A Cognitive Regularizer for Language Modeling

TL;DR

This paper introduces a regularizer based on the uniform information density hypothesis to improve language modeling, demonstrating enhanced perplexity and lexical diversity across multiple languages, especially with limited data.

Contribution

It operationalizes the UID hypothesis as an inductive bias in language models, showing consistent perplexity improvements and more diverse text generation.

Findings

UID regularization improves perplexity across ten languages.

Enhanced lexical diversity in UID-regularized models.

Greater benefits observed with limited training data.

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 language modeling. Specifically, we augment the canonical MLE objective for training language models 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 language models, having a larger effect when training data is limited. Moreover, via an analysis of generated sequences, we find that UID-regularized language models 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 language modeling, but also provide an alternative validation of the UID hypothesis using modern-day NLP tools.