Assessing the Importance of Frequency versus Compositionality for Subword-based Tokenization in NMT

TL;DR

This paper investigates the relative importance of frequency and compositionality in subword tokenization for neural machine translation, finding that frequency alone explains most of the performance gains attributed to subword methods.

Contribution

The study introduces a Huffman coding-based tokenization method that isolates frequency from compositionality, revealing frequency's dominant role in translation quality.

Findings

Frequency accounts for 90-95% of BPE performance.

Compositionality has less impact than previously believed.

Subword frequency is the primary factor in effective tokenization.

Abstract

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.