Hi, I'm at the point I can understand most of what I listen just fine but still have a lot to work on. One of the things is date and basic arithmetic in Japanese.
When someone mentions a date, a day of the week, or basic arithmetic, I choke.

Do you have similar experiences? What did you do in these occasions?

Numbers are always a weak spot with me. Counters are rough, too. It's just hard to find a lot of example sentences that use counters.

I'd love some sort of program that just randomly spit out audio japanese numbers and dates, using pre-recorded bits assembled for each one. It might sound a little robotic, like when a computer telephone operator repeats back your input, "You said, 999-333-4000. Is that correct? If so, press 1 or say 'yes'."

I would listen for hours, just to get used to hearing and quickly constructing numbers in my head at conversational speed.

It can't be that difficult. I can almost visualize it in vbscript, at least.

Perhaps there's an audiobook on amazon.co.jp (or elsewhere) which is designed to teach elementary-level arithmetic to native speakers? Or maybe even a DVD of a kids maths program?

plumage's recommendation is the best idea i've ever heard (and i've heard it before) for numbers and arithmetic. it is easy to put together a script that generates one-off cards with TTS audio for numbers practice. unfortunately i've got a lot on my plate now or i'd look into doing it.

Isn't there some learning games on kids.yahoo.co.jp that do that? Not sure of the audio, but the math is there.

Numbers and dates, how to work with time etc were the first things we learned at my school in Japan so I'm all fluent with that now. As for dates, we simply repeated them constantly until the order was stuck. Might feel a bit brutal when one is used to mnemonics and SRSing, but I'd say it's one of those things which is worth rote memorizing. It's not all that much and it's useful to be extremely fast with it.

I wonder if you could set up Asterisk or Trixbox to use a Japanese language option, then just play with it over your own phone... on the PBX, of course, not over the phone lines... might be interesting.

One system that generates those dates is a voice mail system. Not sure if you have access to a Japanese phone, but it may be a possibility.

Another possibility is commercials for events.

If you want to hear lots of numbers, maybe you can do a youtube search in Japanese, keywords - math class　or counting or phone number etc.. See what comes up.


数える

電話番号

The series of CDs 「羊でお休み」might be helpful. They are CDs with seiyuu counting sheep up to 400 on each CD (some CDs split the counting among 2 seiyuu, some CDs have one seiyuu doing the whole 400). I found a mediafire link to one of the CDs through google: http://www.mediafire.com/?72nhzmfi0mm
Not sure which seiyuu it is though. I have an iPod designated for sleeping time and I have a couple CDs of sheep counting on it. It's helped me get used to how counting sounds during "half-awake" periods. If you have yet to used to how the number+counter combination sounds (for example, ippiki, and then H sound for 2: nihiki) this should help at least a little.

Yeah, I was thinking you could write an asterisk script that could be a sort of date/number quiz thing... if there's a Japanese voice built-in to it. Might be something to play with when you're bored.

To learn dates/days, I just drilled them by rote. It was dull, but effective. Right now, I'm searching through Bk1 and Yahoo for some decent books on counters. I found a few promising ones, but no way to tell how good they are:

数え方の辞典, ISBN: 978-4095052014, 408 pages, 2310 yen.
数え方でみがく日本語, ISBN: 978-4480687180, 160 pages, 756 yen.
数え方クイズ100, ISBN: 978-4094186123, 141 pages, 520 yen.

It would probably help to search for 数え方, though, and see what you get.

I did Indonesian for 5 years at highschool and still choked on longer numbers and sequences of dates. I think that if you don't get extra practice from somewhere (and this was my pre-SRS days, and my pre-AJATT days so I was just plodding along to class) then numbers and dates are one of those things that are very difficult to grasp.

So anyway, I also think that program idea is great. I've wished a couple of times that I could get something to spit random japanese numbers at me. Unfortunately I have no programming experience.

My idea would be that it would just loop, continually generating a random number and playing it, then pausing a moment and then repeating. You wouldn't have to do anything but start it and then do something else while it recites numbers/dates all day.

Perhaps this could go in the "Group Study" area?

Creating the actual program is extremely simple, I'm pretty sure anyone here could do it or learn to do it in a week.

The problem is japanese audio, voice-systems are extremely hard to program (which is why they are so expensive to buy).

