Systems I've seen are also capable of extracting noun phrases and verb phrases, and weighing the relevence of those. If 'the lazy sleeping dog' occurres a couple of times (especially across paragraph boundries) it will score high to be part of an overall summary of the document. Ferreting out the parts of a verb phrase can be quite a bit more difficult, because they can be bigger, containing noun phrases, prepositional phrases, they can be nested, yada yada. You need to have a lexicon so the system can pick out a word, and classify it as a noun, adjective, adverb, verb, and the heirarchial relationship they are allowed to have with one another. And after all that, you have to remember that in english (and alot, but not all other languages) word position is syntactically significant, and that there's more than one (syntactically relevant) way to say the same thing.Which is why it would be hard.
But it wouldn't be impossible. There's a company in Canada that does software like this (in Englis, German and French, I believe) called Nstein. I've seen a demo and its very impressive.
Slashdot Mirror
Systems I've seen are also capable of extracting noun phrases and verb phrases, and weighing the relevence of those. If 'the lazy sleeping dog' occurres a couple of times (especially across paragraph boundries) it will score high to be part of an overall summary of the document. Ferreting out the parts of a verb phrase can be quite a bit more difficult, because they can be bigger, containing noun phrases, prepositional phrases, they can be nested, yada yada. You need to have a lexicon so the system can pick out a word, and classify it as a noun, adjective, adverb, verb, and the heirarchial relationship they are allowed to have with one another. And after all that, you have to remember that in english (and alot, but not all other languages) word position is syntactically significant, and that there's more than one (syntactically relevant) way to say the same thing.Which is why it would be hard.
But it wouldn't be impossible. There's a company in Canada that does software like this (in Englis, German and French, I believe) called Nstein. I've seen a demo and its very impressive.
Isn't it the case that the new 2.4-test1 kernel has the beginnings of LVM?