A Cognitive Regularizer for Language Modeling (2105.07144v3)

Abstract: The uniform information density (UID) hypothesis, which posits that speakers behaving optimally tend to distribute information uniformly across a linguistic signal, has gained traction in psycholinguistics as an explanation for certain syntactic, morphological, and prosodic choices. In this work, we explore whether the UID hypothesis can be operationalized as an inductive bias for statistical LLMing. Specifically, we augment the canonical MLE objective for training LLMs with a regularizer that encodes UID. In experiments on ten languages spanning five language families, we find that using UID regularization consistently improves perplexity in LLMs, having a larger effect when training data is limited. Moreover, via an analysis of generated sequences, we find that UID-regularized LLMs have other desirable properties, e.g., they generate text that is more lexically diverse. Our results not only suggest that UID is a reasonable inductive bias for LLMing, but also provide an alternative validation of the UID hypothesis using modern-day NLP tools.