Instructions to use Awesome-x/voxira-yoruba-phaseXII with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use Awesome-x/voxira-yoruba-phaseXII with Transformers:
# Use a pipeline as a high-level helper from transformers import pipeline pipe = pipeline("automatic-speech-recognition", model="Awesome-x/voxira-yoruba-phaseXII")# pip install -U transformers accelerate # Load model directly from transformers import AutoProcessor, AutoModelForCTC processor = AutoProcessor.from_pretrained("Awesome-x/voxira-yoruba-phaseXII") model = AutoModelForCTC.from_pretrained("Awesome-x/voxira-yoruba-phaseXII", device_map="auto") - Notebooks
- Google Colab
- Kaggle
- Voxira Yorรนbรก Phase XII: YorรนbรกโEnglish code-switching ASR
Voxira Yorรนbรก Phase XII: YorรนbรกโEnglish code-switching ASR
Phase XII = newly trained acoustic weights + a rebuilt 4-gram KenLM + a pyctcdecode bundle. The acoustic model is a 10-epoch fine-tune of facebook/wav2vec2-large-xlsr-53; the language model is a pruned 4-gram trained on a 141,476-line corpus balanced 40 / 30 / 30 (Yorรนbรก / English / code-switched). Unlike Phase XI, which kept the Phase X acoustic weights and changed only the decoder, Phase XII changes both. Decoding requires a blank-fusion step (see below); plain processor.batch_decode will not work.
Status: release candidate. Decoder comparison complete; independent held-out verification pending. The numbers below come from a small held-out set of read studio recordings (one corpus) on which decoder settings were also compared. They are not a generalisation estimate.
At a glance (all 49 valid held-out clips, 749 words): greedy with blank fusion 26.70% WER โ KenLM at default settings 19.76% (26% fewer word errors). The shipped configuration scored 18.52% on the 44 clean takes (greedy there: 26.96%), but its word list and settings were tuned on those same clips, so the default-settings figure is the more honest one.
What changed from Phase X / XI
| Phase X | Phase XI | Phase XII | |
|---|---|---|---|
| Acoustic weights | Phase 10 fine-tune | Identical to Phase X | Newly trained (checkpoint-2950, 10 epochs) |
| Output vocabulary | 42 tokens, no Q / X / Z (per the Phase XI card) | same acoustic model | 46 entries in vocab.json (adds Q=42, X=43, Z=44, '=45) + <pad> at row 46 โ 47 output rows |
| Blank handling | standard processor decoding | standard processor decoding | blank probability split across rows 0 and 46; fusion required |
| LM text | ~366k lines, ~85.6% Yorรนbรก / ~1.2% English | 50k lines, entropy-balanced, templated openings capped, banking/conversational text up-weighted 4ร | 141,476 lines, 40 / 30 / 30, boilerplate removed, templated openings capped at 100 per prefix |
| KenLM | 5-gram | 4-gram c50k_prune, 4.1 MB, --prune 0 1 1 2 |
4-gram phaseXII_40_30_30_prune.bin, 10.83 MB, --prune 0 1 1 2 |
| Unigram list | large unigram list | 47,017 words (lm_vocab) |
50,259 words |
| alpha / beta | 2.0 / 1.5 | 0.3 / 0.5 | 0.25 / 0.5 stored in attrs.json (0.30 / 0.50 also tested) |
Evaluation: read this honestly
Held-out set. The last 50 unique prompts of the AR_metadata code-switching corpus (AR_396.wavโAR_449.wav) were held out and removed from every LM training bucket. Of those 50 audio files, AR_418.wav is empty (0 bytes), leaving 49 valid clips; 5 of those are aborted or cut-off studio takes (AR_409.wav, AR_414.wav, AR_424.wav, AR_430.wav, AR_440.wav), leaving 44 clean takes. The 44 clean takes are a subset of the 49, so the two tables are not independent.
