travisvroman.com archives

This site contains archived posts made before July 2022. If you're looking for the live site, you can find it here.

Podcasts + Text = Podcastle

Tue May 08 2012
Part of immersion is listening to audio, such as podcasts. However, most Japanese podcasts do not come with written transcripts. This represents an issue when studying because it can be difficult to correctly understand what you hear. This is where Podcastle comes in. Podcastle keeps up (for the most part) with the audio currently playing. Podcastle is a project funded by the Japanese government for foreign Japanese language learners. It provides a speech recognition system to automagically convert the audio into text. While you play the audio, it tries to keep up but can sometimes become "desynchronized". It helps if you know enough Japanese to be able to tell when it's messed up, as it can sometimes be off by a few words. Knowing kana at a minimum may be enough to be able to tell this. Podcastle's community can edit transcripts for accuracy. While its speech recognition is not all that accurate (let's face it, though - none are yet), Podcastle provides methods for its users to make "edits", or corrections, as well as confirm correct portions. Some clips have hundreds - even thousands - of edits, indicating that the user base is quite large. The "Full Text" option allows you to see all of the text at once, which is easier to read. Podcastle also works with YouTube videos. Check it out! See Also: