Conceptual Tasks CTs for the Personalized Electronic News System PENS

1
Conceptual Tasks (CTs)for the Personalized
Electronic News System (PENS)

• Jie Zhang

Adwert Group
2
Conceptual Tasks (CTs)

• CTs are responsible to provide information about
  the user, about the users usage, and content of
  the site.
• Provided data shapes the dynamic aspect of the
  system
• Are application specific in general
• Communicate with the Synthesis Engine (SE)
  frequently through the interface between the SE
  and CTs (SE2CT)
• Currently implemented conceptual tasks include
  the User Profile Manager, the Usage Group
  handler, the News Feeder and the Association Miner

3
Communication between a CT and SE2CT

• Accepts requests from SE2CT which are relayed
  from SE
• Generates responses which will later be relayed
  to SE through SE2CT
• Format of packet received from SE2CT
• Example of parameters defaultpassworddefault
• Format of packet sent from CT to SE2CT

4
User/Group Model in Intermediate Format

• Includes domain specific user/group model
  vocabulary and application specific user/group
  model
• They are described by using the Resource
  Description Framework (RDF) serialized in XML
• They will be interpreted and used for
  initializing database tables and instantiating
  user/group profiles
• They are configurable through an authoring tool

5
Example of User Model Vocabulary

• ltrdfsDescription rdfID"UserModel"gt
• ltrdftype rdfresource"http//www.w3.org/2000/0
  1/rdf-schemaClass" /gt
• ltrdfscommentgtUser Model is a collection of
  properties about an userlt/rdfscommentgt
• lt/rdfsDescriptiongt
• ltrdfProperty rdfID"dgmodel"gt
• ltrdfsdomain rdfresource"UserModel" /gt
• ltrdfsrange rdfresource"DGModel" /gt
• ltrdfscommentgtdemographic information about the
  userlt/rdfscommentgt lt/rdfPropertygt
• ltrdfsClass rdfID"DGModel"gt
• ltrdfscommentgtUser model that represents all
  the demographic information about the
  userlt/rdfscommentgt lt/rdfsClassgt
• ltrdfProperty rdfID"age"gt
• ltrdfsdomain rdfresource"DGModel" /gt
• ltrdfscommentgtUser's age is represented by a
  interger numberlt/rdfscommentgt
• ltrdfsrange rdfresource"xsiinteger" /gt
  lt/rdfPropertygt

6
Example of User Model

• ltumdvUserModel rdfID "defaultUserProfile"gt
• ltumdvdgInfor rdfresource "dgInfor"/gt
• ltumdvinterestedTopics rdfresource
  "interestedTopics"/gt
• ltumdvvisitedNews rdfresource
  "visitedNews"/gt
• ltumdvgroupInfor rdfresource "groupInfor"/gt
• lt/umdvUserModelgt
• ltumdvDGModel rdfID "dgInfor"gt
• ltumdvagegt28lt/umdvagegt
• ltumdvgendergtmalelt/umdvgendergt
• ltumdveducationLevelgtbechalorlt/umdveducationLeve
  lgt
• ltumdveducationFieldgtComputer Sciencelt/umdveduca
  tionFieldgt
• ltumdvoccupationgtprogrammerlt/umdvoccupationgt
• ltumdvjobFieldgtITlt/umdvjobFieldgt
• ltumdvenglishLevelgtmediumlt/umdvenglishLevelgt
• ltumdvemailgtias_at_unb.calt/umdvemailgt
• lt/umdvDGModelgt

7
Example of Group Model Vocabulary

• ltrdfsDescription rdfID"GroupModel"gt
• ltrdftype rdfresource"http//www.w3.org/2000/0
  1/rdf-schemaClass" /gt
• ltrdfscommentgtGroup Model is a collection of
  properties about a group of users.lt/rdfscommentgt
  
• lt/rdfsDescriptiongt
• ltrdfProperty rdfID"groupID"gt
• ltrdfsdomain rdfresource"GroupModel" /gt
• ltrdfscommentgtGroupID uniquely identifies each
  group by a number.lt/rdfscommentgt
• ltrdfsrange rdfresource"xsiinteger" /gt
• lt/rdfPropertygt
• ltrdfProperty rdfID"academic"gt
• ltrdfsdomain rdfresource"GroupModel" /gt
• ltrdfscommentgtThe interests value of academic
  news for the center of the group is represented
  by a float number.lt/rdfscommentgt
• ltrdfsrange rdfparseType"Literal" /gt
• lt/rdfPropertygt

