An Intelligent System forAn Intelligent System forDynamic Online Allocation of Dynamic Online Allocation of

Information on Demand Information on Demand from the Internetfrom the Internet

Thamar E. Mora, Rene V. MayorgaThamar E. Mora, Rene V. Mayorga

Faculty of Engineering, Faculty of Engineering,

University of Regina, University of Regina,

Regina, Saskatchewan, Regina, Saskatchewan,

CanadaCanada

Intelligent System and ObjectiveIntelligent System and Objective

Proof of concept of Intelligent SystemProof of concept of Intelligent System

Intelligent System based on Fuzzy Inference SystemIntelligent System based on Fuzzy Inference System

To Customize and Allocate Dynamically Online Information from the To Customize and Allocate Dynamically Online Information from the Internet Internet

IntroductionIntroduction

Advances in Computer and Communications Technology have led Advances in Computer and Communications Technology have led to to - - Information ConvergenceInformation Convergence

No longer Video on Demand; but rather No longer Video on Demand; but rather - - Information on DemandInformation on Demand

The Internet contains plenty of data, leading to The Internet contains plenty of data, leading to - - Information SaturationInformation Saturation

BackgroundBackground

Already available some tools for: Already available some tools for:

- Interfaces - Interfaces - Browsers - Browsers

- Customized Web sites- Customized Web sites

The authors recently proposed: The authors recently proposed:

- - Intelligent SystemIntelligent System, based on a , based on a

- - Fuzzy Inference SystemFuzzy Inference System, for , for

- - Dynamic On-Line Portal CustomizationDynamic On-Line Portal Customization, and , and

- - Intelligent Web AdvertisingIntelligent Web Advertising

The authors also recently proposed: The authors also recently proposed:

- - Intelligent SystemIntelligent System, based on a , based on a

- - Fuzzy Inference SystemFuzzy Inference System, for , for

- - Dynamic On-Line Dynamic On-Line

- TV Programming Allocation- TV Programming Allocation from from

- - TV Internet BraodcastingTV Internet Braodcasting

Proposed Intelligent SystemProposed Intelligent System

The user provides as inputs the type of The user provides as inputs the type of informationinformation customizationcustomization that he/she desires to receivethat he/she desires to receive

According to the According to the user preferred selectionuser preferred selection, a data gathering process , a data gathering process (if the information is not already available in a database) is started(if the information is not already available in a database) is started

This data is processed though a This data is processed though a - Fuzzy Inference System - Fuzzy Inference System - prompting as - prompting as outputoutput - the - the kindkind and and amountamount of information, and of information, and - the most appropriate - the most appropriate mediamedia from which the from which the - information is to be received- information is to be received

User Preferred SelectionUser Preferred Selection The user makes a The user makes a preferred selectionpreferred selection to receive information based on: to receive information based on:

Language - Country; Language - Country;

Level of preference about certain topics Level of preference about certain topics - World, business, politics, technology, - World, business, politics, technology,

entertainment, sports, health, weather, etc.;entertainment, sports, health, weather, etc.;

Level of preference about certain media Level of preference about certain media - Television, Radio, Newspapers, Magazines, - Television, Radio, Newspapers, Magazines,

Journals, Photojournalism, etc.; Journals, Photojournalism, etc.;

Level of desired detail in the output diagnostic Level of desired detail in the output diagnostic - Low, Medium, High - Low, Medium, High

It is an input to our Intelligent Agent, but not the FISIt is an input to our Intelligent Agent, but not the FIS

Fig.2. Screen to input the preferences

Inputs to the FIS Inputs to the FIS

Six inputs to the Fuzzy Inference System:Six inputs to the Fuzzy Inference System:

Three inputs with a high level of preference, each to set the level of Three inputs with a high level of preference, each to set the level of preference of interest for a particular topic preference of interest for a particular topic

Three inputs with a high level of preference, each to set the level of Three inputs with a high level of preference, each to set the level of preference to receive information from a particular media preference to receive information from a particular media

The number of inputs can be changed for a larger or smaller The number of inputs can be changed for a larger or smaller numbernumber

In this project the number of inputs is considered relatively small in In this project the number of inputs is considered relatively small in order to provide better-customized options, and not just a large listorder to provide better-customized options, and not just a large list

Outputs from the FISOutputs from the FIS

Two outputs: Two outputs:

The The kindkind (particular links), and the (particular links), and the amount amount (number of links) (number of links) of information to be displayed, and of information to be displayed, and

And the most appropriate And the most appropriate mediamedia from which the information is to be from which the information is to be displayeddisplayed

A portal-type customization is dynamically generated online with proper A portal-type customization is dynamically generated online with proper links according to the user preferenceslinks according to the user preferences

The number of links plays the role of pondering the importance in the The number of links plays the role of pondering the importance in the decisiondecision

Fuzzy Inference SystemFuzzy Inference System

FIS is a Mamdani typeFIS is a Mamdani type

Uses the Uses the centroidcentroid as the defuzzification as the defuzzification method method

The membership functions (MFs) for all the linguistic values are The membership functions (MFs) for all the linguistic values are triangular triangular

MatlabMatlab based based

Figure 3. Fuzzy Inference System Structure

Rules StructureRules Structure The linguistic values for the inputs are The linguistic values for the inputs are

“ “not too muchnot too much””, “, “regularregular”” and “ and “too muchtoo much””

For the outputs, numbers are defined as the labels.For the outputs, numbers are defined as the labels.

