Distributed Agents to Retrieve Web Intelligence
email E-mail    
HOME                 Registration        About us  Alliances  Contact us |
 Attractions    Bibliography    Classified Papers    Search    Infomaps    Site Map    i-Web    Images Bank    White Papers  
Login Area 11111
Forgot your password?

free counters
Agents Tour | AI-Lab | Darwin Tour | Faq | News | Newsletter | Pag's | Press Releases | Search Tutorial

How Darwin-FIRST Agents work

In the figure below we depict how our agents work. Among Darwin-FIRST family of Agents we choose Procurebot and Retrievebot, Procurement and Retrieve Agents respectively, to illustrate how they make the Cognitive Objects Collection evolve. Evolution is performed along several concurrent ways, namely Retrieving and/or Updating i-URL’s down or with obsolete content, and empowering the cognitive offer to users. Let’s see then how procurebot agents work.

In the upper right corner is shown the matchmaking process between a i-Website and its users. As you may see in Darwin-FIRST Tour and in our Darwin-FIRST White Papers, Website owners and users “speak” different jargons and even “think” different. Users speak and think apparently in a chaotic way, querying, browsing and clicking at will being their “messages” expressed by “keywords” and “instances” of browsing and clicking. On the owners side, owners present their Cognitive Offer organized by “subjects” of a Logical Tree (Menus, Indexes, Catalogues, Data Maps, etc).

The messages go in both directions through a particular matchmaking interface that resembles a “permeable” osmotic e-membrane (yellow slabs). Sometimes What User Wants (WUW) matches What Owners Offer (WOO) and sometimes don’t. As you will find along the pages of this site our search strategy is based on pairs (k, s), keyword, subject, pertaining to different realms: keywords pertain to the users’ realm and subjects to the owners’ realm and as you may see in our White Papers the become the People’s realm and the Web Space Establishment of the Web Society respectively.

When mismatch occurs, a user tracking agent belonging to the family of welcome agents (wellcomebots), stacks the unsatisfied pair in a mismatches stacking. It means that our i-Map doesn’t have any cognitive object to satisfy the user demand, something considered crucial. Periodically procurebot agents are patrolling the stacks and assume as their mission to procure valuable documents to satisfy the oldest demands stacked. A procurebot active in mission proceed then to look for how to satisfy the unsatisfied demand concurring to a Pool of Conventional Search Engines. They have a well precise commitment, predefined and periodically adjusted by the Chief Editor in a tune-up process for each major Subject (discipline). This commitment is defined in terms of Preferences, Mask/Filters, Priorities and Core keywords, (see what they mean in Thesaurus), just the same as an instructive given to a human being, to a professional searcher.

With this commitment the agent, equipped with a fast convergent algorithm, “play” something like an “Interception game” with the Web using the Conventional Search Engines as a gateway trying to reduce the uncertainty from millions to a few that are then conveyed to another stack of “suggested URL’s” to be checked and approved by the Chief Editor.

The Chief Editor in its turn through the Darwin-FIRST Desktop, as you may see below, inspects the answers to the unsatisfied pairs, in the form ordered lists of suggested URL’ for each pair, and selects one or as many as necessary to fully satisfy the pair. Now i-URL’s editing cycle starts.

For each URL selected as either an authority or as a hub deserving to be part of our i-Map, the Editors must edit its corresponding i-URL.

Note: This is a human task that requires specialized skill but it has low volume relative to the whole i-Map, from 0,1% to 1% monthly depending of the “volatility” of the content. We have to take into consideration that the appearance of unsatisfied pairs (k, s) has low probability because the richness of our Thesaurus. Most probable is the appearance of documents considered poor and consequently rejected by users. In these cases the retrieving task of similar but better documents is performed by our retrievebot agents that work similar.

Once a document is retrieved and when document/s that satisfies previously unsatisfied pairs is/are edited, a welcome agent warns the unsatisfied users by e-mail.

The Chief Editor is able to make the agents work stepwise, like when programmers debug their programs, seeing how well/wrong and efficient/inefficient they work in each step, and adjusting parameters accordingly. Darwin-FIRST agents are built with high redundancy concerning routines and scripts that browse the uncertainty space in as many dimensions as “a priori” imagined. The Chief Editor could enable/disable these “boxes” in order to fix agents for each Major Subject.

How It Looks Like in the Web

Back to the last page

Copyright © 2003-2013 Darwin! Inc. All rights reserved.