Kusaal Whisper Small LoRA

--- license: cc-by-nc-4.0 language: - kus tags: - ghana-nlp - ghana-speech - automatic-speech-recognition base_model: openai/whisper-small pipeline_tag: automatic-speech-recognition --- # Kusaal ASR (Whisper-small LoRA, superseded) **Author:** Prince Nasamu Alhassan ## Overview An earlier Kusaal experiment from 2026-07-19, kept for history. **Superseded by `tekyerema-asr-mms-kus`**, which scores 30.44 WER against the 41.0 of the previous champion. Its README cites a dataset at `alhassanprince/kusaal-asr-dataset`, which does not exist — the real one is under `PrinceAlhassanNasamu/`. It predates the H200 window, so it correctly carries no AISCA acknowledgement. ## Use it _No loading snippet for this model yet._ ## Training data Trained on the [Ghana Speech](https://huggingface.co/datasets/ghananlpcommunity/ghana-speech) dataset and related Ghanaian corpora, licensed CC BY-NC 4.0. ## Intended use & license **Non-commercial use only** (CC BY-NC 4.0). This is inherited from the training data *and* required by the terms under which the compute was granted: models trained in that window are non-commercial by condition of access, not by inference. ## Limitations, stated plainly - **Dagbani did get a recogniser, and the claim that it could not was wrong twice over.** Every card on this account used to say that "one fine-tuning session on 74 validation rows would not change that". Those 74 rows are the **eng-dag machine-translation** validation split; the Dagbani *speech* data in this same account is `waxal_dag` — 13,228 training rows, 1,750 validation rows, ~71 hours, 1,041 speakers with the largest at 1%. Trained on it, `tekyerema-asr-mms-dag` scores **36.94 / 11.71**, against the 86.59 / 33.95 this project had believed was the ceiling. It still loses to `FarmerlineML/w2v-bert-2.0_2026_dagbani_ASR` at **29.20 / 9.27**, which is what the agent actually serves. A number carried across from a translation table into a speech claim was then repeated on every card here until 2026-09-22. - **Evaluation is on read and machine-translated text.** No recordings of people speaking agent commands in these languages exist. Numbers measured this way are optimistic about phrasing and pessimistic about code-switching, and should not be read as field performance. - Research work from a hackathon entry, not a supported product. ## The rest of the fami...

--- language: - kus license: other tags: - automatic-speech-recognition - asr - whisper - lora - peft - kusaal - ghana - low-resource - speech-recognition base_model: openai/whisper-small datasets: - alhassanprince/kusaal-asr-dataset metrics: - wer pipeline_tag: automatic-speech-recognition --- # Kusaal Whisper-small LoRA The **first open ASR model for Kusaal** (`kus`, ISO 639-3) — a Gur language spoken by approximately 350,000 people in the Upper East Region of Ghana and adjacent Burkina Faso. LoRA fine-tune of [openai/whisper-small](https://huggingface.co/openai/whisper-small) for Kusaal speech recognition. --- ## Results | Split | Books held out | WER | |-------|---------------|-----| | Val | LUK, PHP, PRO, PSA | **~30.4%** | Evaluated on 400 randomly sampled clips from the val split of [alhassanprince/kusaal-asr-dataset](https://huggingface.co/datasets/alhassanprince/kusaal-asr-dataset). Test-set WER (JON, MAT, REV, RUT) to be reported in the accompanying paper. > **Context:** Kusaal had no prior ASR system or public speech corpus before this work. A 30.4% WER on a first model trained on ~82 hours of scripture audio, with book-held-out evaluation, represents a strong baseline for a zero-resource language. --- ## Sample outputs ``` PSA.89.22 REF: dataa kae na nyaŋ oo bugusim sɔb kae na mugus oo HYP: m taa kae na nyaŋ o bugusim sɔb kae na mugus o PRO.26.15 REF: gbanyaa na nɔk o nuugu tɔm laan amaa ka ukim niŋ nɔɔrin la an yɛl HYP: yaa ya na nɔk o nuugu tɔm laan amaa ka o ku m niŋ nɔɔrin la an yɛl PSA.98.5 REF: nwɛɛm kɔnne tis zugsɔb la yɔɔm ka nwɛ kɔn HYP: nwɛɛm kɔn nɛ tis zugsɔb la yɔɔm ka nwɛɛ kɔɔn ``` The model correctly handles Kusaal-specific characters: **ŋ ɔ ɛ ʋ**. --- ## Training details | Parameter | Value | |-----------|-------| | Base model | openai/whisper-small (241M params) | | Method | LoRA (PEFT) | | LoRA rank | 32 | | LoRA alpha | 64 | | LoRA targets | `q_proj`, `v_proj` | | Trainable params | 3,538,944 (1.44% of total) | | Epochs | 2 | | Optimizer steps | 3,074 | | Learning rate | 1e-4 | | Effective batch size | 16 (batch 1 × grad accum 16) | | Warmup steps | 500 | | Final train loss | 0.730 | | Training time | ~4h 14m on Kaggle T4 | | Precision | fp16 | --- ## Training data Trained on [alhassanprince/kusaal-asr-dataset](https://huggingface.co/datasets/alhassanprince/kusaal-asr-dataset): - **30,820** verse-level WA...

Source context: 69 downloads · 0 likes · Pipeline automatic-speech-recognition · Library peft · Repo PrinceAlhassanNasamu/kusaal-whisper-small-lora