Early days for video web-search

Today I wanted to watch the speech Hugo Chavez' UN speech. The task proved harder than I expected:

I started by entering "Chavez UN speech" in Google. I found newspaper reports and commentary, but nothing that obviously gave me access to the footage itself.

Fair enough. Google is a general-purpose search service, and it can't know I'm looking for video. So I tried "Chavez UN speech video". This was better: the fist item linked to a blog post with commentary and an embedded YouTube player with parts of the speech taken from Fox News, dubbed in English. I didn't know if this was the full version.

I then tried Google Video: I got three items: a "dramatization", an AP newscast with 20 seconds from the speech, and something from a Long Beach City Council meeting

So I tried YouTube. Again, all I found was news reports with only excerpts from the speech and commentary, all taken from TV newscasts.

Then I tried Blinkx, which is known for aggregating content from many sources. Again, however, all I found was news reports with only parts of the speech.

Finally, I went to the old trusty Alta Vista, which I had nearly forgotten about. I chose the Video tab and typed "Chavez UN speech". The first link pointed to a page at a nonprofit organization's site which, in turn, finally, linked to the full speech, without commentary (in the Spanish original, and also in English), hosted at webcast.un.org.

 

Why was this so hard? Why wasn't I offered the genuine article when I first tried? Going through this quest made me think of various issues:

• Neither Google Video and YouTube are web-search services in the traditional sense. They are hosting services and destinations, and they probably hope that the sheer amount of content they host will make the difference meaningless. This may make business sense, but it makes me wonder if Google is letting an opportunity pass them by.
• Blinkx is closer to being a web-search service, but its apparent reliance on a limited number of sites (generally broadcasters) means that they can't leverage the long tail of video.
• It's not obvious that offering a genuine web-search service for video is straightforward. Algorithms like PageRank would link to the videos that most people have in mind when writing about Chavez' speech, and these are likely to be news packages on YouTube.
• Even if a service were to overcome these technical difficulties (as AltaVista apparently does), it's not obvious what it should link to. If the "genuine" video is embedded on the "originating" site, presumably the link should be to the page that mentions it - but there's no guarantee that the video will feature prominently on this page (as in YouTube), or that it will be the only one on it. On the other hand, embedding someone else's video in a search-results page might be contentious without publishers' consent(see this); and if search services have to seek consent from all publishers then they can't offer the long tail of content.
• The people at AltaVista are aware of this. From their site:

  The video image you see on AltaVista results is a scene from the video itself. This image will rarely appear on the "found on" page. Here are some tips for finding the video:

  * Look for text that might link to the video.
  * Look for another image that might link to video. The image might be another scene from the video.

• And then, what does it mean to speak of the "genuine" video or the "originating" site? Linking to a broadcaster's site (as Blinkx does) is, in a way, a link to an "original", hosted at an "originating" site.
• My doubts about which versions were unedited point to tricky questions of attribution and integrity of content.
• Finally, it wouldn't be practical to always link to the "official" source: the UN's servers would probably crash if millions of people were to stream video from them (but then, this is hardly prime-time stuff)

Lots of questions, no answers.

 

: Update 06/12/06: Embarrasingly, when I wrote this I hadn't looked at Yahoo Video Search, which - like Altavista - delivers exactly what I was looking for (see also this Forrester report for a good overview of the competitors).

Comments

[Bertram]

I think the problem is that it's the HTML-Code that's indexed. There should be a microformat for video containing all the required informations about the video, the creator, the "originating" site and so on. Then you could access the video with its associated informations without the need to visit the "originating" site.

[Nico Flores]

Bertram - Are you sure of that? Google Images seems to work quite well without any microformats, fancy metadata, etc. I really don't think this is a technical problem - it's one of business models (Google prefer to host video), etiquette (is it OK to embed someone else's video without permission?) and legislation (are there any legal precedents?). Metadata is usually oversold, imho.

[Bertram]

Well I agree, injecting Metadata is quite a hassle and I'd love a better solution. But for the time being it could solve some problems. Why is Googel Images working so well? Because there is a convention for embedding images and you can see the full result in advance . I wager you could guess the images of any given news report just by reading the caption.
Speaking of business models you could create a new one by microjunking video and freeing it from html. My problem with video is usually where to link. If I link to the site containing the video there's too much clutter around and if I link directly to the video-file no context is given and nobody will watch the video.
So just find a way in the middle and that's what I meant with microformat: a way to link to and display a video with just the associated relevant information (by the way you could include licence information solving the etiquette and legal problem). With time I hope one could convince the YouTube and GoogleVideos to adopt.

[Timo Stein]

In addition to your post http://ondemandmedia.typepad.com/odm/2006/10/how_youtube_wor.html, you elaborate a bit on the tedious video search trial everyone has to suffer in order to just find a certain video. I think http://blinkx.com is close, however - I don't like their interface.
For me, new video search engines like http://www.lumerias.com are a better fit, as they combine the simple google search experience with a focused video search. But lumerias at the moment seems to be building its index and has not 100% coverage yet...