Evaluating Morphological Compositional Generalization in Large Language Models

TL;DR

This paper systematically evaluates how well large language models understand and generalize morphological structures, revealing significant limitations in their ability to handle novel words and complex morphology in agglutinative languages.

Contribution

The study introduces a novel suite of tasks to assess morphological compositionality in LLMs, highlighting their struggles with systematic generalization in agglutinative languages.

Findings

LLMs perform poorly on morphological generalization with novel roots

Models show limited systematicity in morphological composition

Performance declines with increased morphological complexity

Abstract

Large language models (LLMs) have demonstrated significant progress in various natural language generation and understanding tasks. However, their linguistic generalization capabilities remain questionable, raising doubts about whether these models learn language similarly to humans. While humans exhibit compositional generalization and linguistic creativity in language use, the extent to which LLMs replicate these abilities, particularly in morphology, is under-explored. In this work, we systematically investigate the morphological generalization abilities of LLMs through the lens of compositionality. We define morphemes as compositional primitives and design a novel suite of generative and discriminative tasks to assess morphological productivity and systematicity. Focusing on agglutinative languages such as Turkish and Finnish, we evaluate several state-of-the-art instruction-finetuned multilingual models, including GPT-4 and Gemini. Our analysis shows that LLMs struggle with morphological compositional generalization particularly when applied to novel word roots, with performance declining sharply as morphological complexity increases. While models can identify individual morphological combinations better than chance, their performance lacks systematicity, leading to significant accuracy gaps compared to humans.