Reproduce “GigaSpeech 3: Scaling Multi-Dialect Arabic Speech Recognition with Weakly Labeled Data” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Junzhe Liu, Yujie Tu, Yifan Yang, Peng Wang, Haoran Wang, Guodong Lin, Leyan Liao, Chengquan Zhang and 8 more
The authors build weakly labeled corpora for five Arabic dialects using small manual seeds to train recognizers that transcribe web audio. Independently initialized Zipformer models trained on these labels outperform their label-generating models across all five dialects. On GigaSpeechBench, their macro-average word error rate is 23.86%, versus 28.62% for a third-party system, but they lead on only three dialects.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.