What makes a language easy to deep-learn? Deep neural networks and humans similarly benefit from compositional structure

TL;DR

This study compares neural networks and humans in learning languages with varying compositional structures, showing neural networks learn more systematically from more structured languages, similar to human patterns.

Contribution

It demonstrates that deep neural networks, like humans, find languages with higher compositionality easier to learn and generalize, highlighting the importance of structure in language learning models.

Findings

Neural networks generalize better with more compositional languages

Neural networks show greater agreement across agents for structured languages

Neural networks' learning patterns resemble human language acquisition

Abstract

Deep neural networks drive the success of natural language processing. A fundamental property of language is its compositional structure, allowing humans to systematically produce forms for new meanings. For humans, languages with more compositional and transparent structures are typically easier to learn than those with opaque and irregular structures. However, this learnability advantage has not yet been shown for deep neural networks, limiting their use as models for human language learning. Here, we directly test how neural networks compare to humans in learning and generalizing different languages that vary in their degree of compositional structure. We evaluate the memorization and generalization capabilities of a large language model and recurrent neural networks, and show that both deep neural networks exhibit a learnability advantage for more structured linguistic input: neural networks exposed to more compositional languages show more systematic generalization, greater agreement between different agents, and greater similarity to human learners.