In general, a Fuzzy knowledge model consists of a set of rules of the form: In general, a Fuzzy knowledge model consists of a set of rules of the form:

- If - If xx is is AA then then yy is is BB

The current prototype includes 54 rules. The current prototype includes 54 rules.

These rules are determined according to the smoothness of the rules surface.These rules are determined according to the smoothness of the rules surface.

The structure of the rules follows the following pattern The structure of the rules follows the following pattern

- - Topic1Topic1 and and Topic2 Topic2 and and Topic3 Topic3 LinksLinks - Media1 - Media1 and and Media2 Media2 and and Media3 Media3 MediaMedia

Figure 4. Example of a View of the Rules’ Surface

Rules StructureRules Structure

It is also possible to consider the option of allowing the specification It is also possible to consider the option of allowing the specification of the of the mediamedia’’s preference for each s preference for each topictopic. .

However, this will give a Fuzzy Inference System similar to the one However, this will give a Fuzzy Inference System similar to the one shown in Figure 5.shown in Figure 5.

In this case a much larger set of rules is needed than the current In this case a much larger set of rules is needed than the current prototype.prototype.

Fig.5. Option that allows setting the preference for media in each topic

Example 1Example 1

The user preferred selection:The user preferred selection: - Language Country (USA); - Language Country (USA);

- Level of interest for a particular Topic: Fig. 6; - Level of interest for a particular Topic: Fig. 6; - Level of preference for a particular - Level of preference for a particular media: Fig 6; media: Fig 6;

- Detail Level: Medium- Detail Level: Medium

According to the user preferences in Fig. 6, our Intelligent According to the user preferences in Fig. 6, our Intelligent System prompts the FIS output taking into account the System prompts the FIS output taking into account the preferred level of detail to display the informationpreferred level of detail to display the information

Fig. 6. User preferences for Example

Example 1Example 1

The Intelligent System output is:The Intelligent System output is:

Display Display - three links of news around the - three links of news around the WorldWorld, , - three links related to - three links related to BusinessBusiness news; and news; and - two links for the latest on - two links for the latest on PoliticsPolitics

From From - - NewspapersNewspapers sites sites - in - in United States of AmericaUnited States of America

Fig. 7. Output of the agent for the preferences shown in Figure 6 for Example 1

Example 1Example 1

Figure 7 shows the corresponding icons: Figure 7 shows the corresponding icons: the gateway for the user to reach the desired information. the gateway for the user to reach the desired information.

Here the example is illustrated with icons, but behind the icons the Here the example is illustrated with icons, but behind the icons the corresponding web site addresses are:corresponding web site addresses are:

http:/nytimes.com/pages/http:/nytimes.com/pages/worldworld/index.html/index.html http://nytimes.com/pageshttp://nytimes.com/pages/business/business/index.html/index.html http://nytimes.com/pages/http://nytimes.com/pages/politicspolitics/index.html/index.html

Example 2Example 2

The user preferred selection:The user preferred selection: - Language Country (USA); - Language Country (USA);

- Level of interest for a particular Topic: Fig. 8; - Level of interest for a particular Topic: Fig. 8; - Level of preference for a particular - Level of preference for a particular media: Fig 8; media: Fig 8;

- Detail Level: Low- Detail Level: Low

According to the user preferences in Figure 8, our Intelligent According to the user preferences in Figure 8, our Intelligent System prompts the FIS output taking into account the preferred System prompts the FIS output taking into account the preferred level of detail to display the informationlevel of detail to display the information

Fig. 8. User preferences for Example 2

Example 2Example 2

The Intelligent System output is:The Intelligent System output is:

Display Display

- two links of news around the - two links of news around the WorldWorld, ,

- two links related to - two links related to TechnologyTechnology news; and news; and

- one link for the latest on the - one link for the latest on the WhetherWhether

From From

- - TelevisionTelevision sites sites

- in - in United States of AmericaUnited States of America

Fig.9. Output of the agent with links for Television programs about news around the World broadcast in the Internet

Fig.10. Links for Television programs related to Technology

Fig.11. Link for a Television program with the latest about current Weather news broadcast in the Internet

Example 2Example 2

The Agent prompts the links for programs available at the time. The Agent prompts the links for programs available at the time.

Here, the television screen obtained after selecting and clicking a Here, the television screen obtained after selecting and clicking a link is shown. link is shown.

The user has access to the different broadcast options inferred from The user has access to the different broadcast options inferred from his/her preferences as shown Figures 9 - 11.his/her preferences as shown Figures 9 - 11.

For the special case of For the special case of Television, Television, our Intelligent System prompts our Intelligent System prompts online programming or the latest recorded programs that were online programming or the latest recorded programs that were broadcast in the net.broadcast in the net.

ConclusionsConclusions

An Intelligent System as a Proof of ConceptAn Intelligent System as a Proof of Concept

Dynamic Online Information Allocation from the InternetDynamic Online Information Allocation from the Internet

FIS architecture as a framework for intelligent decisions about FIS architecture as a framework for intelligent decisions about the the kindkind, the , the quantity of informationquantity of information, and , and the the mediamedia from which it is to be displayed from which it is to be displayed

Current Intelligent System, a Generalization from our own Current Intelligent System, a Generalization from our own previous Intelligent System: previous Intelligent System: Intelligent System for Dynamic On-Line Intelligent System for Dynamic On-Line TV Programming Allocation from TV Programming Allocation from TV Internet Broadcasting TV Internet Broadcasting - IASTED ISC’2001- IASTED ISC’2001

Thanks !Thanks !