fastabx: A library for efficient computation of ABX discriminability

TL;DR

fastabx is a Python library that enables efficient construction and computation of ABX discrimination tasks, facilitating broader use in evaluating learned representations across domains.

Contribution

It provides a flexible, high-performance framework for ABX tasks, addressing previous tool limitations and supporting systematic analysis of learned representations.

Findings

Enables rapid ABX task creation and distance computation

Facilitates evaluation of phonetic discriminability in speech models

Supports analysis of representations across multiple domains

Abstract

We introduce fastabx, a high-performance Python library for building ABX discrimination tasks. ABX is a measure of the separation between generic categories of interest. It has been used extensively to evaluate phonetic discriminability in self-supervised speech representations. However, its broader adoption has been limited by the absence of adequate tools. fastabx addresses this gap by providing a framework capable of constructing any type of ABX task while delivering the efficiency necessary for rapid development cycles, both in task creation and in calculating distances between representations. We believe that fastabx will serve as a valuable resource for the broader representation learning community, enabling researchers to systematically investigate what information can be directly extracted from learned representations across several domains beyond speech processing. The source code is available at https://github.com/bootphon/fastabx.