Beta 7 Audio Error Thread

mikelove

皇帝
Staff member
I'm making a separate thread for these, since with the full set of audio files now available there'll probably be a lot more people making heavy use of audio, and having them spun off here will make it easier to check for duplicates. So if you find an entry that doesn't match up with its audio recording, post the pinyin/gender here and we'll make sure it gets fixed. (don't report static-y/noisy ones here unless they're completely inaudible - we already know about the issues with those but unfortunately we can't fix them until after 2.0's out)
 

Alexis

状元
Hi Mike,

Is there more value in testing with the "Skip words without multi-syllable audio" setting enabled or disabled?
 

mikelove

皇帝
Staff member
That wouldn't matter as far as finding mistakes - if there is a multi-syllable recording for a word in the database it'll pull it up regardless of whether or not that option is enabled.
 

sych

榜眼
All the audio errors that I had found in B6 (but hadn't reported yet) are fixed now, so I've got nothing to report anymore!
 

sych

榜眼
Male and Female Voices: bu4zhi4.

bu4zhi4 is always read as bu2zhi4. This is ok where "bu" is 不, but AFAIK not OK in other cases, e.g. 布置 should be read as "bu4zhi4", not "bu2zhi4".

I've asked a Chinese friend and he confirms 布置 should be 4th-4th. We couldn't find any other case where bu4 can change to bu2 unless it's 不. If you'd like me to do some more research on this (e.g. ask my teachers at BLCU), I'd be happy to.
 

mikelove

皇帝
Staff member
Good catch on that, not something we'd actually thought of in the design (embarrassingly enough) - there aren't too many cases like this (looks like about 10 for bu and 2 or 3 more for yi out of 34,000+ samples), fortunately, but we definitely need to check for them / only play back the sound when it's correct. We actually even have a correct recording for 布置, in fact, just didn't support it in the database. For now I think we'll probably just add a quick-and-dirty little check which only plays the pinyin for those couple of recordings when the first character is a 不 or 一, since not playing the recording at all is better than playing it incorrectly.
 

ipsi

状元
One of the speakers (Female, maybe?) says Han4bao3bao3 instead of han4bao3 for hamburger (can't write Chinese at work...)

Also, in the Tuttle entry for gao1ji2, it has two examples at the end, both of which start gao1ji2xiao3xue2. The second one should be gao1ji2zhong1xue2. Should be pretty obvious what I mean, especially as the pinyin is right but the characters are wrong :).
 

goog1e

举人
bao3bao3 宝宝 male is pronounced like bao3bao1

I can't say if this is really wrong but it's different from the female who says bao3bao3
 

Alexis

状元
I'm going through the Oxford Dictionary in Beta 7 starting from "a1", (I'm at currenkt at ben3) and have found 170 or so mistakes so far. I am currently adding them to flashcards as I go along. Probably take me a week or two before I can finish going through it. Will it be sufficient to send the flashcard file when I'm done (without specifying the male/female, you'd get to figure that out! :) ?

- Alexis
 

mikelove

皇帝
Staff member
Alexis - 170 mistakes before ben3? That's odd - in our earlier (pre-B7) testing we ran through about 5000 for each speaker and only found 20 or so mistakes. Anyway we'd certainly love to have that flashcard file when you're done, sending that along would be all we need - many many many thanks for doing this. If anyone else wants to do audio testing in a methodical way like this, perhaps we can find a way to divvy up the alphabet to cut down on the workload.
 

Alexis

状元
Hi Mike,

I double-checked again and I exaggerated a bit, it was in about the 130 region or so. I'll double check the ones I identified for A/B and send the flashcard file your way.

Would be great if we could divy up the testing a bit by letter. :) Any takers?

- Alexis
 

Alexis

状元
The good news is that i discovered that having the volume cranked pn my ipaq 210 and listening to the speakers made a lot of the "b"s sound like "m's" and "l's" at the beginning of words. I've started to cut down the list by listening with headphones and the number has come down drastically! (A few 10's).

- Alexis

PS. Is it possible to get the audio to play from the "manage flashcards" screen?
 

mikelove

皇帝
Staff member
Great! Was worried there, if something like 1/5 of our audio samples were wrong that would be a serious, delay-the-release-for-a-month type thing... no way to play it from Manage Flashcards, unfortunately.
 

marsch

举人
Hi. A few things sound different from what I would expect. But that could also be due to poor hearing and/or poor earphones...

1. zhang3shang4xing4dian4nao3, fem: zhangshangdiannao
2. mei2tou2 – brows - female meitouzi, male: meikou
3. te4zhi3 – to refer in particular to - male sounds like ke4zhi3
4. qu1xiang4 – trend, tendency – male/female sound like xuxiang
5. wa3 – watt – fem. like ba3
6. xiang4ce4 – photo album – fem. like xiang4ke4
7. wen1huan3 - male “gun1huan3”
 

mikelove

皇帝
Staff member
A lot of these I think may be headphone / audio-compression issues; #s 1 and 2 female are definitely wrong, though. With #7, do you mean "wen1nuan3"? There's no entry for "wen1huan3" in any of our dictionaries.
 

anchan42

探花
很想 should pronounce hen2xiang3, pleco pronounce hen3xiang3
挺爽 should pronounce ting2shuang3, pleco pronounce ting3shaung3
ie. 3,3 sound change to 2,3 sound problems.

省 women sound like a man. Not sure it is a bug though. :)
 

mikelove

皇帝
Staff member
That's happening because Pleco doesn't have specific recordings for those words, so it just pieces them together out of the component syllables; coding in the rules for sandhi tone transformations like 3,3 to 2,3 is really tricky, so we opted not to do that in 2.0 at least. If you'd rather Pleco not play audio when it doesn't have an exact recording for it, enable the "Skip words..." option in the Misc panel of Preferences.
 
Top