The Interspeech 2025 Speech Accessibility Project Challenge

TL;DR

The 2025 Interspeech SAP Challenge advances speech recognition for individuals with speech disabilities by providing a large, diverse dataset and evaluating new models that outperform existing baselines in accuracy and semantic understanding.

Contribution

This paper introduces a large, publicly available dataset and a competitive challenge to improve ASR systems for speech-impaired individuals, setting new performance benchmarks.

Findings

12 teams outperformed the baseline in Word Error Rate

17 teams surpassed the baseline in Semantic Score

Top system achieved 8.11% WER and 88.44% SemScore

Abstract

While the last decade has witnessed significant advancements in Automatic Speech Recognition (ASR) systems, performance of these systems for individuals with speech disabilities remains inadequate, partly due to limited public training data. To bridge this gap, the 2025 Interspeech Speech Accessibility Project (SAP) Challenge was launched, utilizing over 400 hours of SAP data collected and transcribed from more than 500 individuals with diverse speech disabilities. Hosted on EvalAI and leveraging the remote evaluation pipeline, the SAP Challenge evaluates submissions based on Word Error Rate and Semantic Score. Consequently, 12 out of 22 valid teams outperformed the whisper-large-v2 baseline in terms of WER, while 17 teams surpassed the baseline on SemScore. Notably, the top team achieved the lowest WER of 8.11\%, and the highest SemScore of 88.44\% at the same time, setting new benchmarks for future ASR systems in recognizing impaired speech.