Friday, 5th October 2007
Data Mining, Search, and Summarization: Music, Critics, & Crowds
We've been posting a lot lately about a category of search that we call search+data mining (value added services). Search tools in this category include:
Airfare pricing and prediction
Historical consumer product pricing info
Aggregating/summarizing disparate consumer reviews
Aggregating disparate consumer reviews and giving them a score based on Wize's algorthim.
Today, Jon Fine offers an excellent introduction to a new member of our datamining/search category: Critical Metrics. CM offers music recommendations.
Jon Fine writes:
Rather than harnessing the collective preferences of an audience, Critical Metrics aggregates current and past critical opinion for, as of this writing, more than 23,000 songs on its beta site, play.criticalmetrics.com....Critical Metrics scans opinions from mainstream pubs like Rolling Stone down to single proprietor music-geek blogs like fluxblog. Such catholic sourcing, at least in theory, aggregates critical consensus and cancels out individual prejudices. Rolling Stone may worship the new Springsteen album, but some Critical Metrics-scanned blogs will find it emetic.
Again, we see the trend of mining/aggregating and delivering a summary of what's found from disparate sources. In this case, sources that have been hand-picked by thee Critical Metrics team.
Services like MovieLens from the University of Minnesota have been offering user-powered recommender system for movies has been online for years as part of the GroupLens project.
Look for more tools like this in the future. Why? So much information, so little time. Of course, sites like Newsblaster* (from Columbia University) and NewsinEssence** (from the University of Michigan) were auto-summarizing news from disparate news sources years ago.
Of course, another service that involves music recommendation is a ResourceShelf favorite and one we use daily. Pandora from the Music Genome Project. Not only is Pandora fun to use and a great way to find/identify/listen to new music but the service also illustrates the power of quality metadata, aka quality cataloging.
Pandora employs professional musicians/music experts to listen to CD's and
catalog the track using more than 400 criteria. When your using Pandora you can click, see the criteria, and how it relates to other songs you've said you like or dislike. It's truly fascinating and yes, even fun.
Here's how Pandora explains the process:
...fifty musician-analysts has been listening to music, one song at a time, studying and collecting literally hundreds of musical details on every song. It takes 20-30 minutes per song to capture all of the little details that give each recording its magical sound - melody, harmony, instrumentation, rhythm, vocals, lyrics ... and more - close to 400 attributes!
So, there you have it. Both auto-summarization/recommendation services as well as professional cataloging of music.
We will continue to track new tools that offer these and similar types of services.
* The NewsBlaster home page from December 12, 2005
** The NewsinEssence home page from January 26, 2002.