Improving Speech Recognition for Video Indexing

Today I played with PODZINGER, a company powered by BBN’s speech recognition technology. This company is a spin-off of BBN.

Speech recognition has been improving and is a good step in audio and video indxing. PODZINGER is different from other sites in its coverage of podcasts. The quality of speech in podcasts is not as high as in broadcast News video and audio. This is definitely a good step in making podcasts more useful. From technology perspective it is a good step in showing that application of speech recognition are incereasing and slowly extending to lower quality audio.

Their website says

Podcasts have been subjected to the same primitive search through categorization … until now. PODZINGER looks inside podcasts, not just the metadata, letting you search podcasts in the same way that you search for anything else on the web.

When you type in a word or terms, PODZINGER not only finds the relevant podcasts, but also highlights the segment of the audio in which they occurred. By clicking anywhere on the results, the audio will begin to play just where you clicked. There are also controls that let you back up, pause, or forward through the podcast. Or you can download the entire podcast.

Leave a Reply