arXiv cs.CL· Amr Abdelkhalik·· 13 小时前AI 评分29
Gemma 4 语音编码器不改权重能走多远？冻结与适配器增强编码器的沙特及海湾阿拉伯语语音识别
How Far Can the Gemma 4 Speech Encoder Go Without Changing Its Weights? Saudi and Gulf Arabic Speech Recognition with a Frozen and an Adapter-Augmented Encoder
AI 导读
研究在不改动 Gemma 4 与 EmbeddingGemma 2 共享的 305M 参数音频编码器权重的前提下，用于沙特和海湾阿拉伯语语音识别。冻结系统仅训练 22.6M 参数 CTC 读取器，在六项人工标注条件中的五项上字符错误率与 2.07B 参数的 Cohere Transcribe Arabic 统计无差异，CPU 贪心解码快 3.4 倍。
正文 · 原文
Abstract:We study the 305M-parameter audio encoder shared by Gemma 4 and EmbeddingGemma 2 (identical except for its final projection) for Saudi and Gulf Arabic speech recognition, without changing any of its weights. A frozen system trains only a 22.6M-parameter CTC reader over a mix of encoder layers; an adapted system adds 3.1M low-rank adapter parameters, a 41.5M-parameter hybrid CTC/attention reader and joint decoding with a sub-word language model. Against Cohere Transcribe Arabic (2.07B parameters), re-run on identical audio, the frozen system trails in word error rate (WER) but is statistically indistinguishable in character error rate (CER) on five of six human-labelled conditions, while running 3.4 times faster on CPU with greedy decoding. The adapted system is statistically indistinguishable in WER on a public Saudi customer-service call (19.20% against 19.27%) and on clean Emirati speech, is significantly better on telephone-coded Emirati speech, has a significantly lower CER on all six conditions of these sets, and beats a 0.6B recogniser fine-tuned on SADA in WER on all six; it still trails Cohere in WER on long-form Saudi speech. On a blind, human-corrected public call-centre sample it ranks 4 of 15 systems, behind three commercial APIs (significantly only one) and ahead of the other self-hostable systems by point estimate. An optimised joint decoder runs it at 139 times real time on one GPU. We also show that references made by correcting machine drafts favour the drafting system. Code and public-set results are released.
| Comments: | 24 pages, 15 tables, 5 figures. Code and public-set results: this https URL |
| Subjects: | Computation and Language (cs.CL); Sound (cs.SD); Audio and Speech Processing (eess.AS) |
| Cite as: | arXiv:2610.28673 [cs.CL] |
| (or arXiv:2610.28673v1 [cs.CL] for this version) | |
| https://doi.org/10.48550/arXiv.2610.28673 arXiv-issued DOI via DataCite |
Submission history
From: Amr Abdelkhalik [view email]
[v1]
Sun, 18 Oct 2026 21:06:44 UTC (143 KB)
来源：arXiv cs.CL · arxiv.org