Beta 7 Audio Error Thread
- Thread starter mikelove
- Start date
Yeah, that's actually a software bug rather than mistake in the recording - the program that encodes the audio files ignored / skipped over ...s, so since there's both a 左...右... and a 左右 in the audio database it's pretty much random which one it hits. This should be fixed in b8.
Alexis
状元
Hi Mike,
I emailed you a-j errors flashcard file. Includes both man and woman.
I've ignored the main "accents" in the list (which my ears think they're hearing). So far:
Women
ca->ta
cai->tai
ce->se
chi->shi
chuan1->chun1
*uan -> *un
Man
Zi->ze (sometimes on the end of a syllable)
quan->chuan
Regards,
Alexis
I emailed you a-j errors flashcard file. Includes both man and woman.
I've ignored the main "accents" in the list (which my ears think they're hearing). So far:
Women
ca->ta
cai->tai
ce->se
chi->shi
chuan1->chun1
*uan -> *un
Man
Zi->ze (sometimes on the end of a syllable)
quan->chuan
Regards,
Alexis
Thanks very much for that. The accent situation should improve considerably at least for the male files with b8/rc1 - it turns out re-encoding them at just a 50% higher file size gets rid of a lot of the accent issues, so if other people feel similarly we may even make those slightly-larger files the only versions available.
(for me) The larger the better 
I know my Chinese listening is not good, but I have not issues with the man a woman voices. Maybe the man sounds clearer but both are Ok, and I feel the experience of to listen one a voice and next the other very convenient. So bigger files, I don't think in my ears or handset will make a difference. But, of course, can make for other ones.
I also want to confirm the limit of the Palm Tungsten E2 is 1GB. So, I found from I started to use Pleco Beta 2 I almost run out of space on my 1GB card. It is enough for Pleco 2 and other many programs, but I is not to put the mp3 and video files I use to study more Chinese. So, I have just decided to buy a new program (PowerSDHD) to have enough room for all this files in a 4GB card.
I also want to confirm the limit of the Palm Tungsten E2 is 1GB. So, I found from I started to use Pleco Beta 2 I almost run out of space on my 1GB card. It is enough for Pleco 2 and other many programs, but I is not to put the mp3 and video files I use to study more Chinese. So, I have just decided to buy a new program (PowerSDHD) to have enough room for all this files in a 4GB card.
2.0b7(d)
te4dian3 (characteristic) male voice sounds like ke4dian3
I'd also like to see higher quality audio files for both voices, especially the female. While the male might be better even after re-encoding the female voice, it's useful for me to hear both. And yeah, the male quan->chuan stands out because I've also been confused about this point. Certain discussions say it should be a certain way ( e.g. http://www.chinese-forums.com/showpost. ... ostcount=2 ) but it seems like native speakers even deviate.
te4dian3 (characteristic) male voice sounds like ke4dian3
I'd also like to see higher quality audio files for both voices, especially the female. While the male might be better even after re-encoding the female voice, it's useful for me to hear both. And yeah, the male quan->chuan stands out because I've also been confused about this point. Certain discussions say it should be a certain way ( e.g. http://www.chinese-forums.com/showpost. ... ostcount=2 ) but it seems like native speakers even deviate.
te4dian3 definitely sounds better with the new files - all of those hard consonants seem to come through a lot more clearly. The female files aren't going to be updated again until after 2.0 is out, though certainly we'll encode them at higher-quality once we do update them. (at this point it looks like we may not be able to do much with noise removal, but we can at least clean up all of the obvious mistakes) And yeah, there's a lot of deviation both in certain syllables and in tones - the 1/4 divide especially is something that both speakers seem to occasionally differ from our dictionaries a bit on (in ways other the standard sandhi transformations).
Wouldn't it just be easier to rerecord some of the words, even if with a different woman's voice? Especially for a very few "bad quality" words.
I don't know about mixing voices - if they're not fairly consistent they're not all that useful, even though we tried to pick people with relatively standard accents they still each have their distinctive ways of pronouncing some things. The unsalvageable ones we'll probably just delete/drop, at least for now.
Yeah, there are a couple of those in there - we weighed whether to take them out (couldn't find a way to recover the original recordings) but decided to leave them in as long as the second syllable was reasonably clear; to my ears at least there's just enough of an 'ng' at the tail end of that to distinguish it from wa/wan.
The female voice pronounces han4bao3 wrong. She says han4bao3bao1. The male voice of 2.0.0 pronounces it correct.
By the way. Will be an update of the female voice in the nearer future? I like it very much to hear the words spoken by two different speakers. Even if the quality of the female voice is worse it is still usefull. But I would appreciate if you could increase the quality a little bit like yo did with the male voice.
Daniel
By the way. Will be an update of the female voice in the nearer future? I like it very much to hear the words spoken by two different speakers. Even if the quality of the female voice is worse it is still usefull. But I would appreciate if you could increase the quality a little bit like yo did with the male voice.
Daniel
Thanks - I think we have that one down already, actually.
Not sure how soon the female voice update will be out; hopefully along with 2.0.1, though if that takes longer than expected we may push it back a bit - there's actually been very little e-mail on the subject so far, most people seem quite happy with the one voice.
Not sure how soon the female voice update will be out; hopefully along with 2.0.1, though if that takes longer than expected we may push it back a bit - there's actually been very little e-mail on the subject so far, most people seem quite happy with the one voice.
Dear Mike,
I use both voices, and I generally find the female voice better, except for a very small number that have some static. Even after upgrading the male voice to the bigger, better files, I find the male b, d and f pronunciation to be unclear. The female voice is much better on those consonants. (I have notes on what specific words are problematic for the female and male voices but haven't had time to type them in and send them to you.)
John
I use both voices, and I generally find the female voice better, except for a very small number that have some static. Even after upgrading the male voice to the bigger, better files, I find the male b, d and f pronunciation to be unclear. The female voice is much better on those consonants. (I have notes on what specific words are problematic for the female and male voices but haven't had time to type them in and send them to you.)
John