Federated learning for exploiting annotators? Disagreements in natural language processing

Rodríguez-Barroso, Nuria, Cámara, Eugenio Martínez, Collados, Jose Camacho ORCID: https://orcid.org/0000-0003-1618-7239, Luzón, M. Victoria and Herrera, Francisco 2024. Federated learning for exploiting annotators? Disagreements in natural language processing. Transactions of the Association for Computational Linguistics 12 , 630–648. 10.1162/tacl_a_00664

PDF - Published Version Available under License Creative Commons Attribution. Download (647kB)

Abstract

The annotation of ambiguous or subjective NLP tasks is usually addressed by various annotators. In most datasets, these annotations are aggregated into a single ground truth. However, this omits divergent opinions of annotators, hence missing individual perspectives. We propose FLEAD (Federated Learning for Exploiting Annotators’ Disagreements), a methodology built upon federated learning to independently learn from the opinions of all the annotators, thereby leveraging all their underlying information without relying on a single ground truth. We conduct an extensive experimental study and analysis in diverse text classification tasks to show the contribution of our approach with respect to mainstream approaches based on majority voting and other recent methodologies that also learn from annotator disagreements.

Item Type:	Article
Date Type:	Published Online
Status:	Published
Schools:	Computer Science & Informatics
Publisher:	Massachusetts Institute of Technology Press
ISSN:	2307-387X
Date of First Compliant Deposit:	8 August 2024
Last Modified:	08 Aug 2024 11:18
URI:	https://orca.cardiff.ac.uk/id/eprint/171076

Actions (repository staff only)

Edit Item