RyeAI/danish-asr-leaderboard

Open Danish ASR Leaderboard — Results

Benchmark results backing the Open Danish ASR Leaderboard — a reproducible, open evaluation of Danish automatic speech recognition models across five independent public test sets.

The results config (shown by default) has one row per evaluated model. Scores are WER / CER (%) — lower is better. Each model also has its own config exposing the raw, un-normalised transcriptions (reference vs hypothesis per utterance) for GPU-free re-scoring and error analysis.

Test sets

Column prefix	Dataset	Split	Domain
coral_conversation	CoRal-project/coral-v3 — conversation	test	Spontaneous conversation
coral_read_aloud	CoRal-project/coral-v3 — read_aloud	test	Read-aloud speech
ftspeech	alexandrainst/ftspeech	test_balanced	Parliamentary / broadcast
cv17_da	mozilla-foundation/common_voice_17_0 — da	test	Crowd-sourced read speech
fleurs_da	google/fleurs — da_dk	test	Read speech

Schema

results config — one row per model:

Column	Type	Description
model	string	Markdown link: [org/name](https://huggingface.co/org/name) for HF models, plain name for hosted APIs
params_b	float	Parameter count in billions from safetensors metadata; NaN for API models
access	string	open = open weights; proprietary = hosted or closed model
mean_wer	float	Macro-averaged WER (%) across the five core test sets
mean_cer	float	Macro-averaged CER (%) across the five core test sets
coral_conversation_wer	float|null	WER on CoRal v3 conversation
coral_read_aloud_wer	float|null	WER on CoRal v3 read-aloud
ftspeech_wer	float|null	WER on FTSpeech
cv17_da_wer	float|null	WER on Common Voice 17 (Danish)
fleurs_da_wer	float|null	WER on FLEURS (Danish)
coral_conversation_cer	float|null	CER on CoRal v3 conversation
coral_read_aloud_cer	float|null	CER on CoRal v3 read-aloud
ftspeech_cer	float|null	CER on FTSpeech
cv17_da_cer	float|null	CER on Common Voice 17 (Danish)
fleurs_da_cer	float|null	CER on FLEURS (Danish)
speed_x	float|null	Audio seconds / wall-clock second (higher = faster). Measured on one RTX Pro 5000 Blackwell; network-bound for API models. NaN if not measured.
submitted	string	ISO 8601 date the result was submitted (YYYY-MM-DD)

Per-model configs (outputs/<model-slug>) — one row per utterance: dataset, id, reference, hypothesis (raw, un-normalised).

Text normalisation

Applied identically to hypothesis and reference before scoring, so WER/CER reflect recognition errors rather than formatting:

1. Unicode NFKC — compatibility composition (folds ligatures, full-width digits, ²→2, …). A near-no-op on Danish speech text, adopted for correctness and consistency with the Danish standard.
2. Danish number canonicalisation — separators within a numeral are stripped (1.234 → 1234, 3,14 → 314).
3. Lowercase.
4. Punctuation / symbol removal — apostrophes inside a word (det's) are preserved; all other punctuation and symbols are removed.
5. Whitespace collapse.
6. Numerals → words — every standalone integer token is expanded to its Danish cardinal words via num2words (4 → fire, 24 → fireogtyve), so digit-vs-word formatting ("4" vs "fire") is not counted as an error. Only standalone integers are converted; digits embedded in larger tokens (decades like 1960'erne, ranges like 1-3) are left untouched. Ordinals (3. → tredje) and symbol/unit expansion (% → procent) were tested and rejected as net-neutral-to-harmful.

An optional filler-word strip (øh, hmm, …) is available in the harness but off by default, since its effect concentrates on spontaneous-speech sets and can shift that column's relative order.

Danish orthographic variants (aa↔å, oe↔ø, ae↔æ) are not normalised — the digraphs occur legitimately as letter sequences. Because the normaliser is parameterised, scripts/rescore.py can re-derive WER/CER from the saved raw outputs under any configuration without re-running inference.

Adding a model

There are two paths, depending on whether you have run the evaluation yourself:

• Request a model (we run it): open a GitHub issue with the model id, backend, and where to find it — we'll run it through the harness and add it.
• Submit a score (you ran it): run the harness from the GitHub repo and open a pull request with results/<model-slug>.json plus the raw outputs/<model-slug>/ transcriptions. On merge, CI publishes both here and updates the leaderboard automatically.

Whichever path a model arrives by, we re-evaluate it independently on our own hardware before publishing — to confirm the scores reproduce and catch configuration differences. Do not modify the normalisation or metrics; run the harness as-is so results stay comparable.

License

MIT — see LICENSE.