Understanding by Understanding Not: Modeling Negation in Language Models
Arian Hosseini Affiliation: Mila/Université de Montréal Affiliation: Montréal, Canada Email: Siva Reddy Affiliation: Mila/McGill University Affiliation: Montréal, Canada Dzmitry Bahdanau Affiliation: Element AI Affiliation: a ServiceNow Company Affiliation: Montréal, Canada R Devon Hjelm Affiliation: Mila/Université de Montréal Affiliation: and Microsoft Research Affiliation: Montréal, Canada Alessandro Sordoni Affiliation: Microsoft Research Affiliation: Montréal, Canada Aaron Courville Affiliation: Mila/Université de Montréal Affiliation: Montréal, Canada
Abstract
Negation is a core construction in natural language. Despite being very successful on many tasks, state-of-the-art pre-trained language models often handle negation incorrectly. To improve language models in this regard, we propose to augment the language modeling objective with an unlikelihood objective that is based on negated generic sentences from a raw text corpus. By training BERT with the resulting combined objective we reduce the mean top 1 error rate to 4% on the negated LAMA dataset. We also see some improvements on the negated NLI benchmarks.
原文 arXiv:2105.03519;中英对照 + 大白话阅读 https://aha.fim.ai/paper/2105.03519v1