http://www.seobythesea.com/?p=5185
interesting ideas on how search engines might return better video results in the future. At present I think they're using some combination of the image search style proximity (including to embedded) and tagging/transcripts plus popularity, but there's interesting stuff in there whch might help bring out the more long-tail stuff.
QuoteA similarity search between images and audio from a video, and a labeled training dataset, which contains stock images and audio clips which have meta data associated with them can help to identify the unlabeled images and audio from the video.
hmm. so include some clear similarities to some items which already rank well for your chosen keywords?
From what I understand, voice recognition will be playing an even greater part in all this too. Lots of companies have been really working on this in particular but I haven't heard any updates in a while.
youtube automated transcription used to be really variable.... might have come on a bit since then though....
On that story in the Mahalo slashes 10% of staff Jason Calacanis in his email to staff points mentions:
QuoteInterestingly, while the search side of Google has impacted us negatively, Google's video unit (YouTube) continues to be our strong partner, encouraging us to ramp up our video production and publish even more of our expert video lessons with them.
Read more: http://www.centernetworks.com/mahalo-reduces-headcount-google-algo-change#ixzz1FR6ftK9i