How Much is Enough? The Diminishing Returns of Tokenization Training Data

Tokenization, a crucial initial step in natural language processing, is often assumed to benefit from larger training datasets. This paper investigates the impact of tokenizer training data sizes ranging from 1GB to 900GB. Our findings reveal diminishing returns as the data size increases, highlighting a practical limit on how much further scaling the training data can improve tokenization quality. We analyze this phenomenon and attribute the saturation effect to the constraints imposed by the pre-tokenization stage of tokenization. These results offer valuable insights for optimizing the tokenization process and highlight potential avenues for future research in tokenization algorithms.

Greed is All You Need: An Evaluation of Tokenizer Inference Methods

While subword tokenizers such as BPE and WordPiece are typically used to build vocabularies for NLP models, the method of decoding text into a sequence of tokens from these vocabularies is often left unspecified, or ill-suited to the method in which they were constructed. We provide a controlled analysis of seven tokenizer inference methods across four different algorithms and three vocabulary sizes, performed on a novel intrinsic evaluation suite we curated for English, combining measures rooted in morphology, cognition, and information theory. We show that for the most commonly used tokenizers, greedy inference performs surprisingly well; and that SaGe, a recently-introduced contextually-informed tokenizer, outperforms all others on morphological alignment.

Tokenization Is More Than Compression

Tokenization is a foundational step in Natural Language Processing (NLP) tasks, bridging raw text and language models. Existing tokenization approaches like Byte-Pair Encoding (BPE) originate from the field of data compression, and it has been suggested that the effectiveness of BPE stems from its ability to condense text into a relatively small number of tokens. We test the hypothesis that fewer tokens lead to better downstream performance by introducing PathPiece, a new tokenizer that segments a document's text into the minimum number of tokens for a given vocabulary. Through extensive experimentation we find this hypothesis not to be the case, casting doubt on the understanding of the reasons for effective tokenization. To examine which other factors play a role, we evaluate design decisions across all three phases of tokenization: pre-tokenization, vocabulary construction, and segmentation, offering new insights into the design of effective tokenizers. Specifically, we illustrate the importance of pre-tokenization and the benefits of using BPE to initialize vocabulary construction. We train 64 language models with varying tokenization, ranging in size from 350M to 2.4B parameters, all of which are made publicly available.

Improving a tf-idf weighted document vector embedding

We examine a number of methods to compute a dense vector embedding for a document in a corpus, given a set of word vectors such as those from word2vec or GloVe. We describe two methods that can improve upon a simple weighted sum, that are optimal in the sense that they maximizes a particular weighted cosine similarity measure. We consider several weighting functions, including inverse document frequency (idf), smooth inverse frequency (SIF), and the sub-sampling function used in word2vec. We find that idf works best for our applications. We also use common component removal proposed by Arora et al. as a post-process and find it is helpful in most cases. We compare these embeddings variations to the doc2vec embedding on a new evaluation task using TripAdvisor reviews, and also on the CQADupStack benchmark from the literature.