Teacher : Dr. Wan-Shiou Yang Student : Jia-Ben Dia

1. F O C IA Personalized Web Intelligence SystemAh-Hwee Tan, Hwee-Leng, Hong Pan, Jamie Ng, Qiu-Xiang Li Teacher : Dr. Wan-Shiou Yang Student : Jia-Ben Dia

2. Outline • Introduction • FOCI System Architecture • User-configurable Clustering • Experiments • Conclusions

3. Introduction • In the knowledge-based era, it has become increasingly risky to do business without intelligence. • It’s labor intensive to compile and organize the information on world wide web. • Copernics, BullsEye, and NorthernLight organize search results into automatically generated folders to facilitate navigation and browsing. (Gathering purpose only)

4. Introduction (cont.) • This paper introduces a system called FOCI(Flexible Organizer for Competitive Intelligence) : To assist knowledge workers to perform competitive intelligence on the web. • To Construct information portfolios. • To Personalization the portfolios.(incorporates users’ preferences in an information clustering system) • Living reportsthat can be published and shared b other users.

5. FOCI System Architecture • Content management module • For organizing and personalizing portfolios • Content mining module • For analyzing information portfolios • Content publishing module • For sharing of portfolios • User interface module • For graphical visualization and user interactions

6. FOCI System Architecture (cont.)

7. User-configurable Clustering

8. User-configurable Clustering (CONT.) A user-configurable clustering system comprises an information clustering engine for clustering of information based on similarities , • User interface module • Personalization module • Cluster structure knowledge based

9. User-configurable Clustering (CONT.)Feature Representation

10. User-configurable Clustering (CONT.)Feature Representation

11. User-configurable Clustering (CONT.)Feature Representation

12. User-configurable Clustering (CONT.)Clustering Engine The information clustering engine is based on fuzzy ARAM that performs a combination of unsupervised learning and supervised learning.

13. User-configurable Clustering (CONT.)

14. User-configurable Clustering (CONT.)Clustering • Given an information vector A • Searches for a F2 cluster J encoding a template information vector that is closest to the information vector A according to a choice function. • Then check if of the selected category matches with A according to match criterion. • If so, the template information vector of the F2 cluster J is modified to encode the input information. • Otherwise, the system repeats to select another cluster until a match is found or a new cluster is created.

15. User-configurable Clustering (CONT.)Personalization • ARAM can also operate in an insertion mode whereby a pair of information and preference vectors can be inserted directly into an ARAM network. • We present the key cluster personalization functions below : • Labeling information clusters • Inserting information clusters • Merging information clusters • Splitting Information clusters

16. Experiments • How FOCI can be used to create, organize, and track specific topics of interests? • The objective of the experiments is twofold : • To show how the personalization functions can be used to support a variety of organization functions. • To demonstrate how a personalized portfolio can serve as a template for organizing new information.

17. Experiments (cont.)Clustering An information portfolio on “text mining” is created by integrating search results of four internet search engines.

18. Experiments (cont.)Personalization A user ma use the various cluster manipulation functions, namely labeling, inserting, merging, and splitting to personalize his/her portfolios.

19. Experiments (cont.)Personalization An exemplary fully personalized portfolio.

20. Experiments (cont.)Tracking We study the effect of tracking and clustering new information using the personalized portfolio.

21. Experiments (cont.)Tracking The system is able to discover novel information groupings while organizing familiar information into the user’s personalized portfolio.

22. Conclusions • FOCI that enables a user to create and manage personal information portfolios. • Personalization in FOCI is achieved via user-configurable clustering.