ZDT wishlist :P

[onebir]

Ok - this is quite cheeky; so first of all thanks very much to Bogleg for actually writing ZDT and giving it away...

To set the scene, one of the frustrating things about learning chinese is the split between the spoken and written language. Especially for me - a not writing, slightly reading, fluent-ish speaker who'd like to improve my vocab for speaking and reading purposes. This takes massive controlled repeated visual and aural exposure to new vocabulary. For me, anyhow...

So I'd like to be able to do is import chinese texts into a corpus, segment them, and automatically generate a list of single or multi-character new vocabulary. And using built in frequency data, to be able to completely exclude items way above my level, or filter them out when being 'tested' on flash cards.

For bridging the 语言/文字 divide, I'd like to have some way of integrating speech and text. Ideally there would be two ways to do this:

1) using text to speech. i'd like to be able to generate chinese english MP3 word lists from the filtered flashcard results.

(i just had a go at doing this with excel, but because the text to speech voice has to be set in the control panel, it's beyond my limited VBA abilities the euroasiasoftware/URL] chinese homework trainer can do it, but doesn't seem terribly stable on my pc)

2) somehow linking the texts with associated sound files (i remember reading there's an XML format for this - i guess ZDT would have to incorporate a viewer of this format) and being able to dig up example sentences from text-MP3 corpus, for use in 'aural flashcard tests' and/or generating MP3 word lists in this format:

chinese word, example in sentence, english word

(incidentally, it would probably make sense to record separate data for recognition of words aurally and visually. (context free) visual recognition is probably harder than aural recognition, which i think would have to be in context because chinese is so homophonic.)

Another difficulty i have is a 'word/character' divide. There are some characters i recognise in commonly used words, but have trouble with standalone. (eg 旅游 - if i see these separately i can never remember which is which). I imagine other people have similar difficulties, so a way of automatically tagging components of multi character words to their components would be useful.

Some way of tagging easily confused characters - based on a pre-built list of associations, or user defined - and using these as the distractors in multiple choice questions (or increasing the probability of their use as distractors) could be helpful.

Taking things to an (even) more complex level, if the software could parse as well as segment chinese, perhaps it could trawl the corpus for sentences with similar grammatical structures to textbook example sentences. (preferably filtering out ones that use to much unfamiliar vocabulary).

A final level of complexity would be to allow parallel corpuses: chinese and english side by side. The user could then search for chinese cognates of english phrases. (and if the software could also parse english grammar based on an example sentence, chinese cognates of english grammatical structures)

I recognise implementing these things would take a hundreds or thousands of programmer hours. But once implemented, it would create an environment that would allow learneres to take any source of chinese text (with or without audio) and get 100% out of the embodied vocabulary and grammar.

And I'm just too lazy to do it the hard way

[bogleg]

Hi Onebir,

Wow, that's quite a list. One of the things I was hoping for when developing the ZDT and by making it open source was that people could extend the program to do more ambitious things like you're suggesting. Unfortunately at the moment, I've been pretty swamped by my real life job so ZDT development has slowed to a standstill since the last release. When I get more time though, I'd love to talk to you in more detail about some of these features and maybe we can look to getting some of them implemented.

Thanks

Chris

[onebir]

Hi Chris

to be honest, i don't understand how someone with a real job has managed to find the time to take ZDT as far as it's gone!

onebir

[Fu Da-Wei]

The only significant item on "my" wishlist is somehow generating worksheets from multi-character hanzi. As it stands now, unless I'm missing something fundamental, you can only generate worksheets from single characters. The user can circumvent this by chopping up the compound into its constituent parts -- but surely that could be automated.

At the moment, I often keep two separate lists of words:

-----Vocabulary I wish to learn (using the flashcard module, for example)

-----That same vocabulary broken down into single characters, so that I can print worksheets.

Just a thought ...

;)

[thph2006]

Three additional requests:

1. Can a control be added to allow a user to restart a flashcard list where it was stopped the previous session? Imagine lists of several hundred words that you just don't have time to complete in a single session. It would be nice to be able to pick up where you left off when you return.

2. Allow for independent sizing of characters, pinyin and definition text in flashcards. Sometimes I need to increase the font size of the characters to really pick up all the detail but increasing the font size for all fields tends to make the definition text dominate the screen, especially when it is lengthy. It would be really nice for my old eyes to have the Chinese characters at something like 28-32 point and English text at 12 point.

