We all take for granted the proliferation of text based search on the Internet with sites like Google and Yahoo. But think about it: What is the number one medium that people flock to? It is video and audio. Think of television and radio – but for the Internet.

I discovered a Seattle-based company who could be the next Google of searchable online audio and video content. I’m not alone: SingingFish.com has just accomplished a significant with over 1 million online audio and video searches per day.

Karen Howe is CEO of Singingfish.com, the largest online audio and video search engine. They got a leg up by working out search integration deals with Microsoft’s Windows Media 9 series player and Real Networks RealOne Player for exclusive media search technology in those players. Those two companies also integrated Singingfish.com audio and video search into the RealGuide and WindowsMedia.com.

Howe recently talked to me about finding their niche with this potentially huge online trend.

Q: Tell us about SingingFish.com and how it works.

Howe: As a search engine, we look specifically for audio and video files on the Internet. A lot of folks are quite familiar with search engines, but they’re thinking about documents and text like Google.com. Think of us as the Google of audio and video.

There are 3,700 specialty search engines in the market today. We are one of the very few that specialize just an audio video search. We spider the Internet and look for rich pockets where we’re going to find great audio and video content. We collect that information, index it, and present it to end-users so they can listen to music or watch a movie trailer.

Q: What kind of categories are listed in your search engine?

Howe: Categorization is really important for search engines because it helps narrow your focus. We’ve indexed 35 million streams. When you’ve got that kind of inventory, you have to get clever about organizing that information so people can do something with it. Music, movies, TV, sports, radio, finance – we categorize it into a top-level type of format. For example, if you are doing a search on Madonna, but you only wanted to hear her radio interviews, or – heaven help you – only wanted to look at her movie trailers, you could separate it out like that so you’re not looking for MP3’s or music videos.

Q: SingingFish.com recently hit a milestone.

Howe: Yes, one million queries per day! It is a terrific milestone and is something that Thompson, our parent company, looks at it as an achievement. What’s nice is that’s just where you start from – and then you keep going. As we gain distribution on different sites and attract more people to audio/video search, those numbers just continue to go up. They don’t go down. Once you start up, you stay on a steady climb.

Q: Give us your unique perspective of what people are searching for. What are the popular trends?

Howe: You could probably say about 40 percent of all queries on any search engine is entertainment-related. The folks at Google gave this percentage to me. It just shows how much people are interested in music, sports and movies. We believe people would be better served if they could do a search and be able to see both text and audio/video results on the same page. Music is No. 1 for us: 60 percent of all our queries are music related. You would expect that since we specialize in audio and video. Right after that are movies, then sports – with news falling in after that.

Q: Do you find music being accessed more than video?

Howe: I would say audio is searched for more than video. It does relate to broadband in the home. We are only seeing 20 percent penetration of broadband right now. But I have to say – at work, those are the same numbers! At work, a lot of people have T1 lines. We find there is a tremendous amount of usage of audio and video streaming during the workday.

Q: How fast or your growing?

Howe: Typically, we work with a player like the RealOne Player. Once we brought them on as a customer, we saw a huge spike in traffic. Once we brought on Microsoft with the Windows Media Player 9 and the WindowsMedia.com web site, we saw another huge spike in traffic. We have other smaller sites such as Sympatico-Lycos, which is the largest broadband portal out of Canada and freshEYE.com, which is a search engine in Japan.

We never tried to turn SingingFish.com into a destination site. Our business model has always been to license the content into other providers of content, such as portals, players or ISP’s.

Q: What do you offer these media players?

Howe: Our customers have the option of how they want to deploy search within their product. For example, within Windows Media you’re going to see the Windows Media format, whereas on Real.com, they tend to show all formats. Microsoft put in categorization, but RealNetworks didn’t.

Q: How is streaming media content indexed?

Howe: We build this engine from the ground up specifically for streaming media, which is an entirely different technology than looking for a text search. Streams are a very different beast. You have to peek inside them, extract header information, and annotate the metadata that’s there. Typically in a text-based search you’ve got all these wonderful words all over the page.

That makes life easier when you’re trying to figure out what this page is about.
We look at copyright, the title, author, artist, duration, bit rates as well as a description. Typically we find that many people encoding right now are not very sophisticated in the way that they create their titles. It’s almost as if – “find the information if you can, but I’m not going to make it easy for you.” Or they will title different streams with the same title or with something that’s nonsensical.

Q: When indexing, does SingingFish display “descriptions” or “titles”?

Howe: We do have a “descriptions” field that we display as part of the database. As a matter of fact, some universities are very good at giving detailed explanations about what that stream is about. In fact, some of the descriptions may be longer than the stream itself. It’s more typical, however, to have the reverse be the case where there is limited or missing information.
We also license third party databases such as Muze for music. When the stream is missing, say, copyright or genre information, we are able to stuff that back in through an automated process. We’re very sophisticated on the database side. In fact we have the largest Oracle license out of our entire corporate structure of Thompson, which is the fourth largest consumer electronics company in the world.

Q: When I encode, I typically put the same information in both the title and description field. To be properly indexed, your advice is to put more information in the description?

Howe: Maybe even slightly different information if you’re trying to get a keyword search hit. We will then be able to pull out ranking and relevancy based on what the query is.

Q: Which is best: complete sentences or keywords?

Howe: Sentences are good because we can search plus (+) words as well.