How often are languages updated?

I’m doing the Hebrew coursed and I’ve noticed quite a few sentences with mistakes and upon looking at them on Tatoeba I see they were updated years ago but the errors remain on Clozemaster, also the Hebrew TTS sucks, there are much better ones out there it would be nice if it was changed.

4 Likes

Thanks for letting us know! Updates from Tatoeba haven’t been pulled in in a while like you’ve noted, and we still need to find someone to help moderate reported sentences. The TTS has been updated for the 100 and 500 Most Common Words collections - might you be able to let me know if you notice a significant difference vs say the Fast Track or 1,000 Most Common Words collections?

4 Likes

It’s hard to do an apples-to-apples comparison of different voices when they’re reading different sentences, but I think the voice for the 100 and 500 Most Common Words collections is more pleasant than the one for the Fast Track. The latter voice sounds like an aggressive hipster, and that gets tiring fast.

To give a concrete example of bad TTS from a pronunciation point of view, here’s one from the >50,000 Most Common collection:

היא מזגה לי ספל תה.

Sentence source:
https://tatoeba.org/en/sentences/show/1447894

The cloze word should be prounounced like “mazGAH”, but is pronounced like “mzGAH” instead.

1 Like

Mike, this is a great change, the new TTS is a vast improvement, thanks! I noticed it this morning and then found this thread to confirm my observation. Not sure what is involved but it would be great to roll it out on the other Collections you mention if possible.

To elaborate, I had been finding the other TTS basically unusable. Today I popped in to try again. At first I thought “Wow, my listening comprehension just got a lot better!” until I was able to convince myself that the TTS had been changed.

2 Likes

Would it be possible to update the voices for the 1,000, 2,000, etc. categories too? Hopefully it’s a relatively simple API switch!