One of the things I've really liked about Skritter (apart from re-learning my character writing) is the visual and auditory reinforcement of tones. However, there are times, especially when just doing the reading/definition mode, where I'd like to repeat the sound on word I'm studying, just to drill it into my brain for when I have to say it. (I'm an English Mandarin interpreter, and I often have to add specialized vocab based on upcoming assignments, so it helps to get used to hearing it.)

Is there any way you could put a repeat button on the audio, or is it already there and I just haven't found it yet?

Thanks!

There is a small speaker symbol next to the star and magnifying glass on the left or right side (depending on your setup) that allows you to repeat the audio.

Hope this helps.

So, I found the repeat button (my mistake), and it works on most, but not all, words. For example, 恫吓, 猛击,and 捆绑. I realize that these are not words you use everyday (hopefully!), but since the sound for each of the characters is already in the sound library, would it be possible to combine 2+ sound files based on the pinyin and tone input to provide pronunciation for less common words?

You might want to check out Help tab for FAQ: What are the keyboard shortcuts?
"A or ': play the pronunciation audio" I use these often.

The concept of combining audio files to fill in words with missing audio has been discussed before, but I think it was ruled out. I think because they wanted to wait to make or get good recordings for all the words.

Combining to make pronunciations for words is what the text-to-speech on the example sentences does. Do you think that that style of pronunciation would be helpful on a word level?

Personally, I think it would be helpful, since the only time I hear words without a playback button is when I type them (they don't give the correct pronunciation after you press return). If I type them incorrectly, there's no auditory reinforcement of the correct pronunciation, even though the spelling and tones come up in the answer.

I personally wouldn't want to hear machine audio/non-native pronunciation. Can it even get tone sandhi vaguely correct? I think pts or someone was saying it didn't. Half third tones would be a tricky thing to sound right.

Byzanti,
I am not familiar with the word sandhi, but from the context I guess it means the way the tone of one syllable is affected by the tone of adjacent syllables. (For example, the half third tone). There is also the case of characters with multiple tone possibilities depending on the meaning, which I don't know if a machine would get right. In cases where the meaning does not change but the tone does, maybe the machine would do okay, for example with the change in the tone for 不，since it would just have to look at the next character's tone. Anyway, for all these reasons I agree that a natural voice is preferable on sentences, but I would say that the machine voice is better than nothing. Maybe when I go to China instead of asking me which country I am from they will ask me which planet I am from when they hear my robot inflected speech ;)

@podster Wikipedia has a pretty thorough rundown. Yes, a first approximation just looks at the tones in context. To really get it right is much more complicated but maybe not completely necessary for Skritter.

http://en.wikipedia.org/wiki/Standard_Chinese_phonology#Tone_sandhi

I imagine pts would still zing it for inaccuracies, but I would personally be happy if the robot voice just did the basic rules - two third tones in a row, "一" and "不", etc. I think the latter are already taken care of in the readings they have in their system.

James

Thing is, with unusual words or phrases, you have the opportunity to put tone marks in when you add it. For 99% of my words, the tones are correct (even when it comes to 3rd tones and other changers).

@jrc02k hmmm, do you mean you put in 3rd tones as 2nd tones? I think the majority of the entries do *not* do that, although they do generally follow the rules for 一 and 不. I think this is because it's the convention in a number of Chinese dictionaries. The style guide says

"Please follow the pinyin standard formatting, with one exception: do indicate tone sandhi on 不 and 一."

http://www.skritter.com/style_guide

The pinyin standard formatting rule is against indicating tone sandhi:

http://pinyin.info/readings/zyg/rules.html#x4.11

James

The last time I commented about the example sentence audio was just an attempt to check the spreading of the “since all the tones are right and so the audio is helpful” idea. The sandhi rules are applied more often than a few of us here may have thought. Take the following sentence with all characters in the third tone as an example: 我想找你給我買五種水果. One way of reading it is: wóxiǎng/zháonǐ/géiwómǎi/wúzhǒng/shuíguǒ, where ‘/’ indicates a pause. Except the last character 果, all the other characters marked as third tone in the pinyin are read as half third tones. But the synthesizer seems to be ignorant of this concept.

Then something more relevant to the current thread. One should not replace the third tones as second tones. For example, it is tempting to mark the 表 in 表演 as a second tone. But in 你表演, we still have to read 你 as a second tone despite 表 looks like a second tone. This is because a third tone read as a second tone is still a third tone, only the reading is changed. So, replacing the third tones will conceal this fact and lead to incorrect pronunciations.

I agree that “the machine voice is better than nothing” and I’m also constantly relying on synthesizers for reference. I can only say that not all the synthesizers are the same and there are better ones. Repeatedly hearing the readings during our practices will have them permanently burned into our heads, which will be extremely difficult to correct if they are wrong in the first place.

“Repeatedly hearing the readings during our practices will have them permanently burned into our heads, which will be extremely difficult to correct if they are wrong in the first place.”

Which is why I'd rather hear nothing... I'm not sure quite how strong that burn in is, but it's significant. When I think of individual words in my head, it's usually in the voice of the lady who does most of the Skritter sounds. I'm not sure I want that replaced by a synth.

What I meant is that with certain words when you add them, it asks for the pinyin with number marks because it's not yet in the Skritter library. In those cases, you choose the correct tone number. For most words where 一 or 不 should be a different tone, Skritter already has this input. I'm essentially talking about audio for compound words like 驾驶执照 or 家庭暴力 where 驾驶 and 执照 are both in the library, but the audio won't come up when they're together.

@jrc02k what do mean by "the correct tone number"? According to the style guide, you shouldn't change third tones to second tones etc. because of tone sandhi rules, only for 一 and 不. And since most of the entries in the database follow that rule, it would seem to pose a problem for a voice synthesizer.

James

When it comes up and asks whether you mean zhong1 or zhong4, you put the appropriate one for that word. I'm not changing anything that shouldn't be changed. Same with when it asks to choose the traditional character version.

@jrc02k I wonder if we're talking about two different things. You wrote

"Thing is, with unusual words or phrases, you have the opportunity to put tone marks in when you add it. For 99% of my words, the tones are correct (even when it comes to 3rd tones and other changers)."

I took that to mean that, if you were adding 你好 to the database for the first time, you would put it in as ni2hao3 rather than ni3hao3, because you said you were making the tones correct "even when it comes to 3rd tones and other changers". Is that accurate or do I misunderstand you?

James

I think we're talking about different things. I mostly mean in situations where the reading of a character changes (zeng/ceng, zhang/chang, xing/hang...) or where the tone of a character is different depending on the word (like 为). In cases like 你好, even though it's 2 3rd tones together, it's easy enough for your voice to naturally do the 2/3 combo without changing the database from the "correct" tones to the tones you would use to say something. Again, the main thing I was saying is that these words are already in the audio library, but that many of the compound words or less common words don't have a replay button, and it would be useful to have the audio files combined. You seem to be hyper-focusing on a small part of what I said, by which I only meant that if I were adding 长大 for the first time, Skritter would ask which pinyin and tone I wanted for 长. I'm not messing up the database with auditory pronunciation revisions. Hope that clarifies.