If you try to look for a clip you’ve seen before, and are not sure how it’s titled or something like that, it’s pretty hard.
The problem can probably be put down to the fact that users, registered or not, other than the video publisher, aren’t able to label the video with metadata. You can leave comments, but that’s not really the ideal context for labeling video with metadata.
The lack of searchable text which comes with video is the main reason it’s hard to search through, so metadata needs to be pimped to ease searching functions.
Flickr does this brilliantly, but to be fair I don’t usually search for specific images on Flickr at all. When doing a search similar to a flickr search, usually an general ‘what do you have on this’ search, YouTube does really well too.
But with specific video YouTube sort of fails a little, and I reckon it’s because of the lack of metadata.
(As a disclaimer - I’m not intimately familiar with YouTube but I can’t quite see a way of doing this.)
I’m looking forward to seeing what Google might have in store. Video is such a cool ‘new frontier’ of search - the algorithms for video can be so interesting and produce such cool results… but the ones required to actually search using a truly dynamic and expanding library of things are a long way off.
Still, watch this space…
I know, it doesn’t quite fit the Web 2.0 tagging ideal.
So how does Flickr stack up when searching for a particular image? Does the metadata actually help you?
What about in the second, further case, where there are 10,000 images with the same metadata search criteria?
Flickr does rather brilliantly, actually. I search for an image often on Flickr (I know what subject I’m after, I just want to find a good picture) and it usually performs quite well.
When you think about what Flickr have done to get these results, it comes down to three things.
1) They’ve provided the methods of entry for the data. Tagging, titling, describing, noting, geotagging, favouriting and commenting.
2) They’ve encouraged their users to contribute metadata either by direct request (less frequent) or by making it quite user-friendly. At least, by providing the tools to enter the data you are encouraging the user to do so.
3) They’ve embraced their data in a variety of interesting ways - the Flickr search tool is actually very good thanks to the different data they have on the images. Geographic searching builds on this and is totally excellent.
Having so much metadata gives you great options when it comes to searching through your database of content.
Flickr works well because there aren’t 10,000 images with the same metadata, except for the ones that are private or under the strict anticommenting/tagging/etc. control of the author.
At least, Flickr has way more and varied data on their content than YouTube, and so won’t face that sort of problem often.
SO by allowing users to enter the data, you help search. If you don’t, you’re only going to hinder your search engine. And if you take that path when you’ve got something like video, which can’t really be (well) indexed at the moment, you’re going to have lousy search.
hmmm
“Flickr does this brilliantly, but to be fair I don’t usually search for specific images on Flickr at all.”
“I search for an image often on Flickr”
see, searching for a image of a particular kind is different from searching for a particular image. Of the seven metadata-adding techniques you mention for Flickr, YouTube already uses five of them, plus the additional “rating” field and a category (as well as the groups, which Flickr also supports of course).
“Flickr works well because there aren’t 10,000 images with the same metadata”
Wrong. Search for “me” or “home” or “dog”.
I think you’re actually looking at this the wrong way around. I think the metadata that YouTube supports is more than adequate (although not perfect), but the search engine doesn’t expose any of it and doesn’t allow you to search by a particular metadata field. If it did, I can well believe that the amount and quality of metadata added to videos would rise.