Prompt mode: metadata
Method note: This is a metadata-conditioned/oracle prompt experiment. The prompt uses speaker severity, disorder type, and etiology from speaker_metadata.csv; compare it separately from deployment baselines.
WER framing: this notebook does not ask the model to remove repetitions, clean disfluencies, summarize, or normalize grammar.

Test Metrics

Normalized WER: 0.506641
Normalized CER: 0.211382
Avg capped WER: 0.425857
Avg capped CER: 0.170021

Training Setup

Dataset: cdli/ugandan_luganda_nonstandard_speech_v1.0
Splits: cleaned TSVs matched by audio_id
Base model: KasuleTrevor/cdli-qwen3-asr-lg-typical-1p7b-base-finetune
LR: 5e-05
Scheduler: cosine
Epochs: 5
Train mode: full
Seed: 42

Metadata Prompt Template

Speaker speech disorder type, speech impairment severity, and etiology are inserted from metadata. The prompt then instructs exact Luganda transcription while preserving repetitions, disfluencies, false starts, and partial words.

KasuleTrevor/cdli-qwen3-asr-lg-uganda-severity-disorder-fixed-cosine1

CDLI Qwen3-ASR Luganda Atypical Speech Prompt Fine-tune

Prompt Setup

Test Metrics

Training Setup

Metadata Prompt Template

KasuleTrevor/
cdli-qwen3-asr-lg-uganda-severity-disorder-fixed-cosine1