Re: Billed-as Ultimate Search Engine...

Autonomy is using Bayesian text classification to classify and cluster documents and possibly to construct a de facto ontology tailored to specific problems using feature extraction techniques.. Bayesian text
classification is a proven winner in this task (as is latent semantic indexing). The only real problem is that it is a supervised learning technique (although this isn't a real problem in the end). They use agents to do most of their processing - presumably they mean something more than moderately intelligent webcrawlers. Actually I have recommended to a company that they look at Autonomy because they want to provide specialized search/directory functionality for their business and don't have an army of editors like Yahoo or dmoz.

Kirt Undercoffer

Woodcock <EVSmalley_at_blackthespam.earthlink.net> wrote in message news:8e8v3p$k4f$1_at_mulga.cs.mu.OZ.AU...
>
>
> I think I may need a reality check on this. Anyone?
>
> Wired mag, 2/00 "The Quest for Meaning": This article pretty much
> convinced me that a piece of software I've never heard of is going to end
> up running the world 5 years from now, which, when I take a step back,
> seems unlikely.
>
> Did anyone else read this, or has anybody used this Autonomy Systems
> software? The gist of the story is that the real currency of the Internet
> and the Internet Age is so-called "unstructured data", which just means
> text in various formats on various topics. But the problem is that even
> though this text is what humans want and need from the Internet,
> computers don't really know what to do with it, and we have to tell them,
> which defeats the purpose of the whole thing. Computers are good with
> numbers and formalized programming languages, but they don't know how to
> read a magazine article and know whether you'll like it.
>
> And so now according to the article the company Autonomy has changed all
> that, by using a hundreds of years old algorithm to give the skills of,
> quote "comprehending context, generalizing from words to an idea, even
> understanding the unspoken by grasping the root concepts beneath the play
> of syntax." And so supposedly the software can read anything in any
> language (without actually knowing what the words mean) and link
> paragraphs in one document with most closely related documents from its
> source (which could The Web, or The Company's Complete Knowledge Record).
> And the implication is that this all works well enough that it can take
> the place of human beans doing the same work.
>
> The article suggest that these people are the only ones with a product
> that can do this and it quotes the CEO saying they're on the way to
> becoming "the Oracle of unstructured information." Meaning, they'll be
> everywhere.
>
> If somebody could point out the flaws, catches, and exaggerations so that
> I can go back to my normal life, that'd be great.
>
> Edward
>

[ comp.ai is moderated.  To submit, just post and be patient, or if ]
[ that fails mail your article to <comp-ai_at_moderators.isc.org>, and ]
[ ask your news administrator to fix the problems with your system. ]