+- kanji koohii FORUM (http://forum.koohii.com)
+-- Forum: Learning Japanese (http://forum.koohii.com/forum-4.html)
+--- Forum: Learning resources (http://forum.koohii.com/forum-9.html)
+--- Thread: Podcastle.jp: The text of Japanese podcasts (/thread-4030.html)
Podcastle.jp: The text of Japanese podcasts - ahibba - 2009-09-18
Podcastle is a service that lets you search the audio of podcasts in Japanese. Voice recognition technology converts the audio into text. Users can then freely edit any recognition errors.
http://podcastle.jp
Podcastle.jp: The text of Japanese podcasts - atylmo - 2009-09-18
Thanks for reminding me of this! I found it a little while ago, never posted about it though (sorry!
)It saved me so much frustration.. I was trying to figure out a couple words that were hard for me to catch (from 安住紳一郎の日曜天国) just by listening and repeating, slowing them down in Audacity, etc.
Then I found Podcastle. And I proceeded to smack my forehead.
Podcastle.jp: The text of Japanese podcasts - Nukemarine - 2009-09-19
ahibba Wrote:Podcastle is a service that lets you search the audio of podcasts in Japanese. Voice recognition technology converts the audio into text. Users can then freely edit any recognition errors.Wait a minute, so if someone loads up the audio of say: GTO or IWGP (Two great dramas without subtitles available) on a site like Dropbox, this thing could convert it to text which might be made into subtitles?
http://podcastle.jp
Podcastle.jp: The text of Japanese podcasts - ruiner - 2009-09-19
This is what I was wondering (re: Nukemarine). I was under the impression that speech-to-text/audio mining was still in its infancy. The most recent examples I can find are here: http://www.beet.tv/2008/07/youtube-has-voi.html - What's the current status w/ those search engines like Blinkx/EverythingZing, I wonder.
More on podcastle: http://www.streetsmartlanguagelearning.com/2009/09/get-text-of-japanese-podcasts-with.html
http://staff.aist.go.jp/m.goto/PAPER/INTERSPEECH2007goto.pdf
Can't find anything about standalone software or adding your own audio.
Relying on 'social correction' to offset the 'beta' aspect of the tech (thus Podcastle is in a state of 'permanent beta' according to the .pdf) is interesting and probably the best we can have for a long while.
It has a web crawler, not sure how that works or whether users can directly add audio, and what the deal is with copyright. I also wonder what software they're using and whether it's available elsewhere.
Still, I can totally see this getting expanded soon. Googling revealed a cpl venture/startup language-learning groups are already excited about Podcastle, maybe they'll try to integrate something similar to smart.fm or something. That would take care of the social correction aspect, though we'd lose out on the customizable/copyright stuff again...