Testing Lexxe

Emily Chang in her eHub blog posted a link to the new search engine called Lexxe. The unusual thing about this engine is that it claims to be able to answer questions written in a natural language. To be more precise, it searches the web and extracts the answers from the pages it found.

I gave it a shot, and it kinda works – sometimes. More often, however, it produces answers that are far from being precise.

Q: Who was Lenin?
A: leader of the bolsheviks a violent
(I don’t know why it picked such a strange fragment, but this is true).

Q: What was the real name of Mark Twain?
A: Samuel Clemens
(Bingo!)

But the same question about O’Henry caused a problem – the answer was marked as “not reliable”, and contained some words related to the “Seinfeld” show.

Q: Who discovered America?
A: China
(What?!)

Q: When was America discovered?
A:1492
(Ok…)

Q: What is the size of proton?
A: 1 Million
(1 Million of what? Hmmm… Let me try again…)

Q: What is the size of a proton?
A: 10-18 Meters One
(Too scientific for me…)

Q:What is the weight of the Earth?
A: 150 Million Km
(…no comments…)

Q: Who invented radio?
A: Horzepa Stan Surfin’…
(Huh?!)

Q:What do butterflies eat?
A: Nectar
(Correct.)

Q:What do elephants eat?
A: Pounds of Food
(Great answer!)

My last question was: What is Lexxe? And the system proudly answered: 3rd generation search engine.

The answers are not the only output of the system: it also produces clusterized search results, links to the definitions of the key words in the question etc. And these features look good. The engine is still in the alpha stage, so we can hope for improvement.

I like the idea of the search based on natural language requests. I definitely will keep an eye on Lexxe, and will add it to my bookmarks. But for the time being to get answers I’d rather use Google+Wikipedia.
Technorati tags: web2.0, search, Lexxe, Natural language