contentious.com

Here’s some good news for die-hard National Public Radio (NPR) fans like me.

According to this CNet article, “Search engines try to find their sound,” NPR recently began using StreamSage technology to transcribe audio content as it is broadcast. This enables search engines like Google to index NPR’s content more quickly.

This process is not 100% accurate, of course, but apparently it works well enough for the search engines. Reliable voice transcription has long been an elusive holy grail of the content world…

The CNet article notes:

“NPR’s strategy is working so far: In recent weeks, NPR audio has begun regularly appearing on the index pages of Google News and Yahoo News, and clips also crop up when people search for news-related keywords, such as “Abu Ghraib,” the name of the notorious prison in Iraq. …Since stories started showing up in results churned out by major search engines, the NPR site has seen record spikes in visitors for high-interest news stories, such as the murder of American hostage Nick Berg in Iraq.”

The thing is, Streamsage appears to work well but it looks like they’re marketing it mainly to larger organizations. I guess that means it’s probably beyond the reach of small content providers. I hope that eventually someone develops a good system like this that’s affordable and accessible to the consumer market – especially now that audioblogging seems to be on the rise.

Of course, the transcription approach focuses on adapting audio content to cater to the limitations of text-based search engines. This gets me wondering whether we’ll ever have true audio- or video-search capability. That may be pie in the sky for now, but in the meantime audio/video search services such as Singingfish are starting to achieve some intriguing results. In particular, this caught my attention on the Singingfish site:

“Our highly automated systems don’t simply locate streams; they extract the internal metadata, and then add more metadata for indexing. By the time a stream is indexed, we often know more about it than anyone else-possibly even the people who created it!”

And that, I think, is the key for now. Until we have computers that can truly process, interpret, and search audio and video, rich metadata will become increasingly important for all sorts of non-text content.

What’s “metadata?” Generally, this term means “data that describes other data.” Here are several more detailed definitions. Robust and consistent metadata is crucial for any system that retrieves and manipulates any kind of content – from articles, to fingerprints, to spreadsheets, to security videos, and much more.

Since I’ve been learning extensible markup language (XML – a flexible way to create standard information formats and share both the format and the data online), I’ve been delving ever deeper into the world of metadata. Frankly, I have a major hunch that metadata is the next frontier for the editorially inclined. More on that tangent in future entries…

In any event, next time an NPR audio feature turns up in one of your Google or Yahoo searches, think about what this means in terms of how the Internet continues to evolve to accommodate the way humans perceive information, rather than just how computers prefer to process it.

Neat stuff, eh?

(NOTE: Thanks to John Battelle’s Searchblog for bringing this topic to my attention.)