44 clean completed takes (664 reference words)
| Decoder | Errors / words | WER | 95% CI (naive) | vs greedy | Relative error reduction | Selection status |
|---|---|---|---|---|---|---|
| Greedy CTC, blank-fused, no LM | 179 / 664 | 26.96% | 23.7โ30.5% | baseline | โ | baseline |
| KenLM, 66,184-word list, alpha=0.30 beta=0.50 | 128 / 664 | 19.28% | 16.5โ22.5% | -7.68 pp | 28.5% (51 fewer errors) | Phase XI default settings; word list from the training corpus only |
| KenLM, 66,184-word list, alpha=0.25 beta=0.50 | 125 / 664 | 18.83% | 16.0โ22.0% | -8.13 pp | 30.2% (54 fewer errors) | settings chosen by comparing on this set |
| KenLM, 66,184-word list, alpha=0.35 beta=0.60 | 125 / 664 | 18.83% | 16.0โ22.0% | -8.13 pp | 30.2% (54 fewer errors) | settings chosen by comparing on this set |
| KenLM, expanded word list, alpha=0.30 beta=0.50 | 125 / 664 | 18.83% | 16.0โ22.0% | -8.13 pp | 30.2% (54 fewer errors) | word list extended after inspecting errors on this set |
| KenLM, expanded word list, alpha=0.25 beta=0.50 | 123 / 664 | 18.52% | 15.8โ21.7% | -8.43 pp | 31.3% (56 fewer errors) | shipped word list + settings; both tuned on this set |
| KenLM, expanded word list, alpha=0.35 beta=0.60 | 123 / 664 | 18.52% | 15.8โ21.7% | -8.43 pp | 31.3% (56 fewer errors) | word list and settings tuned on this set |
All 49 valid clips, including aborted takes (749 reference words)
| Decoder | Errors / words | WER | 95% CI (naive) | vs greedy | Relative error reduction | Selection status |
|---|---|---|---|---|---|---|
| Greedy CTC, blank-fused, no LM | 200 / 749 | 26.70% | 23.7โ30.0% | baseline | โ | baseline |
| KenLM, 66,184-word list, alpha=0.30 beta=0.50 | 148 / 749 | 19.76% | 17.1โ22.8% | -6.94 pp | 26.0% (52 fewer errors) | Phase XI default settings; word list from the training corpus only |
| KenLM, 66,184-word list, alpha=0.25 beta=0.50 | 145 / 749 | 19.36% | 16.7โ22.3% | -7.34 pp | 27.5% (55 fewer errors) | settings chosen by comparing on this set |
| KenLM, 66,184-word list, alpha=0.35 beta=0.60 | 145 / 749 | 19.36% | 16.7โ22.3% | -7.34 pp | 27.5% (55 fewer errors) | settings chosen by comparing on this set |
Why blank fusion is required (5-clip development probe, 118 words)
| Configuration | Errors / words | WER | Change |
|---|---|---|---|
| Greedy, row 46 masked, NOT fused into row 0 | 54 / 118 | 45.76% | โ |
Greedy, rows 0 + 46 fused (logaddexp) |
39 / 118 | 33.05% | -12.71 pp vs unfused (27.8% fewer errors) |
| Fused + KenLM (alpha=0.30, beta=0.50) | 27 / 118 | 22.88% | -10.17 pp vs fused greedy (30.8% fewer errors) |
Frame-level argmax on one clip (Track-18.wav) showed row 2 (|, 195 frames) and row 46 (<pad>, 98 frames) dominating. Masking row 46 without fusing it into row 0 pushed blank frames onto row 0 or row 2, which injected literal PAD strings and stray spaces into words (LEPADCTURE NOWPADTES, EPADXAPADM). On the same probe, raising alpha to 0.60โ0.95 gave 26.27%โ33.9% WER because the LM began overriding correct diacritics and merging compound English nouns. The probe was used while debugging the decoder; it is not an independent estimate.
Example clips
Selection rule: the two clips where the LM helped most, the median clip, and the two where it helped least or hurt (sorted by WER change).
