GigaSpeech 3 advances Arabic dialect speech recognition with massive data
GigaSpeech 3: Scaling Multi-Dialect Arabic Speech Recognition with Weakly Labeled Data
Computation and Language
Summary
Understanding spoken Arabic is tricky because people speak many dialects, and there’s not much labeled data for each one. The authors put together a very large collection of Arabic speech from five dialects, using some manually labeled examples to help guess transcriptions for much more unlabeled speech. This approach helped improve speech recognition models that work better across dialects and even with modern standard Arabic. Their large dataset and models reduced errors noticeably compared to previous systems.
What this means in practice
- •For speech recognition developers: Train more accurate Arabic dialect-specific and multi-dialect speech recognizers using a large, weakly labeled dataset enabling broader dialect coverage.
- •For language technology product teams: Improve Arabic voice assistants and transcription services by integrating models trained on diverse dialects including Saudi and Syrian Arabic.$Commercial implications: The dataset enables building commercial voice products that work well across Arabic dialects, improving user experience and market reach.
Authors
Junzhe Liu, Yujie Tu, Yifan Yang, Peng Wang, Haoran Wang, Guodong Lin, Leyan Liao, Chengquan Zhang, Zhifu Gao, Xiangang Li, Qinyuan Cheng, Xipeng Qiu, Wei-Qiang Zhang, Guoguo Chen, Kai Yu, Xie Chen
Abstract
Dialectal Arabic speech recognition remains constrained by the scarcity of dialect-resolved transcribed speech despite the abundance of web audio. We introduce GigaSpeech 3, a corpus of 39,000 hours and 10 million weakly transcribed speech segments covering Algerian, Egyptian, Iraqi, Saudi, and Syrian Arabic. Starting from 60 hours of manual transcription per dialect, we train labeling models using speech-recognition-biased self-supervised learning and apply them to dialect-matched web audio. Data selection and four rounds of noisy-student refinement for Saudi and Syrian Arabic yield the final corpus. We validate its utility through dialect-specific and multi-dialect speech recognition training and cross-domain evaluation. On GigaSpeechBench, the final dialect-specific models outperform their labeling models across all five dialects and achieve a macro-average word error rate of 23.86%, a 16.6% relative reduction from the strongest third-party system covering all five dialects, while outperforming the best third-party result on three dialects. A multi-dialect model using self-predicted dialect identities achieves a macro-average word error rate of 25.51% while also supporting Modern Standard Arabic. Across eight conditions from three external corpora, a GigaSpeech 3-trained Zipformer achieves the lowest word error rate among six evaluated models in every condition, while adapting the Qwen3 speech recognizer on GigaSpeech 3 improves performance across all eight conditions. We release the corpus and recipes as a foundation for dialect-aware Arabic speech processing.