Segment Any Text: A Universal Approach for Robust, Efficient and Adaptable Sentence Segmentation

TL;DR

The paper introduces SaT, a universal, efficient, and adaptable sentence segmentation model that outperforms existing methods across diverse domains and languages, especially with poorly formatted text, by reducing punctuation reliance and enhancing domain adaptability.

Contribution

We propose SaT, a novel sentence segmentation model with a new pretraining scheme and architectural improvements, achieving robustness, adaptability, and efficiency in diverse text domains.

Findings

Outperforms all baselines across 8 diverse corpora

Achieves threefold speed improvement over previous state-of-the-art

Effectively handles poorly formatted and multilingual text

Abstract

Segmenting text into sentences plays an early and crucial role in many NLP systems. This is commonly achieved by using rule-based or statistical methods relying on lexical features such as punctuation. Although some recent works no longer exclusively rely on punctuation, we find that no prior method achieves all of (i) robustness to missing punctuation, (ii) effective adaptability to new domains, and (iii) high efficiency. We introduce a new model - Segment any Text (SaT) - to solve this problem. To enhance robustness, we propose a new pretraining scheme that ensures less reliance on punctuation. To address adaptability, we introduce an extra stage of parameter-efficient fine-tuning, establishing state-of-the-art performance in distinct domains such as verses from lyrics and legal documents. Along the way, we introduce architectural modifications that result in a threefold gain in speed over the previous state of the art and solve spurious reliance on context far in the future. Finally, we introduce a variant of our model with fine-tuning on a diverse, multilingual mixture of sentence-segmented data, acting as a drop-in replacement and enhancement for existing segmentation tools. Overall, our contributions provide a universal approach for segmenting any text. Our method outperforms all baselines - including strong LLMs - across 8 corpora spanning diverse domains and languages, especially in practically relevant situations where text is poorly formatted. Our models and code, including documentation, are available at https://github.com/segment-any-text/wtpsplit under the MIT license.