# Black-box Adaptation of ASR for Accented Speech

arXiv: https://arxiv.org/abs/2006.13519 · HTML: https://mlsanity.com/paper/2006.13519/

## Key facts

- Cited by 8 papers in the map (data through 14 Oct 2026).
- First posted on arXiv 24 Jun 2020.

## Built on by (most cited first, 8 of 8)

- [Accented Speech Recognition: A Survey](https://mlsanity.com/paper/2104.10747/) (2021)
- [RED-ACE: Robust Error Detection for ASR using Confidence Embeddings](https://mlsanity.com/paper/2203.07172/) (2022)
- [Language technology practitioners as language managers: arbitrating data bias and predictive bias in ASR](https://mlsanity.com/paper/2202.12603/) (2022)
- [Adapting an Unadaptable ASR System](https://mlsanity.com/paper/2306.01208/) (2023)
- [AccentFold: A Journey through African Accents for Zero-Shot ASR Adaptation to Target Accents](https://mlsanity.com/paper/2402.01152/) (2024)
- [Can Open Domain Question Answering Systems Answer Visual Knowledge Questions?](https://mlsanity.com/paper/2202.04306/) (2022)
- [Accented Speech Recognition Inspired by Human Perception](https://mlsanity.com/paper/2104.04627/) (2021)
- [ASR Error Detection via Audio-Transcript entailment](https://mlsanity.com/paper/2207.10849/) (2022)

## Abstract (arXiv)

We introduce the problem of adapting a black-box, cloud-based ASR system to speech from a target accent. While leading online ASR services obtain impressive performance on main-stream accents, they perform poorly on sub-populations - we observed that the word error rate (WER) achieved by Google's ASR API on Indian accents is almost twice the WER on US accents. Existing adaptation methods either require access to model parameters or overlay an error-correcting module on output transcripts. We highlight the need for correlating outputs with the original speech to fix accent errors. Accordingly, we propose a novel coupling of an open-source accent-tuned local model with the black-box service where the output from the service guides frame-level inference in the local model. Our fine-grained merging algorithm is better at fixing accent errors than existing word-level combination strategies. Experiments on Indian and Australian accents with three leading ASR models as service, show that we achieve as much as 28% relative reduction in WER over both the local and service models.