Vous êtes tous cordialement invités à assister au séminaire au DIC, session hiver 2022, en partenariat avec l'ISC et le CRIA.

Abbas GHADDAR – 17 mars 2022 à 10h30   

Titre : Context-aware Adversarial Training for Name Regularity Bias in Named Entity Recognition   

Résumé :

In this work, we examine the ability of NER models to use contextual information when predicting the type of an ambiguous entity. We introduce NRB, a new testbed carefully designed to diagnose Name Regularity Bias of NER models. Our results indicate that all state-of-the-art models we tested show such a bias; BERT fine-tuned models significantly outperforming feature-based (LSTM-CRF) ones on NRB, despite having comparable (sometimes lower) performance on standard benchmarks.To mitigate this bias, we propose a novel model-agnostic training method that adds learnable adversarial noise to some entity mentions, thus enforcing models to focus more strongly on the contextual signal, leading to significant gains on NRB. Combining it with two other training strategies, data augmentation and parameter freezing, leads to further gains.

Bio :

Abbas GHADDAR is a Senior Researcher at Huawei. Abbas received his Master's and Ph.D. degrees in 2016 and 2020, respectively, from the University of Montreal under the supervision of Professor Philippe Langlais. Since 2020, he works as a researcher at Huawei Noah Ark Lab located in Montreal. His main research interest is machine learning methods applied to natural language processing, model compression, robustness, and generalization. He has publications at top conferences and journals, including ACL, ACL, EMNLP, CoNLL, COLING, NeruIPS.

S.V.P. Vous connecter au moins 10 à 15 minutes avant l'heure et inscrire votre nom complet pour aider à vous admettre au séminaire. 

La participation à micro et caméra ouverte est grandement appréciée lors de la période des questions.

Lien Zoom de la rencontre : https://uqam.zoom.us/j/85407268175