I'd guess that maybe .NET has some form of functionality to use Windows built-in voice generator... but that would still require the user to have some form of good japanese computer-voice installed.

Does the audio have to be from TTS? Can't you use audio form a site like this? http://japanese.about.com/od/japanesevoc...umbers.htm

yukamina Wrote:Does the audio have to be from TTS? Can't you use audio form a site like this? http://japanese.about.com/od/japanesevoc...umbers.htm

That would limit the program so it would only be able to work those numbers. Of course, you could play several sound files in a row, such as "ni-hyaku-sanjuu-go" but it wouldn't sound natural and it wouldn't be even close to natural speed anyway. The program would also be enormous with that many sound files included.

Tobberoth Wrote:The problem is japanese audio, voice-systems are extremely hard to program (which is why they are so expensive to buy).

In python,

Code:

import codecs
from comtypes.client import CreateObject

stream = CreateObject("SAPI.SpFileStream")
tts    = CreateObject("SAPI.SpVoice")
format = CreateObject("SAPI.SpAudioFormat")
voices = tts.GetVoices()
for i in range(len(voices)):
    if voices[i].GetDescription() == u'VW Misaki':
        tts.Voice = voices[i]
        break

from comtypes.gen import SpeechLib

format.type = 34
stream.Format = format
stream.Open("out.wav", SpeechLib.SSFMCreateForWrite)
tts.AudioOutputStream = stream
tts.Speak(u"今日は世界", 0)
stream.Close()

mafried Wrote:

Tobberoth Wrote:The problem is japanese audio, voice-systems are extremely hard to program (which is why they are so expensive to buy).

In python,

Code:

import codecs
from comtypes.client import CreateObject

stream = CreateObject("SAPI.SpFileStream")
tts    = CreateObject("SAPI.SpVoice")
format = CreateObject("SAPI.SpAudioFormat")
voices = tts.GetVoices()
for i in range(len(voices)):
    if voices[i].GetDescription() == u'VW Misaki':
        tts.Voice = voices[i]
        break

from comtypes.gen import SpeechLib

format.type = 34
stream.Format = format
stream.Open("out.wav", SpeechLib.SSFMCreateForWrite)
tts.AudioOutputStream = stream
tts.Speak(u"今日は世界", 0)
stream.Close()

Assuming the person running the program is in possession of a TTS voice which can output Japanese decently (Your example Misaki is only available illegally). It's coding that voice which is hard, not using the voice in a program.

Tobberoth Wrote:Assuming the person running the program is in possession of a TTS voice which can output Japanese decently (Your example Misaki is only available illegally). It's coding that voice which is hard, not using the voice in a program.

Misaki is available from Voiceware. If you don't want to use an existing TTS engine, then do what was suggested and chain together freely available sound files. That's how a TTS engine works under the hood.

mafried Wrote:

Tobberoth Wrote:Assuming the person running the program is in possession of a TTS voice which can output Japanese decently (Your example Misaki is only available illegally). It's coding that voice which is hard, not using the voice in a program.

Misaki is available from Voiceware.

Wow, they won't even write out the price? I'm going to have to assume it's insanely expensive.

Yes, I assumed it would utilize very small, mono, low bitrate mp3 files or something for each portion of the number/date being spoken. Apparently there are ways for a server-side script to build an .m3u file dynamically--this would seem to be the best way to keep the files from sounding way choppy and broken up. I mean, it still wouldn't be "natural" but if you need something like this, the very short (but unnatural) pauses would be a welcomed mental processing aid.

Adding in a timed wait and then looping would get you constantly churned out numbers without having to reclick over and over.

Why use TTS to make snippets? Why not just download the MP3 audio from Smart.FM? Surely they have all the words for numbers, then just join those up. TTS may sound better though if you give it an entire number, like 1234 and not 1, 2, 3, 4 individually.

If you're just doing numbers, you can use the cheaper, previous-generation NeoSpeech JP voices, which run ~$30. They don't sound as good as Misaki, but they'll do the job.

On a slightly different tract from a random number generating program, what about using natural sources? Take snippets of numbers from news or TV shows. It might not be exactly the number you want and it would be more work, but it would sound natural. It could always be supplemented by a TTS program?