Sunday, 15 February 2009

Powerset

I like it. It's got potential.

As described on ReadWriteWeb:

Powerset is a much-hyped semantic search engine that uses natural language processing to "understand" concepts in web content and match pages to queries. Right now it only searches Wikipedia (and Freebase). 

www.powerset.com

The string 'Semantic search' always makes my ears prick up. I had a play with powerset, and it is not a bad start in the whole Natural Language Processing search experience. (Incidentally, I am so sad that NLP, which is one of my Favourite Things, shares an acronym with corrupt, pseudo-scientific, fraudulent, thieving, and cultish self-help psycholbabble. Oh well.)

Powerset has a smashing feature for information discovery: Factz. A search for 'Gorilla' (NB: 'Gorillas' takes you to the band 'Gorillaz') gives us this: 

1. A short overview ('Gorillas are the largest of the living primates'...)
2. Scientific name, Taxon rank, etc. 
3. Factz. (This is the fun bit.)

The factz, each of which are linked to info tell us that...

Gorilla uses: tools, branches, lick, rocks, stick, saplings, and trunk
Gorilla eats: plants, fruits, shoots, bamboo, herbs, and leaves
Gorilla live in: forests, groups, corners, troops, and park.

Not bad. (And if you're wondering - Gorillas lick objects to identify them. Who knew? I didn't, hence me being impressed by the discoverability potential - you can learn by accident, which is my favourite way.)

It all gets a bit stickier with more complex topics - like people. A search for Marlon Brando opens with his bio, provides his Date of Birth, Date of Death, Place of Birth, Nationality... very impressive. But then we go to factz:

Marlon Brando played: Jor-El, role, owner, survivor, movie, Corleone, drums, characters,
Marlon Brando said something about: Kashfi, Tango, Paris, Brando, Last, Bertolucci, call, Marilyn Monroe
Marlon Brando portrayed: father, leader, Don Vito Corleone, Johnney Strabler, Jor-El 

So... what's the semantic difference between 'portrayed' and 'played'? And how did 'role' get through the net? (Answer: the sentence '...in the role eventually played by Marlon Brando'.)

But that's a quibble. I think the problem can be solve in the first instance by not paying heed to obviously off-target factz (like 'role'). And I love 'said something about' - it's a little misrepresentative because it is, by definition, limited to available quotes, but still - how nifty!

I hope this one takes off, it's terrific. If it does fly, the big danger is that it may be so good that people take the results as being comprehensive - let's all promise not to do that, shall we?

No comments:

Post a Comment