Understanding by Understanding Not: Modeling Negation in Language Models
This addresses a core linguistic challenge for natural language processing, but it is incremental as it builds on existing models and objectives.
The paper tackled the problem of language models incorrectly handling negation by augmenting the language modeling objective with an unlikelihood objective based on negated sentences, reducing the mean top-1 error rate to 4% on the negated LAMA dataset.
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.