English

On the Robustness of Text Vectorizers

Computation and Language 2023-06-13 v2 Machine Learning Machine Learning

Abstract

A fundamental issue in machine learning is the robustness of the model with respect to changes in the input. In natural language processing, models typically contain a first embedding layer, transforming a sequence of tokens into vector representations. While the robustness with respect to changes of continuous inputs is well-understood, the situation is less clear when considering discrete changes, for instance replacing a word by another in an input sentence. Our work formally proves that popular embedding schemes, such as concatenation, TF-IDF, and Paragraph Vector (a.k.a. doc2vec), exhibit robustness in the H\"older or Lipschitz sense with respect to the Hamming distance. We provide quantitative bounds for these schemes and demonstrate how the constants involved are affected by the length of the document. These findings are exemplified through a series of numerical examples.

Keywords

Cite

@article{arxiv.2303.07203,
  title  = {On the Robustness of Text Vectorizers},
  author = {Rémi Catellier and Samuel Vaiter and Damien Garreau},
  journal= {arXiv preprint arXiv:2303.07203},
  year   = {2023}
}

Comments

Accepted to ICML 2023. 33 pages, 10 figures

R2 v1 2026-06-28T09:14:22.902Z