3. Allow for display of pinyin using tone marks or tone numbers. This is another old eyes problem. Sometimes it's hard for me to distinguish the different tone marks when the font is below 18 point. Numbers are easier to distinguish than tone marks. Of course this feature wouldn't be necessary if Request #2 were implemented since I could just jack up the pinyin font size.

thanks!

[Luobot]

I recently discovered ZDT, here on the forums, and really love this program more than any other that I’ve paid good money for. There are a few enhancements that I would like to see, but the single-most important thing to me was mentioned by Onebir, already, and I would just like to add to it a little bit.

In short, the single-most important thing is sound, not text-to-speech, but good old-fashioned, warm-blooded human speech. I have lots of real, live native Mandarin speaking sound files that I would like to attach to the words in ZDT. I can also see exchanging sound files with others, just like the ZDT word lists, though my primary goal at the moment is to link up my own sound files.

This would not be to replace the current sound function (F3), but would be in addition to it (say, F4 or whatever), so that it would be possible to hear an entry both ways (e.g., F3 and F4). Ideally, there would be two ways to attach a sound file, manually (one at a time), and in batch mode (as in the import function). The manual method could work by right clicking a word and selecting “Add SoundFile”, causing a “Browse” dialog window to pop, where the user then locates the file on their hard drive. The user can enter the path or browse to it. The batch method could work like the current import a category function. The user would identify to ZDT a text file that contains either characters or pinyin, followed by a delimiter, followed by the path where the sound file is located. If characters are used, then the sound file would be attached to that exact matching character combination in the dictionary. If pinyin is used, then the sound file would be attached to all exact matching pinyin entries. So, for example, 是 would result in one row being updated. But ‘shi4’ would result in about 39 rows being updated (using the CEDICT dictionary as an example). Using pinyin is therefore more efficient for this purpose, as a single sound file would be applied to all exact instances of this phonetic.

I wonder if this is a feature that others would also love to have … and what about the idea of being able to exchange sound files with each other, just like word lists can be exchanged?

[bogleg]

Hi Luobot,

That's an interesting suggestion for adding sound files. What format are your sound files in? Are they sentences/phrases or are they single sounds? If it's the latter, you can already create your own custom sound plugin and use them. But attaching a sound (in general) to an entry is a cool idea. I won't promise anything, but I will look into it.

Chris

[Luobot]

Hi Chris –

Thanks for responding! I’m creating sound files in WAV (uncompressed PCM) format, but I can convert them to another format, if that matters. Note, when I say “creating” I mean downloading from other sources and editing them. I use WAV so as to preserve sound quality while editing.

I’m doing words (consisting of any number of syllables), phrases (consisting of a few words), and longer sentences (the sort that you would put a period at the end of). I am presently interested in matching sounds up to the words that already exist in your version of the CEDICT dictionary, though it should work the same for any dictionary that contains the matching word. Including phrases and sentences would be a very interesting possibility, but it’s probably a separate project. I’m guessing that it wouldn’t be hard: Just create a custom dictionary of phrases and sentences and presumably everything else should work the same. But I’m new to ZDT, so you would know better than I.

I was thinking that it would be easy for ZDT to pass the fully qualified drive:pathfilename.ext to the Operating System (i.e., Windows) and the System would then run the user’s default player for that filetype. Therefore, it doesn’t matter what format the file is in, as long as you have a player for it on your system. It could even be a video file that describes the history of that particular word.

This would provide several benefits:

1 – Chaining together single syllable sounds doesn’t take into account tone changes. A simple example is, ‘Bu4 Yao4’ should be pronounced ‘Bu2 Yao4’. And of course there are all the intricacies of consecutive third tones. Using a soundfile that was originally spoken as a single word provides a more accurate pronunciation tool.

2 – Provides a more natural sounding alternative to F3.

3 – Affords the ability to keep improving the quality of the sound file by substituting a better one, whenever the opportunity arises.

4 – Affords the ability to play other media files, such as video files that the user wishes to associate with the word.

Please let me know what you think … Thanks!

