
DOI
EID Scopus
BibTeX
@INPROCEEDINGS{FITPUB12839,
   author = "Sergio Burdisso and Martin Faj\v{c}\'{i}k and Pavel Smr\v{z} and Petr Motl\'{i}\v{c}ek and et al.",
   title = "IDIAPers @ Causal News Corpus 2022: Efficient Causal Relation Identification Through a Prompt-based Few-shot Approach",
   pages = "61--69",
   booktitle = "Proceedings of the 5th Workshop on Challenges and Applications of Automated Extraction of Socio-political Events from Text (CASE 2022)",
   year = 2022,
   location = "Abu Dhabi, AE",
   publisher = "Association for Computational Linguistics",
   ISBN = "978-1-959429-05-0",
   doi = "10.18653/v1/2022.case-1.9",
   language = "english",
   url = "https://www.fit.vut.cz/research/publication/12839"
}
Back to top