Afrispeech-Dialog: A Benchmark Dataset for Spontaneous English Conversations in Healthcare and Beyond
This work addresses the problem of underperformance in speech technologies for African-accented English, particularly in healthcare and other sectors, by providing a dataset and benchmarks, though it is incremental as it builds on existing methods for new data.
The authors tackled the performance gap of speech technologies on African-accented English by introducing Afrispeech-Dialog, a benchmark dataset of 50 simulated conversations, and found a 10%+ degradation in ASR and speaker diarization compared to native accents, while also exploring the impact of ASR errors on medical summarization with LLMs.
Speech technologies are transforming interactions across various sectors, from healthcare to call centers and robots, yet their performance on African-accented conversations remains underexplored. We introduce Afrispeech-Dialog, a benchmark dataset of 50 simulated medical and non-medical African-accented English conversations, designed to evaluate automatic speech recognition (ASR) and related technologies. We assess state-of-the-art (SOTA) speaker diarization and ASR systems on long-form, accented speech, comparing their performance with native accents and discover a 10%+ performance degradation. Additionally, we explore medical conversation summarization capabilities of large language models (LLMs) to demonstrate the impact of ASR errors on downstream medical summaries, providing insights into the challenges and opportunities for speech technologies in the Global South. Our work highlights the need for more inclusive datasets to advance conversational AI in low-resource settings.