Sunday, June 28, 2009

hi ho hi ho, it's off to the mine we go




Data mining seems such a naughty thing, but most tech is neither good nor bad; its all in how it is used.

Take Marconius' wonderful index of all things RR, which in a sense mines data from several sources into one great resource for us RR folk. That's pretty much a mechanical labor of love; no one's going to write a program to automate for such a small group.

But the folks at LyricMiner have apparently been thinking along the same lines as Marconius, and think they have a way to tap into the growing mixtape/playlist trend.

Their almost-out utility will be free at least at first, and runs on Mac, Linux or Windows. What it does is scan your music collection, checks it against 3 large on-line lyric databases, and creates a searchable keyword file. Because it is cataloguing only words, it is a realtively small text fie. It takes about an hour to catalogue 30,000 song files, and after that its miscroseconds. (There is a 'watch folder' option to automatically add new songs to the catalogue.

I'll use this week's RR topic to illustrate. Use 'pets' for instance, and it'll give you a list of songs in your collecion that mention pets. It also uses a synonym and alternate spellings database to suggest other likely choices, PLUS - and I think this is sheer brilliance - it has some algorithm thingee which lets its servers monitor reputable sites dedicated to song meaning discussions, and from that adds other songs from your collection likely to be about the keyword.

And the list is ranked by likely matches. Eerie? Just a bit? But it's pretty amazing. You can have a sneak preview here.

7 comments:

tincanman said...

I probably should get a life

saneshane said...

man.. I read that and thought I would really spoil my fun of discovering stuff... but it's pretty far out...
unfortunately it came up with 'straight girls don't lick pussy' from my porn house collection for pets.. still not pufffect then!
nice one tin.

Carole said...

Being a boring database technical expert type person IRL, I am more than slightly familair with the concepts behind data mining.

Nice to see them put to an interesting purpose for a change.

TatankaYotanka said...

Will there be redundancies here then?
I could invest my payout in a vinyl mine.

goneforeign said...

Well don't tell anyone, we're onto a goldmine here, with this we can lock up the weekly A&B lists from here to eternity!

ToffeeBoy said...

@ goneforeign - some of us still wouldn't have a look in!

DarceysDad said...

[*sound of utterly perplexed, and very over-tired DsD scratching slightly-sunburnt head*]