Next Gen Search: Thinking Engines

jmiller · #1 (**permalink**) 08-02-2005, 02:34 PM

For those who’ve been keeping up, keyword relevance-based searches will soon look like top-loading VCR’s; they still may work but who has the time? Soon search engines will think, as evidenced by Engenium’s latest patented application, Semetric V.4.0.

Yahoo, Google, and MSN are looking into a more contextualized search experience, indeed this is the next wave, but now we have the technology “ConceptSpace” which powers the new Semetric.

According to Engenium, Semetric “learns” as users search, familiarizing itself with the nuances of the language. The algorithms designed by Dallas based Engenium are modeled on human speech and thinking.

By creating relationships between words, the technology scans the content of documents, judges how each word, phrase, or concept is related to other words, phrases and concepts in the document in order to retrieve more precise results.

“Because Engenium Semetric® operates on concepts -- not keywords -- search and retrieval isn't constrained by the language or lexicon used in queries. Advanced mathematics is leveraged to facilitate the retrieval of relevant content, even when disparate content elements do not share any words or objects in common. Thus, users benefit through more accurate, precise search results,” as posted on their Website.

To borrow from Business Week, here’s an example.

“Say you want to search for documents on sedimentation, water rights, run-off, and environmental impact. When you ask [the program] to do a contextual search for those terms, the engine might also return other unexpected terms such as riparian (an obscure word used to describe lush vegetation on the sides of waterways). And it might return the name of someone who you hadn't expected to be involved in these matters.”

Semetric is invisible to end-users using existing interfaces to run queries while it completes each search “behind the scenes.” Engenium says the application also works for users unskilled in constructing search queries, for as soon a document fits what a searcher is looking for, the entire document can be used as a query for further searching.

Currently, the product is for businesses and other entities, like law firms and government agencies, who need to search volumes of material on their own databases. But it won’t be long until contextual search is applied to the Internet.

Business Week says, “The contextual search function lets you find things you don’t even know you’re searching for.”

Huron Consulting Group’s Jim Mitchell tested the system out on a securities suit.

"It was bringing up material that hadn't been looked at when you search exclusively on those key terms," says Mitchell. "That's big."

kgun · #2 (**permalink**) 08-03-2005, 11:40 AM

The Back Office Robot is already a fact.

Examples:

http://www.hotcomm.com/IMLive.asp

http://www.oddcast.com/home/

I wrote a master thesis in mathematics on nonlinear, chaotic and fractal structure in finance (295 pages in Norwegian) in 1996.

Neural Networks (AI) and Fuzzy Logic may be used to train a bot. You get a database of correspondence (logs) that is fed into to Bots memory, and it gets smarter and smarter.

MFIBot:
This is the Bot of MultiFinanceIT talking. How are you today? (First 20 words, first 20 seconds very important).

I am a Bot and not a human being, so you know that, but I shall try to serve you as best I can. Do you have any questions about our solutions (not products)? Yes (hit Y) / No (hit N)?

Do you have any additional questions? If I cannot answer you, you may give me your phone number or your email address and my friend, Uncle Sam will call (contact) you tomorrow.

Have a nice day. Nice to serve you.

Do you see the advantages, n to 1 relationsships. Bandwith etc. sets the limit for communication capacity.

After a while the bot may get better answers than a human being because of full overview of products and solutions. And it is always polite.
http://www.webproworld.com/viewtopic...r=asc&&start=0

In the same way you may use contextual search to train a search engine.

Did you ask me a question?

Kjell Gunnar Bleivik
http://www.multifinanceit.com/
http://www.blognorway.com/

AcmeSEO · #3 (**permalink**) 08-03-2005, 05:36 PM

Actually, "contextual" indexing and retrieval has been part of the algo at the "theme-based" engines (Google, Alltheweb and Lycos) for a couple years now at least. Some of the components are Latent Symantic Indexing and Term Vectoring. If you've got the patience and time, Study those subjects. My Clients and I have been prospering from knowledge of how this works for almost two years now. And I agree that this area is the most interesting and where many of the coolest innovations of search marketing are and will be appearing.

