# Towards stable AI systems for Evaluating Arabic Pronunciations

arXiv: https://arxiv.org/abs/2508.19587 · HTML: https://mlsanity.com/paper/2508.19587/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 7 of its 14 references that are in the map.
- First posted on arXiv 27 Aug 2025.

## Builds on (most cited first, 7 of 7)

- [Explaining and Harnessing Adversarial Examples](https://mlsanity.com/paper/1412.6572/) (2014)
- [Towards Deep Learning Models Resistant to Adversarial Attacks](https://mlsanity.com/paper/1706.06083/) (2017)
- [wav2vec 2.0: A Framework for Self-Supervised Learning of Speech Representations](https://mlsanity.com/paper/2006.11477/) (2020)
- [Unsupervised Cross-lingual Representation Learning for Speech Recognition](https://mlsanity.com/paper/2006.13979/) (2020)
- [Audio Adversarial Examples: Targeted Attacks on Speech-to-Text](https://mlsanity.com/paper/1801.01944/) (2018)
- [Deep Speaker Embedding Learning with Multi-Level Pooling for Text-Independent Speaker Verification](https://mlsanity.com/paper/1902.07821/) (2019)
- [Adversarial Attacks on Speech Recognition Systems for Mission-Critical Applications: A Survey](https://mlsanity.com/paper/2202.10594/) (2022)

## Abstract (arXiv)

Modern Arabic ASR systems such as wav2vec 2.0 excel at word- and sentence-level transcription, yet struggle to classify isolated letters. In this study, we show that this phoneme-level task, crucial for language learning, speech therapy, and phonetic research, is challenging because isolated letters lack co-articulatory cues, provide no lexical context, and last only a few hundred milliseconds. Recogniser systems must therefore rely solely on variable acoustic cues, a difficulty heightened by Arabic's emphatic (pharyngealized) consonants and other sounds with no close analogues in many languages. This study introduces a diverse, diacritised corpus of isolated Arabic letters and demonstrates that state-of-the-art wav2vec 2.0 models achieve only 35% accuracy on it. Training a lightweight neural network on wav2vec embeddings raises performance to 65%. However, adding a small amplitude perturbation (epsilon = 0.05) cuts accuracy to 32%. To restore robustness, we apply adversarial training, limiting the noisy-speech drop to 9% while preserving clean-speech accuracy. We detail the corpus, training pipeline, and evaluation protocol, and release, on demand, data and code for reproducibility. Finally, we outline future work extending these methods to word- and sentence-level frameworks, where precise letter pronunciation remains critical.