Analyzing BERT Cross-lingual Transfer Capabilities in Continual Sequence Labeling - Information, Langue Ecrite et Signée
Conference Papers Year : 2022

Analyzing BERT Cross-lingual Transfer Capabilities in Continual Sequence Labeling

Juan Manuel Coria
Mathilde Veron
Sahar Ghannay
Guillaume Bernard
  • Function : Author
Hervé Bredin
Olivier Galibert
Sophie Rosset

Abstract

Knowledge transfer between neural language models is a widely used technique that has proven to improve performance in a multitude of natural language tasks, in particular with the recent rise of large pre-trained language models like BERT. Similarly, high crosslingual transfer has been shown to occur in multilingual language models. Hence, it is of great importance to better understand this phenomenon as well as its limits. While most studies about cross-lingual transfer focus on training on independent and identically distributed (i.e. i.i.d.) samples, in this paper we study cross-lingual transfer in a continual learning setting on two sequence labeling tasks: slotfilling and named entity recognition. We investigate this by training multilingual BERT on sequences of 9 languages, one language at a time, on the MultiATIS++ and MultiCoNER corpora. Our first findings are that forward transfer between languages is retained although forgetting is present. Additional experiments show that lost performance can be recovered with as little as a single training epoch even if forgetting was high, which can be explained by a progressive shift of model parameters towards a better multilingual initialization. We also find that commonly used metrics might be insufficient to assess continual learning performance.
Fichier principal
Vignette du fichier
analyzing_bert.pdf (370.94 Ko) Télécharger le fichier
Origin Files produced by the author(s)

Dates and versions

hal-03824597 , version 1 (21-10-2022)

Identifiers

  • HAL Id : hal-03824597 , version 1

Cite

Juan Manuel Coria, Mathilde Veron, Sahar Ghannay, Guillaume Bernard, Hervé Bredin, et al.. Analyzing BERT Cross-lingual Transfer Capabilities in Continual Sequence Labeling. First Workshop on Performance and Interpretability Evaluations of Multimodal, Multipurpose, Massive-Scale Models, Oct 2022, virtual, South Korea. ⟨hal-03824597⟩
241 View
71 Download

Share

More