Everything in Its Place(s)

I have been exploring Pandora.com, an on-line, real-time music service that specializes in finding new artists for the listener based on the characteristics of their self-chosen music. Pandora has meta tagged the songs on over 30,000 cds with the categories of tone, vocal, electronic or acoustic, beat, melody, etc. Pandora employs music theorists to properly identify the characteristics of songs in the database. The service provides a “Why” for each song played, so the listener can understand the similarities of the music. I have found it very interested to see how the meta tags do pull similar music in type, but can totally miss on nuance. Bananarama and Britney Spears are NOT the same.

Another database I am looking into is Zappos.com. It is a web shoe store for men, women and children, with seemingly all available sizes. The shoes are tagged by designer, style, heel height, gender, size, width, price, and sale. There are many ways to identify a shoe.

These approaches on information management both involve identifying all of the different facets of a single item. Thinking of cancer.org – how will we make its content accessible in the new world of pull-down information driven by consumer demand? Who will do the best job at tagging the data – experts in the field, such as at Pandora, or constituents, who are in greatest need of the data? Can the two be combined? Do you just start, tagging information broadly first and then drill down as feedback is received and relationships are realized? How to allow for feedback? How to keep the data accessible as new tags arise? How to incorporate new data? How to do it cost effectively?

I am drafting an idea for a starting point for this expedition into the meta verse. I want to be able to learn what types of cancer I am at risk for as easily as I can find ten styles of brown two-inch heel loafers and songs that have female vocals in a major tone, electronica influences, and repetitive melodic phrasing.


  1. Yeah, the world has a huge and growing info management problem. But I can’t get excited about tagging. I’m too lazy. After writing something you have to think up labels for it. I was never good about that back when we did file folders. It’s me doing more work rather than the computer. I can’t believe that we still don’t have software that can figure out what 1000 words of text really means. Why can’t they supply the labels? That’s why I’m skeptical about AI. It’s still people supplying all the smarts.

  2. Tagging stuff in the right way will be major. I can envision marketing MBA’s in tagging.