| Clip | Reference | Greedy (fused) | KenLM | Greedy WER | KenLM WER |
|---|---|---|---|---|---|
| AR_439.wav | MO RA TRAVEL INSURANCE NรTORร Pร รRรNรJร ร Jแป OHUN Tร A Lร FOJร KรRร Sร | MOO TRAVEL INSแนขURANCE NรTORร Pร รRรNรJร OJรHรN TI ALร FรJร K | MO TRAVEL INSURANCE NรTORร Pร รRรNรJร OJร OHUN TI A Lร FOJร E | 75.0% | 37.5% |
| AR_428.wav | WแปN BรรRร FรN WRITTEN STATEMENT MI | WแปN BรรRร FรN WRITEN STATEMENT M | WแปN BรรRร FรN WRITTEN STATEMENT MI | 33.3% | 0.0% |
| AR_420.wav | MO Nร LรTI Lแป Sร HEAD OFFICE LแปLA NรTORร Pร WแปN TI BLOCK GBOGBO ONLINE TRANSACTIONS Tร MO Fแบธ แนขE | MO Nร LรTI Lแป Sร HEAD OFFICE LแปLA NรTORร Pร WแปN TI BLOCK GBOGBO ONLINE TRANSACTIONS Tร MO Fแบธ | MO Nร LรTI Lแป Sร HEAD OFFICE LแปLA NรTORร Pร WแปN TI BLOCK GBOGBO ONLINE TRANSACTIONS Tร MO Fแบธ แนขE | 5.0% | 0.0% |
| AR_413.wav | GBOGBO รWแปN Tร A INVITE Sร WEDDING RECEPTION YแบธN GBแปDแป MแปRA Kร WแปN Mร BAร MISS OUT LรRร BUFFET แปSรN Nรร | GBGBO รWแปN Tร A INVITE Sร WEDDING RECEPTION YแบธN GBแปGUN MแปRA Kร WแปN Mร Bร MISS OUT LรRร BUFFET แปSรN | GBOGBO รWแปN Tร A INVITE Sร WEDDING RECEPTION YแบธN GBOGUN MแปRA Kร WแปN Mร BA MISS OUT LรRร BU FE แปSAN N | 19.0% | 28.6% |
| AR_446.wav | Bร A Bร PรN HOUSEHOLD RESPONSIBILITIES รGBรSรAYร WA ร RแปRรN | B A Bร PรN HOUSEHOLD RESPONSIBILITIES รGBรSรAYร WA ร RRร | Bร A Bร PรN HOUSEHOLD RESPONSIBILITIES รGBรSร AYร WA RรN | 20.0% | 40.0% |
Caveats
- Small and narrow. 749 words from one corpus of read studio recordings; speaker count is not recorded. Naive 95% intervals above are about ยฑ3 pp and are too narrow because errors cluster within clips.
- Tuned on the evaluation clips. The alpha/beta comparison was run on these clips, and the shipped unigram list (
50,259words) includes 3,000 extra words (common English from the Brown corpus plus a hand-curated list). The curated words were added after inspecting errors on these clips (for exampleRABBITafter theAR_407.waverror). The rows marked "tuned" are therefore not clean held-out numbers. The cleanest row is the one with Phase XI's default settings and the corpus-only word list. - Domain match. The LM text includes the 376 training prompts from the same
AR_metadatacorpus, repeated 15ร (none of the 50 held-out prompts). The held-out clips come from the same prompt set, which favours the LM compared with unseen speech. - Not comparable with Phase X / XI numbers. Per the Phase XI card, Phase X acoustic fine-tuning used most of
AR_Audios(55 of ~550 clips were excluded), and its 4.5% figure came from a dev set the decoder had been tuned on. Phase XII acoustic training did not includeAR_Audiosaudio (mix: FLEURS Yorรนbรก, Google FLEURS English (en_us), YECS). The evaluations measure different things. - Scoring. Not recorded in this card. (The Phase XI card used uppercase + NFC + punctuation removal on both sides.) Whether
phonetic_map.jsonpost-processing was applied to the WER figures is not recorded in this card. - Not measured: WER by language (Yorรนbรก vs English vs code-switched stretches), tone-mark accuracy, code-switch-point accuracy, named entities, noisy or 8 kHz audio.
Acoustic model
- Base:
facebook/wav2vec2-large-xlsr-53(Wav2Vec2ForCTC); 424 F32 tensors, 315.5M parameters, 1.26 GB. - Frozen: the convolutional feature extractor and the lower 12 transformer layers; per the training script only the upper layers and
lm_headwere updated. - Schedule: 10 epochs, 2,950 steps, wall time
5:47:44, Kaggle multi-GPU, started fresh from the base weights. - Data: Google FLEURS Yorรนbรก (
yo_ng), Google FLEURS English (en_us), and the YECS YorรนbรกโEnglish code-switching corpus. Per-source hours and mixing ratios are not recorded in this card. - Output head:
lm_head.weight[47, 1024]. Before training, the script foundpad_id(46) equal to the matrix size (46) and expanded the head by one row.config.jsonhasvocab_size: 47andpad_token_id: 46. - Vocabulary:
vocab.jsonhas 46 entries (ids 0โ45):[PAD]=0,[UNK]=1,|=2,<s>=40,</s>=41. Added token:<pad>=46 (added_tokens.json).
