Paraphrasing in Affirmative Terms Improves Negation Understanding
This addresses a common linguistic challenge for natural language processing applications, but the approach appears incremental as it builds on existing paraphrasing techniques.
The paper tackles the problem of language models struggling with negation in natural language understanding tasks by using automatically generated affirmative paraphrases, showing improvements on CondaQA and five other tasks.
Negation is a common linguistic phenomenon. Yet language models face challenges with negation in many natural language understanding tasks such as question answering and natural language inference. In this paper, we experiment with seamless strategies that incorporate affirmative interpretations (i.e., paraphrases without negation) to make models more robust against negation. Crucially, our affirmative interpretations are obtained automatically. We show improvements with CondaQA, a large corpus requiring reasoning with negation, and five natural language understanding tasks.