tim: i read an article about aimee mann. the audience was not quite big enough for her to warrant the attentioof a producer. she was kind of in a no man's land. all these hundreds of musicians i'd known, all living in obscurity, there was all this talent nobody knew about. one thing i had learned to do was understand someone's taste. i did that by essentially putting them for a musical interview where i would play a song for them. based on their reaction to what i played for them, i would kind of home in on their taste or what they wanted for their movie. i thought in that moment, what if i could codify that and then build a discovery tool? emily: so that was the seed that became the music genome project which is sort of a part human, part algorithm way to figure out what people want to listen to. tell us how it works. tim: we built a 450 attribute musical taxonomy, which is essentially every dimension of melody and harmony, kind of like musical dna. we have had trained musicians analyze songs one at a time and score them at