Different Tokenization Schemes Lead to Comparable Performance in Spanish Number Agreement
This addresses tokenization impact on language model performance in Spanish morphology, but is incremental as it confirms existing patterns without major breakthroughs.
The study investigated how different tokenization schemes affect number agreement in Spanish plurals for language models, finding that morphologically-aligned tokenization performs similarly to other schemes, with models generalizing morphological patterns to new items without strict requirement for performance.
The relationship between language model tokenization and performance is an open area of research. Here, we investigate how different tokenization schemes impact number agreement in Spanish plurals. We find that morphologically-aligned tokenization performs similarly to other tokenization schemes, even when induced artificially for words that would not be tokenized that way during training. We then present exploratory analyses demonstrating that language model embeddings for different plural tokenizations have similar distributions along the embedding space axis that maximally distinguishes singular and plural nouns. Our results suggest that morphologically-aligned tokenization is a viable tokenization approach, and existing models already generalize some morphological patterns to new items. However, our results indicate that morphological tokenization is not strictly required for performance.