February 6, 2007

The Future of Online Search for Multimedia Dance Content Part I of II

In this post you'll find highlights from the latest developments in Internet search technology.

This is a two-part post. In the next post, I'll describe how the search tools discussed below, with additions and modifications, might be put to work to provide dancers, choreographers, educators, students, researchers and others with new ways to search for video and other forms of multimedia content.

Searching for dance videos or specific parts of dance videos may not strike you as a pressing issue for the dance world. But it will soon become one as more and more dance videos are uploaded to the web.

How I Find Dance Videos Today

For my recent multimedia guides to Katherine Dunham, Mark Morris and Merce Cunningham, you might be wondering how I find online video clips for each of these dancers.

I start by using three video-dedicated search engines that index videos across multiple video sharing sites:

- Vdoogle.com
- SearchVideo
- Blinkx

When using the above search engines, I usually start by typing in a line of text such as the following with quotation marks included:

"Katherine Dunham" dance

Next I turn to individual video sites including YouTube, Google Video, Yahoo Video and others - although for the most part I'm probably going to get the same results as I did with the 3 videos search engines above.

Then, finally, I go to the regular search engines such as Google and search again for videos by entering text such as:

"Katherine Dunham" dance video

The reason I go to the general-purpose search engines is because none of the dedicated video search engines is comprehensive so I have to use a number of additional tools to conduct a search.

I also came across an upcoming video search site called CastTV. TechCrunch has good things to say about CastTV's search technology, but as you can see, it hasn't lunched yet.

Image Search

When we think about search, we're really thinking about text-based searches. We type words into a search box to find matching results. But why limit ourselves to using text strings when what we're looking for is multimedia content?

Take a look at the Like Visual Shopping site. You can conduct searches where the input consists of images and not text. So if you click on the picture of a red shoe, you'll be presented with matches that have similar color, size, shape and design of the red shoe you started with. This same visual matching technology can also be used for matching people - more about that when I talk about new search applications for dance.

Here are two screen shoots that show how Like Visual Shopping Works. In the first picture - click on it for larger image - I clicked on the image of the red shoe. Like looked for other items in the database with similar visual characteristics and returned the results, which you'll see on the second image.

Pixsta also offers a visual browse and search technology. Visit Net-a-porter.com to see how this matching technology is used to search for clothing and accessories. [Via Read/WriteWeb]

One advantage with this visual matching approach is that you don't have to describe with words what you're looking for. You can simply start with a picture of what you like and these visual matching tools will find similar items.

Visual search is also available for mobile devices. With Mobot mobile video search, you can use your camera phone to take pictures of movie posters, magazines, billboards, logos and other items, and the application will return relevant purchasing information to your mobile device.

Audio Search

Here are two applications with different takes on searching for audio content:

Podzinger searches for audio content within audio and video files. Podzinger uses speech-to-text recognition technology to convert the spoken word to text files. Then you can use the search engine on Podzinger to search these text transcripts.

I used the Podzinger search engine to find references to "modern dance". The first match is for a radio program where the term "modern dance" is spoken at 13:18 minutes into the program. I can click on this time stamp and I'll be taken to a point in the audio program a few seconds before "modern dance" is uttered.

With Midomi you can search for songs by singing or humming a few bars into a microphone. This voice technology search tool will then find songs that match your input. Here's a video on Midomi on how to use this voice matching tool. The Mashable blog is sold on this technology.

Videos Authentication Tools

What I was hoping to find on the cutting-edge of video search tools is the ability to search for moving bodies within a video. I didn't find this although it may exist.

What I did find are automated video content authentication tools that allow content owners to ensure that their videos are not being used and distributed illegally. So this technology can be used by movie studios who don't want their copyrighted materials to be uploaded to YouTube for example.

This video technology is usually referred to a video fingerprint filtering. With this software a large number of video files can be examined to see if any of them contain a specific "fingerprint" that matches the videos of a content owner. I haven't been able to figure out yet how each video is actually examined - is it by color, movement, shapes or other elements?

If specific objects - such as a human body - cannot be identified and tracked using this fingerprint technology, then it really won't address the video search challenge for dance that I'll address in my second post about this topic.

Here are three companies that offer this video filtering technology:

- Advestigo just introduced Advestigate, a filtering technology to analyze user generated content (UGC). [Via SearchEngineWatch]

- Guba, which offers downloadable video content for free or for a fee, uses a proprietary software to filter video so that all content is being legally used. [Via TechCrunch].

- Audible Magic offers tools for fingerprinting and protecting audio and video copyrights from infringement. Here's November press release for its new video identification service.

Trackback Pings

1 Comments