My Photo

Photo Albums

Research Nuggets

Blog powered by Typepad

« Why Apple Won't Buy TiVo | Main | Brightwhat? »

February 26, 2005


David Sabel

It's all about the meta data and the context. Right now text based information has a huge edge in this area over audio and video. Over time that would need to change for audio and video to catch up to text.

As far as meta data goes, web pages have titles, anchor text, keywords etc. As do PDF files. Google et. al. have spent millions making sense of all these meta data clues in order to make searching them fast, easy. The problem with audio, is other than artist, album and track names, there is very little *consistent* other information available in an electronic form to be searched and indexed. So while there are liner notes, which might have details on who sat in on what song and where the album was recorded, that information is invisible to all except those that have the physical CD in their hand. For video I think it is going to take not only voice to text transcription, but some heavy duty liner notes (locations, actors/people, genre, plot, etc.) either assembled by humans or by some yet-to-be-invented *super* recognition software.

As for context, I think that the key to opening up the long tail for audio and video (short of gigantic advances in the availability of meta data)is to have systems that can infer, for each individual work, what the right context is. For audio this can be done by looking at sales, or by looking at actual user collections ( we just launched a site yesterday that does the latter - ) to find the human context. Finding something new and interesting in a familiar context will help long tail audio artists to get found, but granted, that will only go so far for video as most folks don't store videos in an accessible place like they do for songs.

The comments to this entry are closed.