Finding Fuzziness in Neural Network Models of Language Processing

TL;DR

This paper investigates whether neural network language models naturally encode fuzzy concepts like 'hot' or 'cool' by analyzing their responses in a natural language inference task, revealing promising but noisy correlations.

Contribution

The study demonstrates that state-of-the-art language models exhibit fuzzy-membership patterns similar to classical fuzzy logic, highlighting their potential to represent imprecise linguistic concepts.

Findings

Models show fuzzy-set like patterns in temperature perception

Responses are similar to classical fuzzy logic formulations

Significant noise indicates room for improvement

Abstract

Humans often communicate by using imprecise language, suggesting that fuzzy concepts with unclear boundaries are prevalent in language use. In this paper, we test the extent to which models trained to capture the distributional statistics of language show correspondence to fuzzy-membership patterns. Using the task of natural language inference, we test a recent state of the art model on the classical case of temperature, by examining its mapping of temperature data to fuzzy-perceptions such as "cool", "hot", etc. We find the model to show patterns that are similar to classical fuzzy-set theoretic formulations of linguistic hedges, albeit with a substantial amount of noise, suggesting that models trained solely on language show promise in encoding fuzziness.