Evaluating Subword Tokenization: Alien Subword Composition and OOV Generalization Challenge
This addresses the need for better evaluation methods for subword tokenization in NLP, though it is incremental as it builds on existing tokenization improvements.
The paper tackles the problem of evaluating subword tokenizers, which often ignore morpheme boundaries and affect model performance, by proposing a combined intrinsic-extrinsic evaluation framework. The results show that alien tokenization leads to poorer generalization in language models, with the UniMorph Labeller achieving 98% accuracy.
The popular subword tokenizers of current language models, such as Byte-Pair Encoding (BPE), are known not to respect morpheme boundaries, which affects the downstream performance of the models. While many improved tokenization algorithms have been proposed, their evaluation and cross-comparison is still an open problem. As a solution, we propose a combined intrinsic-extrinsic evaluation framework for subword tokenization. Intrinsic evaluation is based on our new UniMorph Labeller tool that classifies subword tokenization as either morphological or alien. Extrinsic evaluation, in turn, is performed via the Out-of-Vocabulary Generalization Challenge 1.0 benchmark, which consists of three newly specified downstream text classification tasks. Our empirical findings show that the accuracy of UniMorph Labeller is 98%, and that, in all language models studied (including ALBERT, BERT, RoBERTa, and DeBERTa), alien tokenization leads to poorer generalizations compared to morphological tokenization for semantic compositionality of word meanings.