1 / 16

Personalization

Personalization. Speaker: Ping-Tsun Chang 3/7/2002. Personalization of WWW10. Designing Personalized Web Applications Session: Personalization in E-Commerce Gustavo Rossi, Daniel Schwabe, Robson Guimaraes, Dept. of Informatics, PUC-Rio , Brazil. Personalizing Web Sites for Mobile Users

Download Presentation

Personalization

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Personalization Speaker: Ping-Tsun Chang 3/7/2002

  2. Personalization of WWW10 • Designing Personalized Web Applications • Session: Personalization in E-Commerce • Gustavo Rossi, Daniel Schwabe, Robson Guimaraes, Dept. of Informatics, PUC-Rio , Brazil. • Personalizing Web Sites for Mobile Users • Session: Content Transformation for Mobility • Corin R. Anderson, Pedro Domingos, Daniel S. Weld, Department of Computer Science, University of Washington.

  3. Motivation • Different scenrios of personalization covering most existing applications • Object-Oriented Hypermedia Design Method (OOHDM) • Personalized Web applications by refining views according to users’ profiles or preferences

  4. Scenrios of Personalization • Link Personalization • Content Personalization • Node structure customization • Node content customization

  5. CD Date: date Order Date: date PaymentMethod Name: String Performer Text: String Comment Name: String Description: [String+photo] Keywords: {String} Price: Real Size: String Section: {Section} … DeliveryTime: string Customer Name: String Address: … CdDiscount Recommendation OOHDM: Conceptual Model • Conceptual Model for a CD store

  6. User Order CD Name: String Address: … Comment Date: date Performer Name: String CD Text: String Name: String Description: [String+photo] Keywords: {String} Price: Real Size: String Section: {Section} … DeliveryTime: string Name: String Description: [String+photo] Keywords: {String} Price: Real Size: String Section: {Section} … DeliveryTime: string OOHDM: Navigation Model • Different Navigation Schemata for different profiles includes hasComment boughtBy

  7. Hot-spots • In the conceptual model: by explicitly representing users, roles and groups and by defining algorithms that implement different (business) rules for different users. • In the navigational model: by defining completely different applications for each profile, by customizing node contents and structure and by personalizing links and indexes. • in the interface model: by defining different layouts according to user preferences or selected devices.

  8. Designing Personalized Views • Link Personalization • Content Personalization Personalizing content in a node Link personalization in OOHDM According to some data related with the user’s buying history, his category, etc. NODE Customer.CD FROM CD:c, user: Customer Name: String Price: Real [Subject.price – user C Discount ] … Comments: Anchor [Comments] Link Recommendations, user: Customer SOURCE HomePage TARGET CD:C WHERE C belongsTo user recommendations

  9. Customer Recommend Algorithm Recommendation Decoupling users from Recommendation algorithms If we want to improve the use of recommendation algorithms, we can model the assignment of differnet algorithms to different users by using strategies Recommentations() Recommender getRecomm recommender getRecomm CollaborativerFiltering getRecomm SimpleRecommend getRecomm SpecialRecommend

  10. A Link A Customer A RecommAlgorithm recommendations getRecomm A Link A Customer ThirdParty Adapter ThirdParty Recomm recommendations getRecomm recommInterface Recommendation: Implement Sequence Diagram for recommendation strategies Accommodating third party products

  11. Context Personalization Navigation Diagram of Conference Paper Review system scenrio Paper Paper by Topic by Topic Review by Author My Reviews by Reviewer Context Specification Card Reviewer by Paper

  12. Reusing Specifications Extending a Node Specification for different user profiles NODE CD FROM CD:C Name: String Price: Real NodeCustomer.CD Extends CD Description: Image Comments: Anchor [Comments] Node Manager.CD Extends CD Comments: Set Select text From Comment: Co Where C hasComment Co

  13. Goal of Personalization • A Web Personalizer can • Make frequently-visited destinations easier to find • Highlight content that interests the visitor • Elide uninteresting content and structure • A Web site personalizer adapts the site for the mobile visitor in a two-step process • The personalizer mines the access logs to build a model for each visitor • The personalizer transforms the site to maximize the expected utility for a given visitor

  14. Personalization for Mobile Users • Problem Definition • V={v0,…vm} as m indivial visitors • Vi=(R, D) a visitor is represented as his history and demographics • R=<ro,…,rt> requests ordered by time • ri=(us, ud, t, c) request is the orginating page, destination page, time, and client • D=(d0,…dn) demographic information is an n-tuple of data items • An Evaluation Function F(W, u, v)->R

  15. Web Site Model Evaluation • Expected Utility F(W, u, v) = E[Uv(p)] E[Uv(pi)] = E[Uv(si0)] • The excepted utility of a screen is the sum of its intrinsic and extrinsic utilities E[Uv(sij)] = E[IUv(sij)] + E[EUv(sij)] • Extrinsic utilities measure the value of screen by its connection to the rest of the web site E[EUv(sij)] = P(scroll)(E[Uv(si,j+1)]-rs) + ∑[P(lijk)(E[Uv(dijk)]-rl)]

  16. Intrinsic Uility • intrinsic utility of a screen as a weighted sum of two terms, which related to how the screen’s content matches the • visitor’s previously viewed content • how frequently the visitor viewed the screen. IUv(sij)] = wsjm . simV (Tij ) + wfreq . freqV (Sij ) simV (Tij ) = (wTij . wV)/(||wTij ||. ||wV||) Yahoo!

More Related