Performance of two French BERT models for French language on verbatim transcripts and online posts - Archive ouverte HAL Accéder directement au contenu
Communication Dans Un Congrès Année : 2022

Performance of two French BERT models for French language on verbatim transcripts and online posts

Résumé

Pre-trained models based on the Transformer architecture have achieved notable performances in various language processing tasks. This article presents a comparison of two pretrained versions for French in a three-class classification task. The datasets used are of two types: a set of annotated verbatim transcripts from face-to-face interviews conducted during a market study and a set of online posts extracted from a community platform. Little work has been done in these two areas with transcribed oral corpora and online posts in French.
Fichier principal
Vignette du fichier
Performance_of_two_French_BERT_models_for_French_language_on_verbatim_transcripts_and_online_posts.pdf (111.91 Ko) Télécharger le fichier
Origine : Fichiers produits par l'(les) auteur(s)

Dates et versions

hal-03891881 , version 1 (09-12-2022)

Identifiants

  • HAL Id : hal-03891881 , version 1

Citer

Emmanuelle Kelodjoue, Jérôme Goulian, Didier Schwab. Performance of two French BERT models for French language on verbatim transcripts and online posts. 5th International Conference on Natural Language and Speech Processing, Dec 2022, Trento, Italy. ⟨hal-03891881⟩
28 Consultations
28 Téléchargements

Partager

Gmail Facebook X LinkedIn More