Datasets:

---

google

/

WaxalNLP

like 214

Follow

Google 47.8k

Thank you for creating this dataset.

I'm trying to use the "ewe_asr" portion of the data but it seems it has some issues with many digits in the transcriptions, unlike the TTS subset which seems relatively clean.

Here's a Colab notebook.

These are the first 3 transcriptions:

E3u ge2ee yi ke wo5e amadedewo le abe sika ene wole te5e sia, wole m4dodoa dzi eye wole amehawo ameha ge2e dome, ke ameawo h7 le wo dome tom. 
Ame a2ewo le s4l4mex4 me. Wo domet4 woame ene wole klo dzi, bubuwo h7 le tsitre, s4fo g7wo le v4samlekpui la dzi le gbe dom 2a 2e wo ta. Wodo awu l4b4l4b4wo ke`.  
~utsu a2e kple esr- kple eviawo wole a5eme, wo kat7 wo2o nukomo, `utsua kple ny4nua wo2o kuku, won4 anyi 2e a5ea me hele atikutsetsewo dom. Ati h7 le megbe na wo.