8
Example of Group Model

• ltgmdvGroupModel rdfID "defaultGroupProfile"gt
• ltgmdvgroupIDgt-1lt/gmdvgroupIDgt
• ltgmdvgroupPopugt-1lt/gmdvgroupPopugt
• ltgmdvnewPopugt-1lt/gmdvnewPopugt
• ltgmdvfinancialgt-1lt/gmdvfinancialgt
• ltgmdvacademicgt-1lt/gmdvacademicgt
• ltgmdvsocialgt-1lt/gmdvsocialgt
• ltgmdvpeoplegt-1lt/gmdvpeoplegt
• ltgmdvmiscgt-1lt/gmdvmiscgt
• lt/gmdvGroupModelgt

9
(No Transcript)
10
User Profile Manager

• is responsible to provide information about the
  user and about the users usage
• is also in charge of instantiating the profile
  from user model and updating the profile fields
• In the initialization stage, it reads the RDF
  description, creates database tables for the user
  model, and inserts the default user profile into
  the database

11
Class Diagram
12
Services Provided by UPM
13
Profile Editor A Tool Relates to UPM

• Is a tool to maintain user profiles directly
  through user
• Implemented by using J2EE technique
• A link on the front page will pass the current
  users ID to the J2EE server. A servlet page
  parses the users ID and retrieves user
  information from the database. User can change
  her property values through the Profile Editor.
  The request of changes will be replied by the
  server. Changes will be made on the database as
  well.
• Makes use of services provided by UPM

14
Usage Group Handler

• is responsible to provide information about the
  group of users
• is responsible for creating and maintaining
  groups of users that can be clustered based on
  their interests or behavior.
• is also in charge of grouping existing users
  according to their characteristics
• will assign a new user or reassign an older user
  into a group based on how close they are to the
  group centers
• reads the RDF description for group model and
  creates database for the group model

15
Class Diagram
16
Visualization Tool A Tool Related to User Groups

• Is used for visualizing user groups
• Each user is represented as a dot
• Members of different groups are represented in
  different colors
• Groups are separated
• The current user will be highlighted
• Group interests are listed

17
Implementation of Visualization Tool

• On the client side, there is an applet to show
  the graphical representation of users and groups.
• On the server side, The Visualizer sends back the
  transformed current user and all the user groups.
• All the user groups will be re-transformed after
  the groups are re-generated.
• The Transformer retrieves information about each
  user and group information from the database. It
  then calls matlab to compute the principal
  components and transforms the data (user
  information). The two most important principal
  components will be selected and the corresponding
  coordinates of transformed data will be sent back
  to the Interface, and then passed to the client
  applet.

18
Association Miner

• is responsible to provide recommendations to the
  user when she is reading a news item according to
  the association relationship between the current
  news item and other news items discovered by the
  mining process
• mines the usage data (users navigation history,
  for example), generates association rules
  according to the usage data, and provides
  recommendations according to the discovered
  association rules and the users current visiting
  news item

19
Class Diagram
20
Cooperation between the UGH and the AM
21
Recommendation Algorithm

• First In each group the Miner mines visiting
  history performed by the users in the group to
  discover association rules through the Apriori
  algorithm
• Second In each group the Miner also finds out
  the 6 most popularly visited news stories from
  all users visiting history
• Third If the number of the most popularly
  visited news stories is no more than 6 (this case
  happens when system starts and not too much
  visiting history), some latest news stories
  should be added
• Forth The Recommender finds out related news
  stories with the current reading news story from
  association rules
• Fifth If the number of the associated news
  stories is not enough for what has been requested
  (no more than 6), the Recommender will add the
  rest from the popular news found out in the
  second and third steps.

22
New Feeder

• provides information about news to other
  components in the system
• continuously retrieves news from news feeds,
  extracts useful information and stores them in
  the news database
• The news feeds are written in XML based on RSS
  specification.
• retrieves news feeds through HTTP protocol. It
  also interprets the XML files and extracts the
  information that the system needs. Useful
  information is stored in the database without
  redundancy. After a certain time, the NR will
  read the news feeds again and check whether there
  is new news story or not. If so, the new news
  story will be stored in the database.

23
Class Diagram
24
Instructions Provided by the NF