11
Dynamic Bookmarks for WWW Managing Personal Navigation Space by Analysis of Link Structure and User Behavior Hajime Takano Stanford University / NEC Corporation Terry Winograd Stanford University

Hajime Takano Stanford University / NEC Corporation Terry Winograd Stanford University

Embed Size (px)

DESCRIPTION

Dynamic Bookmarks for WWW Managing Personal Navigation Space by Analysis of Link Structure and User Behavior. Hajime Takano Stanford University / NEC Corporation Terry Winograd Stanford University. Problems. WWW Space. Bookmark Management is tiresome. forget to add useful URLs. - PowerPoint PPT Presentation

Citation preview

Page 1: Hajime Takano Stanford University / NEC Corporation Terry Winograd Stanford University

Dynamic Bookmarks for WWWManaging Personal Navigation Space

by Analysis of Link Structure and User Behavior

Hajime TakanoStanford University / NEC Corporation

Terry WinogradStanford University

Page 2: Hajime Takano Stanford University / NEC Corporation Terry Winograd Stanford University

Problems

Bookmark Management is tiresome. forget to add useful URLs. Useful URLs are varying. Useless URLs are remaining.

Users want to focus on navigation itself. Decision to add or not Evaluation for Social Filtering

Users should explore again around a bookmarked URL.

Bookmark

Personal Info. Space

Mailing list, E-News, Web Magazine, etc…

WWW Space

Page 3: Hajime Takano Stanford University / NEC Corporation Terry Winograd Stanford University

Requirements

Organizing History Well-organized history helps users to revisit sites/pages.

Evaluating Site/Page Importance Good sites/pages should be prominent.

Automatically Creating & Managing Navi. History Less efforts are expected to manage bookmarks.

Page 4: Hajime Takano Stanford University / NEC Corporation Terry Winograd Stanford University

WDB Bookmark Viewer

Page 5: Hajime Takano Stanford University / NEC Corporation Terry Winograd Stanford University

Organizing History

Meta Page Sources to sites ex) Search results, link lists, etc.

Home Page Front page of a sites ex) http://www.stanford.edu/

Landmark Page Useful pages in a site ex) http://www.stanford.edu

/home/academics/schools.html

Yahoo HotBot Link page

www.cs.stanford

3-Layer Management

Page 6: Hajime Takano Stanford University / NEC Corporation Terry Winograd Stanford University

Home

www.Stanford.EDU

Home

www-cs.Stanford.EDU

Root

Reference links

Segmentation Strategies

Clustering by Hostname / directory

Finding Home Page filename: ex) index.html directory depth: ex) /htakano/

Finding links among clusters Reference Links

Page 7: Hajime Takano Stanford University / NEC Corporation Terry Winograd Stanford University

The Second-order Connectedness

in Sougata, Hypertext97 paper

The number of nodes can be reached by following at most two links.

Income links + Outgo links

Second-Order Connectedness

Back Second-Order Connectedness

• shows a chance to reach more info.

• shows a reliability referred by more info.

Page 8: Hajime Takano Stanford University / NEC Corporation Terry Winograd Stanford University

Consideration on User Behavior

Emphasizing

Number of Visited Pages

Number of Following Links add visit count on a link

Reducing

Number of Unvisited links useless for a user use Logarithm scale

Expiring old behaviors

Page 9: Hajime Takano Stanford University / NEC Corporation Terry Winograd Stanford University

Algorithm - equations

)(0

i

N

kiki ULfuncVLFOC

)(0

)0(0

)0(

N

kVLk

N

kVLki FOCfuncFOCSOC

)(0

i

N

kkii UBfuncVLBFOC

iiiiii VCwBSOCSOCwBFOCFOCwPR 321 )()(

ki PRSR

)(0

)0(0

)0(

N

kVLk

N

kVLki BFOCfuncBFOCBSOC

Page 10: Hajime Takano Stanford University / NEC Corporation Terry Winograd Stanford University

Automatic Archiving

Implemented as “Intelligent” Proxy Server Tracking Every HTTP Access Following “Referrer: “ field Analyzing HTML Archiving into URL DB Analyzing data in URL DB

Implementation Java 1.1.5 ObjectStore PSE for Java

WWW Browser

WWW Space

Access T

racking

HTMLAnalysis

URL Database

ImportanceCalculator

ViewerEngine

BookmarkViewer

WDB

Page 11: Hajime Takano Stanford University / NEC Corporation Terry Winograd Stanford University

Future Work

Evaluation How well this algorithm evaluate page/site importance ? Does WDB really help users’ navigation task?

Further Researches Shared Bookmark Architecture Integration with other techniques:

Content Analysis, Classification, etc.Hierarchical Visualization.