[flameproof]

1: To make it start faster

2: Import from Pleco

3: can paste a group of characters into a folder (currently can only do one at a time)

[onebir]

1 – Chaining together single syllable sounds doesn’t take into account tone changes.

I think text to speech would do this - although another thread here suggests the free chinese voice isn't great. It would also allow ZDT to read out example sentences. Flashcards (visual or audio) in the format 'chinese word/english word' have a major flaw in the sense that they don't provide examples of how to use the word (eg where it goes in the sentence) - so an example sentence field would be a major plus.

Even if text to speech is too tricky - & I have no idea of the technicalities - being able to associate english (or whatever) sound as well as chinese would make it possible to produce chinese-english audio flashcards - far more useful than isolated chinese words (especially when they're monosyllables )...

[Luobot]

Tone changes:

I think text to speech would do this

Question: How natural does it sound? If you're imitating it, you don't want to sound like a monosyllable robot, or like something from a text-to-speech answering machine. If it’s something that you wouldn’t mind sounding like in a job interview or dating situation, then fine.

Presently, if you listen to the example that I cited in ZDT, you'll notice that it isn't doing tone changes at all. Bu4 Yao4 is Bu4 Yao4, not Bu2 Yao4. This is because these sounds are produced as individual syllables spoken sequentially, rather than produced as one word.

Anyway, it would be great to have the freedom to link up any file of any type – voice, video, whatever – which is what I’m suggesting.

This can coexist with the current or perhaps improved text-to-speech solution, as you suggest, Chris willing.

[onebir]

Question: How natural does it sound?

The free MS chinese voice isn't great, but it doesn't pronounce separate syllables separately, I think it handles tone sandhi and some sentence intonation. There are better commercial voices apparently (see MandarinBoy's thread)

If you're imitating it

No one suggested imitating the computer. Text to speech would give the ability to generate audio flashcards from text-only sources - which are more common than text + audio material - for learning new words. People at/above about intermediate level will generally be able to pronounce these, because they'll have already learned to pronounce the same syllables occurring in other words...

If you have (much rarer) text+audio, sound files are the way to go for pronounciation purposes. But creating sound files manually by editing audio is quite labour intensive. So people might like to have the option of generating audio flash cards using text to speech even where they have audio.

So I agree it would be ideal to have functionality for both attaching sound files and text to speech (perhaps with an override), which would suit different groups of users needs... (But this would also involve several times more work for Chris I guess...)

[Luobot]

So I agree it would be ideal to have functionality for both attaching sound files and text to speech ... which would suit different groups of users needs

So I guess we agree ... now, if only Chris agrees

[bogleg]

I agree, I agree! =) I would probably say that (with no investigation) that attaching sounds to entries would be a more likely feature for inclusion. I'm kind of doubtful that I would be able to use a commercial text to speech product in open-source software. But i'll wait and see how mandarinboy responds in his thread.

Chris

[bogleg]

@flameproof

1: To make it start faster.

How long does the ZDT take to startup for you? On the computer I'm using at the moment, it takes approximately 7 secs.

2: Import from Pleco

It's coming... slowly but surely. I'm working on a custom import function that I will make sure supports Pleco.

3: can paste a group of characters into a folder (currently can only do one at a time)

You're talking about pasting into a category right? This should already be possible. You have to put the characters on a separate line though. Let me know how you're trying to do it.

Chris

[onebir]

@Luobot:

No harm done ;-)

@Chris:

The free MS chinese voice probably isn't adequate for listening comprehension/imitation, but I thnk it's good enough it's fine for reading out words for memorization purposes (since you can hear the syllables & tones, and it copes with tone sandhi). But I don't know if there are free voices for Linux/Mac users...

[Luobot]

Onebir -- Sorry

[i removed it from my original post because I misunderstood you, and therefore it was inappropriate.]

Chris – I’m really excited about it!!! Thanks so much for agreeing!!!

[bredmond812陈秉劲]

maybe a link to Palm OS5 would be cool too. of course, it would need to use Big5 for traditional and GB2312 for simplified, or there would need to be some way to convert it.

[bogleg]

@bredmond812

Could you explain in more detail what you mean by "a link to Palm OS5"?

[bredmond812陈秉劲]

sure, i mean a way so that i can do the flashcards on Palm (whose operating system is OS 5) does that make better sense?