I’ve found the link to Codeorgan in my rss feed (via synthgear), and as its name almost suggests, it transforms any web page into music!! The application processes the < body > content of an HTML page to count the number of characters. Then several parameters are used to determine the key, the synth instruments, and rhythm patterns of the music bit.
Needless to say that I processed the homepage of the blog through codeorgan, and voilà Muzicosphere’s music interpreted by Codeorgan. I kind of like it! It may seems useless but I’d like to understand the underlying algorithm behind all this. The basic steps of the algorithm are:
- Scanning of the web page content.
- Filtering the characters: It keeps letters from A to G that are actual musical notes.
- Analysis of the number of characters: even numbers are he page is translated to the major pentatonic scale of that particular note, it becomes minor if there is an uneven number.
- Synth choice based upon the total number of characters.
- Drums pattern choice based upon the ratio of characters on the page versus the number of characters that are actual musical notes.
Though the algorithm is quite blurry to me, I kind of like such experiments because it could be related to all the projects in the field of sentiment analysis and text mining. Obviously, to reach such purposes the analysed text has to be subdivided into words rather than characters, and the filters have to be quite elaborated.
In any case, that is a tricky mind game that totally matches the sentiment analysis questions and outcomes (this field has an increasing popularity because of the web 2.0 circus). If such analyses had to be widely used and spread, I think I would rather hear music bits instead of reading single scores such as negative/positive.
Well, anyway that is over speculative to talk about this, especially if you consider that language is so flexible that it is hard to automatically determine the general sentiment associated with a text. But will that always be the case?