#low-resource languages

29 stories · 1 in 7 days · 9 in 30 days
Stories, products and papers tagged "low-resource languages", newest first.
Nov 13
ModelsBuildersMulti-source13:19
Chinese contentSaidzoda-Kazakh-EmbeddingGemma-300M:单轮微调将哈萨克语检索 recall@1 提升至 78.0%

339M 参数的小模型,用一个 epoch 的合成问答对微调,哈萨克语检索 recall@1 从 60.8% 拉到 78.0%,还压过 OpenAI text-embedding-3-large。做低资源语言检索或想复刻这套合成数据微调流程的工程师都该看看细节。

Event
arXiv: OpenAI@Tohir Saidzoda et al. (6)Source
Nov 3
Oct 30
Oct 28
Oct 22
Oct 21
Oct 20
Oct 15
Oct 13
PapersBuildersOfficialSelected22:47
Apple Research Finds Data Mixing Beats Hyperparameter Tuning for Low-Resource Languages

The practical implication is that teams building models for smaller language markets should prioritise corpus curation and bilingual data mixes over expensive hyperparameter sweeps, saving compute and experiment cycles. Engineers working on Thai, Vietnamese or Bahasa Indonesia models can apply the mixing-ratio approach directly, and it offers a concrete benchmark for comparing against tuning-heavy pipelines used by larger labs.

Primary sourceApple ML ResearchSource
Oct 6
Sep 28
ModelsBuildersSelected73°23:58
Monsoon speech dataset cuts Telugu WER from 92.7% to 16.1%

The takeaway is that low-resource ASR quality can be bought with better data curation rather than bigger models, since a mid-size Whisper checkpoint beat Gemini 3.1 Pro on Telugu purely through fine-tuning. Teams building speech products for Indic and other under-served language markets can use this dataset-first approach instead of waiting for larger architectures. Keeping noisy audio in training, with DNSMOS-based stratification, is the transferable technique.

rohanpaul_ai@rohanpaul_aiSource
Sep 25
Aug 31
Aug 19
Aug 13
Aug 7
Jul 23
Jul 21
Jul 3
Jun 12
Jun 3
Papers10:25
Chinese content用从未发生的对话训练ASR:合成对话提升低资源语言识别

低资源语言ASR团队终于有了可落地的数据增强方案——用LLM+TTS生成对话数据,效果堪比数倍真实数据。做多说话人语音识别的开发者值得一试,尤其适合匈牙利语等小语种场景。

arXiv cs.AI@Máté Gedeon, Péter MihajlikSource
May 22
May 19
May 14