Quote:
Originally Posted by ApK
Conceptually, I think you have it exactly backwards.
It's far easier to create an expert system to analyse word use, language, genre elements, grammatical structure, etc, in text then it is to measure the elements of music that Panadora does. I mean, technically, I could write a thesaurus comparison function for text, I wouldn't know where to start making code figure out if a song was "guitar driven" or "family friendly."
Pandora does a pretty good job with it's choices, but really so does Amazon's recommendation system, and it doesn't use any of those sorts of data, which I think could only make it better if it did.
|
With music, the tone or feel is generally represented by how upbeat things are. With text, most of the tone comes from inferences and context. If you did something like assign values to words and then calculated based on how often certain types of words were used, you could get a general feeling, but it wouldn't always be accurate. Was that really a sad story, or a comedy about goths?