| Step | Training loss | Validation loss |
|---|---|---|
| 500 | 11.866803 | 5.210579 |
| 1000 | 8.790150 | 3.593741 |
| 1500 | 6.799892 | 2.816854 |
| 2000 | 6.284078 | 2.569384 |
| 2500 | 5.873839 | 2.456798 |
Validation loss fell at every logged evaluation (5.21 at step 500 to 2.46 at step 2500), but the gain per 500 steps shrank from 1.62 to 0.11. The run stopped at its fixed 10-epoch budget, not at a stopping criterion; no evaluation is logged at the final step (2,950). The composition of the validation split is not documented here.
Language model and corpus
Master corpus audit. voxira_phaseXII_master_corpus.txt: 381,469 unique normalised lines, including 3,274 with Q, 3,777 with X and 3,219 with Z; 8,274 of them are web-mined sentences collected to cover those letters. Sanitising removed 6,927 web / citation boilerplate lines and capped 28,839 repetitive 3-word Yorรนbรก openings at 100 per prefix, leaving 345,703 clean lines:
| Class | Lines | Share of clean pool |
|---|---|---|
| Yorรนbรก | 270,862 | 78.35% |
| English | 42,443 | 12.28% |
| Code-switched (YECS) | 21,980 | 6.36% |
| Unclassified short fragments | 10,418 | 3.01% |
Balanced 40 / 30 / 30 training corpus (141,476 lines): Yorรนbรก 56,590 (conversational 4โ18-word sentences prioritised), English 42,443 (all clean English, including the Q / X / Z sentences), code-switched 42,443. The code-switched bucket contains the 21,980 unique YECS lines and the 376 AR_metadata training prompts repeated 15ร (5,640 lines); the remaining 14,823 lines are not itemised in this card.
AR_metadata. 449 rows, 426 unique prompts (23 duplicate rows; AR_277.wav absent). Words per sentence: mean 11.63 (SD 5.81), median 11, range 3โ31. Characters: mean 57.26 (SD 29.31), median 54, range 14โ154. 72 sentences (16.04%) contain Q (11), X (45) or Z (20). The last 50 unique prompts were held out; the first 376 went into the LM.
KenLM (lmplz -o 4 --prune 0 1 1 2): 75,796 unigram types (1,832,905 tokens); higher orders kept:
| Order | Kept | Total | Retained |
|---|---|---|---|
| 2-gram | 161,969 | 558,686 | 28.99% |
| 3-gram | 165,990 | 963,410 | 17.23% |
| 4-gram | 48,550 | 1,088,627 | 4.46% |
Unigram list (language_model/unigrams.txt): 50,259 words.
Decoding pipeline
- Run the acoustic model; take the raw logits.
- Fuse the blank rows:
logits[:, 0] = np.logaddexp(logits[:, 0], logits[:, 46]). - Mask control rows listed in
decoder_config.json([1, 40, 41, 46]) to-100. - Beam search with
pyctcdecode(beam_width=100;alphabet.jsonuses""for the blank at row 0 and placeholder symbols for masked rows). - Uppercase, NFC, strip punctuation.
phonetic_map.json(50 rules) is optional and not applied in the quickstart: Several rules were written in response to errors seen on the evaluation clips (for example theRABIES/RABBITcase above), so do not apply the map when benchmarking on those clips.
Known limitations
- Blank fusion is mandatory (see above). Anything that decodes the raw logits directly, including
processor.batch_decodeandWav2Vec2ProcessorWithLM, will emitPAD/[UNK]artefacts and extra spaces. - Phonetic English spellings without the LM. After 10 epochs the raw acoustic output often spells English words phonetically (
ACOUNT BAANC,ORNNOUT,HEADAICH,WRITEN,EXAM UESTIONS). Greedy WER is 26.96% on the clean takes. Q, X and Z are in the vocabulary, but no per-letter evaluation exists, andEXAM UESTIONS(forQUESTIONS) shows a droppedQin greedy output. - High alpha overrides acoustics. Alpha โฅ 0.60 degraded the probe (see above); the tested range with good results is 0.25โ0.35.
