Between words and characters: A Brief History of Open-Vocabulary Modeling and Tokenization in NLP

Abstract

What are the units of text that we want to model? From bytes to multi-word expressions, text can be analyzed and generated at many granularities. Until recently, most natural language processing (NLP) models operated over words, treating those as discrete and atomic tokens, but starting with byte-pair encoding (BPE), subword-based approaches have become dominant in many areas, enabling small vocabularies while still allowing for fast inference. Is the end of the road character-level model or byte-level processing? In this survey, we connect several lines of work from the pre-neural and neural era, by showing how hybrid approaches of words and characters as well as subword-based approaches based on learned segmentation have been proposed and evaluated. We conclude that there is and likely will never be a silver bullet singular solution for all applications and that thinking seriously about tokenization remains important for many applications.

15 page preprint

References in corpus (14)

• Cross-lingual Language Model Pretraining
• CharBERT: Character-aware Pre-trained Language Model
• Charformer: Fast Character Transformers via Gradient-based Subword Tokenization
• ByT5: Towards a token-free future with pre-trained byte-to-byte models
• Hash Embeddings for Efficient Word Representations
• A Call for Prudent Choice of Subword Merge Operations in Neural Machine Translation
• Morphological Word Segmentation on Agglutinative Languages for Neural Machine Translation
• Squared English Word: A Method of Generating Glyph to Use Super Characters for Sentiment Analysis
• Neural Polysynthetic Language Modelling
• Unsupervised Word Segmentation with Bi-directional Neural Language Model
• A Latent Morphology Model for Open-Vocabulary Neural Machine Translation
• Word Shape Matters: Robust Machine Translation with Visual Embedding
• A Masked Segmental Language Model for Unsupervised Natural Language Segmentation
• Crowdsourced Phrase-Based Tokenization for Low-Resourced Neural Machine Translation: The Case of Fon Language

Cited by in corpus (7)

• Linguistically inspired roadmap for building biologically reliable protein language models
• UniMorph 4.0: Universal Morphology
• Natural Language Processing Methods for Symbolic Music Generation and Information Retrieval: a Survey
• Large language models for automated scholarly paper review: A survey
• Language Modelling with Pixels
• ImmunoLingo: Linguistics-based formalization of the antibody language
• SelfSeg: A Self-supervised Sub-word Segmentation Method for Neural Machine Translation