But researchers at the company's laboratories on Thursday showed three separate approaches to searching and categorizing music files and e-mail messages, which Sun hopes will translate to improved scalable server cluster technology.
At the forefront of the research is Paul Lamere, who is using combinations of existing music database technologies to catalog the spectral signatures of music. The result is Search Inside The Music, a 3-D visualization program that can be used to explore a large music collection. You like jazz? Search Inside The Music reads the waveforms of an MP3 file and aligns it with other ones of similar values. The goal is to help the user find songs that are similar in nature to each other and make recommendations based on preferences.
Add in filters for social tagging and the result is Sun's Snapp Radio project. The mashup also adds in tagged pictures from Flickr to give the playlist a relevant look as well as sound. Doug Eck, a member of the Sun research team who specializes in machine learning and classification, suggests that the research is appealing on a couple of levels.
"By combining the technologies behind Radio Paradise and last.fm, we can start to translate the results into a better understanding of how humans can take suggestions from a machine but add in their own preferences," he told InformationWeek. "It's not a perfect system, but it does accommodate for social suggestions that a song might be 'great for a workout' or 'perfect for romance.'"
In an enterprise scenario, the music search and suggestion functions could also be augmented by Sun's MailFinder program. The research led by Jeffery Alexander and Stephen Green has produced an interface that scans more than 3.5 million e-mail messages that were sent to public aliases inside Sun's mail servers.
The research is being translated into different ways to cluster around the author of the message and how to apply it to mail archives.
Sun Labs director Bob Sproull suggested that combinations of all three projects could speed up search relevance for Sun's own Web site, including finding answers to questions, updates to Sun's bug databases, and suggestions for product demonstrations based on a user's traffic patterns.