What podcasting has been waiting for – Audio search
There can be a lot of great stuff in radio downloads and audio blogs.
I’m sure this is true, but I gave up on so-called ‘podcasting’ along time ago because I can’t find the good stuff.
Blogs have been a far more search-friendly medium because the text is its own metadata. Not so for audio, you have to listen to the whole thing to work out if any of it is of interest to you.
Most audio search systems currently work by indexing text metadata which may accompany and audio file, but this metadata is unlikely to accurately represent the contents, particularly if it’s a long programme.
Something which might make internet audio attractive to me again : HearHere. It uses speech recognition to scan audio files and then matches text search terms to the audio. There’s a good graphical interface to show you where in the file your search term occurs.
It resides within Pluggd which is a recommendations engine, and should go some way to helping you discover good things by accident, something else I love about listening to the radio.
However – speech recognition is particularly difficult to do well. How does the technology HearHere is using fair against the variety of accents and the poor recording quality of many audio blogs.