Subword tokenization is the de facto standard for tokenization in neural language models and machine translation systems. Three advantages are frequently cited in favor of subwords: shorter encoding of frequent tokens, compositionality of subwords, and ability to deal with unknown words. As their relative importance is not entirely clear yet, we propose a tokenization approach that enables us to separate frequency (the first advantage) from compositionality. The approach uses Huffman coding to tokenize words, by order of frequency, using a fixed amount of symbols. Experiments with CS-DE, EN-FR and EN-DE NMT show that frequency alone accounts for 90%-95% of the scores reached by BPE, hence compositionality has less importance than previously thought.
@article{arxiv.2306.01393,
title = {Assessing the Importance of Frequency versus Compositionality for Subword-based Tokenization in NMT},
author = {Benoist Wolleb and Romain Silvestri and Giorgos Vernikos and Ljiljana Dolamic and Andrei Popescu-Belis},
journal= {arXiv preprint arXiv:2306.01393},
year = {2024}
}