- KenLM
<unk>penalty on zero-count words. A word can be inunigrams.txt(so the prefix trie allows it) yet have no occurrences in the LM text, so it receivesunk_score_offset(-10.0). InAR_407.wavan acoustically close in-corpus word (RABIES, seen 3 times) beatRABBIT(zero occurrences) when the audio ended in a sibilant. - Evaluation limits (see Caveats): one small corpus, tuned on the evaluation clips, LM favoured by a matching prompt set, no per-language breakdown.
- Data provenance. 8,274 LM sentences were mined from the web, and sources and licences are not itemised here. The code-switched LM bucket contains 14,823 lines whose origin is not itemised. Check the licences of FLEURS, YECS and any web-mined text before commercial use.
- Audio scope. 16 kHz mono read / studio speech. Not tested on 8 kHz telephony, overlapping speakers, background music or noise, or Nigerian Pidgin.
- No punctuation or casing; output is uppercase. Not a speaker-verification model (do not use transcripts as proof of identity). Human review is advised for legal, medical and financial actions triggered by a transcript.
Quickstart
pip install torch transformers librosa pyctcdecode kenlm huggingface_hub
import re, json, unicodedata
import numpy as np, torch, librosa
from huggingface_hub import snapshot_download
from transformers import Wav2Vec2ForCTC, Wav2Vec2Processor
from pyctcdecode import BeamSearchDecoderCTC
REPO_ID = "Awesome-x/voxira-yoruba-phaseXII"
device = "cuda" if torch.cuda.is_available() else "cpu"
repo_dir = snapshot_download(REPO_ID) # acoustic weights + language_model/ + decoder files
processor = Wav2Vec2Processor.from_pretrained(repo_dir)
model = Wav2Vec2ForCTC.from_pretrained(repo_dir).to(device).eval()
decoder = BeamSearchDecoderCTC.load_from_dir(repo_dir) # alphabet.json + language_model/ (alpha/beta stored in attrs.json)
with open(f"{repo_dir}/decoder_config.json", encoding="utf-8") as f:
MASKED_ROWS = json.load(f)["ctc_blank_fusion"]["masked_rows"]
def transcribe(wav_path, beam_width=100):
audio, _ = librosa.load(wav_path, sr=16000, mono=True)
inputs = processor(audio, sampling_rate=16000, return_tensors="pt")
mask = inputs.get("attention_mask") # may be absent (return_attention_mask=False)
with torch.no_grad():
logits = model(inputs.input_values.to(device),
attention_mask=mask.to(device) if mask is not None else None).logits[0].float().cpu().numpy()
logits[:, 0] = np.logaddexp(logits[:, 0], logits[:, 46]) # REQUIRED: fuse the two blank rows
for i in MASKED_ROWS: # mask control rows (incl. row 46 after fusion)
logits[:, i] = -100.0
text = unicodedata.normalize("NFC", decoder.decode(logits, beam_width=beam_width)).upper()
text = re.sub(r"[^\w\s]", "", text)
return re.sub(r"\s+", " ", text).strip()
print(transcribe("sample.wav"))
Tip: split audio longer than ~30 s into chunks before decoding. This exact code was executed against this repository when the card was generated: pyctcdecode 0.5.0, transformers 5.0.0, torch 2.10.0+cu128.
