RETVec: Resilient and Efficient Text Vectorizer

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Bursztein, Elie, Zhang, Marina, Vallis, Owen, Jia, Xinyu, Kurakin, Alexey
Format: Preprint
Published: 2023
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866916217839681536
author Bursztein, Elie
Zhang, Marina
Vallis, Owen
Jia, Xinyu
Kurakin, Alexey
author_facet Bursztein, Elie
Zhang, Marina
Vallis, Owen
Jia, Xinyu
Kurakin, Alexey
contents This paper describes RETVec, an efficient, resilient, and multilingual text vectorizer designed for neural-based text processing. RETVec combines a novel character encoding with an optional small embedding model to embed words into a 256-dimensional vector space. The RETVec embedding model is pre-trained using pair-wise metric learning to be robust against typos and character-level adversarial attacks. In this paper, we evaluate and compare RETVec to state-of-the-art vectorizers and word embeddings on popular model architectures and datasets. These comparisons demonstrate that RETVec leads to competitive, multilingual models that are significantly more resilient to typos and adversarial text attacks. RETVec is available under the Apache 2 license at https://github.com/google-research/retvec.
format Preprint
id arxiv_https___arxiv_org_abs_2302_09207
institution arXiv
publishDate 2023
record_format arxiv
spellingShingle RETVec: Resilient and Efficient Text Vectorizer
Bursztein, Elie
Zhang, Marina
Vallis, Owen
Jia, Xinyu
Kurakin, Alexey
Computation and Language
Artificial Intelligence
This paper describes RETVec, an efficient, resilient, and multilingual text vectorizer designed for neural-based text processing. RETVec combines a novel character encoding with an optional small embedding model to embed words into a 256-dimensional vector space. The RETVec embedding model is pre-trained using pair-wise metric learning to be robust against typos and character-level adversarial attacks. In this paper, we evaluate and compare RETVec to state-of-the-art vectorizers and word embeddings on popular model architectures and datasets. These comparisons demonstrate that RETVec leads to competitive, multilingual models that are significantly more resilient to typos and adversarial text attacks. RETVec is available under the Apache 2 license at https://github.com/google-research/retvec.
title RETVec: Resilient and Efficient Text Vectorizer
topic Computation and Language
Artificial Intelligence
url https://arxiv.org/abs/2302.09207