Optimized regards,

Robert

kgun · #4 (**permalink**) 08-03-2005, 06:26 PM

Does it have a relation to principal components and / or eigenvalues? Hilbert spaces and projections in Hilbert spaces?

Do you have some good references better than a simple search?

"Latent Symantic Indexing and Term Vectoring"

Which technique does
http://www.nedstatbasic.net/
Use?

Good to uncover surfers habits as an alternative to KW popularity.

Kjell Gunnar Bleivik
http://www.multifinanceit.com
http://www.blognorway.com

kgun · #5 (**permalink**) 08-06-2005, 03:10 AM

http://www.subjex.net/servlet/Subjex...?showmain=true

Be sure the boys in the sandbox at Google, have their toys.

Search term:

Latent Symantic Indexing and Term Vectoring

One hit.

Did you mean: Latent Semantic Indexing and Term Vectoring

74 hits.

That is the first step.

My term:
Pattern reckognition, the mapping of fuzzy sets and projections in Hilbert spaces.
0 hits

Google:
Did you mean
Pattern recognition, the mapping of fuzzy sets and projections in Hilbert spaces.
4050 hits.

Be aware: The term includes the AND operator, but that is another story.

Professor GoogleBot, you are the most knowledgable guy at Google. What about your Chinese friend, http://www.baidu.com/? Do you have a good relationship to http://www.asiabot.com/ ? I am sure. You are more polite than us human beings.

It is too early to know the future.

http://www.multifinanceit.com/

TrafficProducer · #6 (**permalink**) 08-11-2006, 03:57 PM

Buthttp://www.oddcast.com/home/ Great Laugh LOL :)

Is is't Pigeons that power Google Page Rank and not Monkeys

>>>><<<
Three Wise Monkeys are joined by another Monkey to make four wise Monkeys.
Adult humour. If you are adult enough Click your cursor at the Three wise Monkeys to see the fourth.

kgun · #7 (**permalink**) 08-12-2006, 12:17 PM

<this is a joke>

asm

{

Yes, it is my pigeons. I just teached them exception handling, name spaces and generics in C++ :-)

}

</this is a joke>

lourdes · #8 (**permalink**) 08-13-2006, 03:36 PM

Very good information.

eleap · #9 (**permalink**) 08-14-2006, 07:37 PM

IMHO isn't all of this dependant on the quality of input into se's ... I mean mining terabytes of garbage still means garbage out yes?

How does the semantic web concept intend to deal with spam?

Don

lourdes · #10 (**permalink**) 08-14-2006, 10:48 PM

Great article.

kgun · #11 (**permalink**) 11-08-2006, 03:02 PM

I have collected some important SE links at SurfToolbar.

TrafficProducer · #12 (**permalink**) 06-22-2007, 06:35 PM

According to Engenium: Maybe the important word is According

I don't believe Thinking/ Intelligence has not been found to exist in Computer Technology, or any other Mechanical device yet

Turing Test, (Wikipedia), a proposal for a test of a machine's capability to demonstrate thought. Described by Professor Alan Turing in the 1950 paper "Computing machinery and intelligence," it proceeds as follows: a human judge engages in a natural language conversation with two other parties, one a human and the other a machine; if the judge cannot reliably tell which is which, then the machine is said to pass the test. It is assumed that both the human and the machine try to appear human. In order to keep the test setting simple and universal (to explicitly test the linguistic capability of the machine instead of its ability to render words into audio), the conversation is usually limited to a text-only channel such as a teletype machine as Turing suggested or, more recently, IRC or instant messaging.
Turing Test, a method to gauge artificial intelligence. (Computer World)
Has the Turing Test become Worthless? (Android World)
The Alan Turing Home Page
What Is Artificial Intelligence?

Thread Tools
Show Printable Version Email this Page
Display Modes
Linear Mode Switch to Hybrid Mode Switch to Threaded Mode