Repository contents and checksums
| Path | Size | SHA-256 |
|---|---|---|
added_tokens.json |
18 B | fc85bf44b931cbd172549d2a7b7eec19276a7900b8c28b00663fe46bcb144f55 |
alphabet.json |
353 B | 84ee61564c87e983391b920967501c4d9ea3793afd0eae8a009597d02b887946 |
config.json |
2.3 KB | 52680cfdf5ea845f72c995cc7b62b04692bd8447b3bbd32d1ff84f76f0707127 |
decoder_config.json |
1.0 KB | a5989b6abf6c22a3595a89bdabb11f2075a9059d4062cfc57faad027f4b572a0 |
eval_results.json |
24.1 KB | 6b1e7deb9c32cc6e59e06fd0226aa118216dc3ca4ac074776a4780105efb2c37 |
language_model/attrs.json |
79 B | 1e8419cf7292cde4f6c0633065524fd9c67aa95a9e70dd5b271251bc12c6bfde |
language_model/phaseXII_40_30_30_prune.bin |
11.35 MB | 7b9ae0e666f10d4637012accd28787567cb6546c00c605a51aa6f71913b4f4b3 |
language_model/unigrams.txt |
404.2 KB | cf52f1839d28525ed44510c09404c2d16169fdbef2aa7eee3bdc111122c72541 |
model.safetensors |
1.26 GB | d3917214a410de5379f399ad2629b5ec80041fac57b0d5cb152dee1e57bb0750 |
optimizer.pt |
1.28 GB | bee74376f2d6730131f871a295189a18d1e51046d821fbeb5ad1042f9889ec3f |
phonetic_map.json |
1.8 KB | a9d6993b27cb8ecaa1a2a0e001acfb0eaa5cd4c761309cc9fd726f8a07611028 |
preprocessor_config.json |
299 B | 8bc3c8d6d97ea61524d5c15d8554b4baf072d2264c0bda1839bc7a822c200d66 |
rng_state.pth |
14.7 KB | 15dbe4459d4ef773061c09e46ebef0fde526a9dd25f0f162fd73d243ecd033c3 |
scaler.pt |
1.4 KB | 56b8967fdc7656e7c10cf5dc67f14539386e5b031d327d5a70d5f68fbdcd5a80 |
scheduler.pt |
1.5 KB | cc59dc5dfd63ab3e3b656944bc3a6031df082267f7df62d6f7e727f91be0a87d |
tokenizer_config.json |
1.4 KB | 15d1f2504bd244010615487c8a003e29831f15de528ea1021a35bb597c384237 |
trainer_state.json |
7.2 KB | 63a30846b66112f3727299e880e20ac502f5e3f5ee8b732efab721ffa668d045 |
vocab.json |
529 B | 4110c1ddbc5a6322867927977ad98a025b556335b00d4c1387fbb654e1ae0c2f |
Reproducibility
| Item | Value |
|---|---|
| Acoustic weights | model.safetensors, 424 tensors, 315.5M parameters (F32), lm_head.weight [47, 1024] |
| Training checkpoint | checkpoint-2950 (10 epochs, 5:47:44) |
| Pad ids | vocab.json [PAD]=0; config.json pad_token_id=46; <pad>=46 in added_tokens.json |
| KenLM | 4-gram, --prune 0 1 1 2, phaseXII_40_30_30_prune.bin, 10.83 MB |
| Decoder defaults | alpha=0.25, beta=0.5, beam_width=100 |
| Raw numbers | eval_results.json in this repo |
| Environment used to generate this card | torch 2.10.0+cu128, transformers 5.0.0, huggingface_hub 1.10.1, pyctcdecode 0.5.0, kenlm 0.3.0 |
| Card generated | 2026-11-13 13:54 UTC |
Deployment notes
The acoustic model is about 1.26 GB (F32), so cold start is dominated by loading weights; the LM is 10.83 MB. For serverless GPU hosts, bake the weights into the image or a volume rather than downloading on every start. kenlm compiles from source on most hosts. fp16 parity and CPU decoding latency have not been tested.
Not yet done
- A larger, independent held-out set (and speakers other than the studio recordings).
- WER broken down by language, tone-mark accuracy, named-entity and domain-term evaluation.
- 8 kHz telephony and Nigerian Pidgin data.
- Phase XIII (planned): a longer acoustic run (30 epochs).
Citation
@misc{voxira_yoruba_phaseXII,
title = {Voxira Yoruba Phase XII: Yoruba-English code-switching ASR},
author = {Voxira},
year = {2026},
url = {https://huggingface.co/Awesome-x/voxira-yoruba-phaseXII}
}
- Downloads last month
- 100
Model tree for Awesome-x/voxira-yoruba-phaseXII
Base model
facebook/wav2vec2-large-xlsr-53Dataset used to train Awesome-x/voxira-yoruba-phaseXII
Evaluation results
- WER (%), greedy CTC with blank fusion, no LM on Voxira AR_Audios held-out, 49 valid clipsSelf-reported: Voxira internal evaluation26.700
- WER (%), KenLM, 66,184-word list, alpha=0.30 beta=0.50 on Voxira AR_Audios held-out, 49 valid clipsSelf-reported: Voxira internal evaluation19.760
- WER (%), greedy CTC with blank fusion, no LM on Voxira AR_Audios held-out, 44 clean takesSelf-reported: Voxira internal evaluation26.960
- WER (%), KenLM, 66,184-word list, alpha=0.30 beta=0.50 on Voxira AR_Audios held-out, 44 clean takesSelf-reported: Voxira internal evaluation19.280