Miker

17th level Hacker

Search SIG

The first meeting of the Search SIG from SDForum was last night. The new forum is run by Dave McClure and Jeff Clavier. They did a great job on this first event, and I’m really looking forward to our combined Search SIG and Mobile Monday event happening next month. I only seen one posting so far. After Niall and Kevin vying for attention during the announcements, I figured they would represent with a summary or something. Here’s mine.

The main section of the presentation was a panel lead by Doug Kaye from IT Conversations. Doug started out pretty early drawing a distinction between different kinds of audio search, with the main distinction being between directories of audio and search services that actually look at the audio and try to index it. Podscope and Blinkx were mentioned in particular. Podscope is used frequently by companies looking for mentions of their name or products. There was one other mentioned, but I must have gotten the name wrong. If someone knows it please leave me a comment, I’ll update the post. Jeff from Yahoo drew a distinction between phonetic search and text recovery search. Making a transcript from audio gives you text that you can search, but frequently gives garbage back if the audio quality is poor or there’s a lot of noise. Phonetic search tends to give better results, but doesn’t give you text to work with. He also pointed out that closed captioning is another fantastic source of info if you’re dealing with media where that’s available. Someone joked that closed captioning tends to be highly innacurate, but apparently there’s a mandate (FCC?) that requires some minimal degree of accuracy (97%?) for closed captioning on TV broadcasts.

Eric Rice from Audioblog.com (recently funded) made that point that most podcasts appear with some kind of text, and getting that text indexed is the key to getting your content found. He said he personally hasn’t had trouble getting his content found. He likes tagging, even though it has a way to go to be a mature technique. In his opinion extracting text from audio isn’t a great technique, you lose lots of nuance like sarcasm. When asked about directories he said that as a publisher he doesn’t like that there are so many directories to deal with. ITunes is really the big kid on the block when you think about directories. The inclusion of podcasts in the ITunes setup doubled traffic for lots of folks. Eric also gets people coming into his site from searches that hit his show notes. He also tends to tag his own stuff on del.icio.us to bootstrap it.

David Marks from Loomia spoke about the different kinds of things that are searched for. Enterprises search for mentions of their name, but people tend to search for things like music. Conversion to text just isn’t going to help at all when it comes to something like finding music. Loomia attempts to do that through recommendation and collaborative filtering. It helps you to find the things you didn’t know to search for. Timeliness is very important with blogs, but it hasn’t hit yet with audio. It will hit with audio, and that should be very interesting. Audio search is very much unlike traditional search because the cost of scanning results is much higher. It’s not like a page you can scan with excerpts randomly clicking through to see which is the right one. Starting up an audio stream and checking to see if it fits your needs is a much more cumbersome activity.

There was some talk about Media RSS, extensions to RSS spearheaded by Yahoo. Media RSS aims to provide additional information about rich media to be transported along with RSS feeds. It’s meant to help democratize the creation of rich media, but currently people have to be pretty techy to use it. Eric said there was really a need for vendors to create tools that emit this kind of extra information with a single click of a button or toggle of a checkbox. It’s what they did with Audioblog for the ITunes extensions. There’s a very simple checkbox that allows people to include the extended attributes in the information that Audioblog emits for them. Normal people can’t and don’t want to edit RSS. Dave from Loomia said that as an aggregator you really have to eat garbage and poo gold. You have to take in all kinds of errors that the user makes in putting up their info, and provide perfectly correct information to your consumers. It would be great if people validated their feeds, but they don’t. And it would be dumb to expect them to. It would be great though. Character set problems are very common. Doug said he’s run into a lot of character set issues as well.

There was some discussion about the ITunes release that doubled traffic. It happened right around the beginning of July. Apple didn’t tell anyone what was going on though. They sucked in the wrong feeds, and none of the podcasters knew about the ITunes extensions cause they weren’t public at that point. As a result the podcast info in the ITunes directory looked like crap. But most people forgave it cause it doubled their traffic.

Ev (from Odeo) made a point about the key to getting found having nothing to do with technology, it’s all about being unique and having interesting things to say. If you do that people will find their way to what you post. He also said that Odeo isn’t really looking at the creation of text to go along with the audio they do. Most current podcasters think of their website/blog as the home of their audio, but Odeo really wants to go after the very non-technical users who don’t have a site to associate with. They just want audio.

Eric mentioned that mobile devices also don’t really encourage the use of text. Blogging behavior really doesn’t apply when most of the consumption of the audio is being done on iPods and other audio players. People aren’t going to link or trackback from their iPod. Eric said that everything that Audioblog does for audio, they do for video as well. He’s a big believer in video. Ev said that he wasn’t as enthusiastic about it however, cause the consumption devices aren’t as widespread. There are portable audio players all over the place, but no so for video. Doug also made the point that audio already dominates drive time and exercise time, there’s an existing consumption model for audio.

Then there was a quick break and there were demos from Loomia, Odeo, and Audioblog. Great demos actually, I really liked this part of the session. I liked the part about the consumption model for audio vs. video. What podcasting allows people to do is pour more interesting content into the time when they’re driving or exercising. Definitely makes a lot of sense. But the same thing is true for video. There is a built in consumption model there. Lots of people sit around in the evening and channel or tivo surf for an hour or two after dinner or before bed. Would the same thing apply if we were able to pour content from online into that space. Of course, it would have to come in to something that would allow it to be displayed in the same way, it would have to appear on your television and not your computer screen. That might not be true in Asia however, where usage of mobile devices is already pretty well positioned for something like video delivery. There has been some effort already to provide this kind of service, I think specificly about both Akimbo and some efforts by Tivo when I say this, but I don’t think anyone has really crossed the line into videocasting like this. Can I subscribe to a Media RSS feed if I get an Akimbo box?

I think someone like MobiTV would be pretty well served by thinking about stuff like this. Mobile video makes sense, but only if there are clips and snippets of the right size worth watching. Clips and snippets of the right size are generally just about movie trailer or commercial length (I know some folks really think the ideal video length for mobile consumption is 4 to 5 minutes, but I disagree). People generally aren’t going to pay for those little videos however. Do I want to pay $3 to $10 to watch a few minutes of video on my phone? Hell no. So you need a different way to make money off that content. How about putting commercials in? No, when the slices of time you have to fill are short already no one wants to wait for commercials to go along with their content. So how about making the whole video clip a commercial? “What the hell are you talking about?” you might say. How about if the video clips that MobiTV carried were basically just coming attractions and short versions of longer content? If I download a clip via MobiTV, for free, and I like what I just saw I can opt to use my IP DRV at home to download the main content (a new movie, a television show, whatever). Television already runs cross promotion on television in order to get folks to change the channel to whatever it is they’re pimping. How about paying to pimp it on mobiles as well. If the content were done right I think people might actually enjoy the bits they get on their mobile, and because they get it for free they’ll be a lot happier with what they do get. It would basicly be using your mobile as a video remote control for your DVR at home. You can whip it out when you have a few minutes free and check out recomendations and new additions instead of doing that part with the remote control while sitting on the couch. When you select stuff you want it activates the box at home that starts downloading. And by the time you park your ass in front of the tube you have a hot steaming dose of fresh video cache to stimulate and amuse and rock you to sleep. Sounds like a good idea to me at least.

Tags: searchsig