Download presentation

Presentation is loading. Please wait.

Published byAmani Dinsdale Modified over 2 years ago

1
A Hypertext Metric Based on Huffman Coding Chris CoulstonTheresa M. Vitolo Penn State ErieGannon University Electrical and Computer EngineeringComputer and Information Science

2
Motivation Relationship between –Foraging navigation patterns –Outcomes measures Metric –Correlate –Semantics

3
Prior Work Botafogo, Rivlin and Shneiderman –Compactness and Stratum Pirolli, Pitkow, Rao –High level regions of interest (Xerox web site) McEneaney –Establishes relationship –Small range of metric values –Semantics of metric

4
Huffman Code Given:Fixed message –Symbols –Frequencies Find:Binary encoding of symbols –Minimize total number of bits in message –Huffman tree –Bits per symbol

5
Example Message –a,a,a,a,a,a,b,b,b,c,c,d Huffman Tree SymbolABCD Frequency6321 SymbolABCD Code010110111

6
Transformation User behavior viewed as decoding process Input –HT topology –User path / Node and link frequencies Output –Bits per symbol –Binary decisions to get to information in the context of the entire hypertext

7
Example HT topology

8
Example User path

9
Example User Path –BFS –Frequency

10
Example User 3.82 BPS

11
Example Optimum 2.89 BPS

12
Example RatioR=2.89/3.82 = 0.78 R in (0,1] –R=1optimal navigation –R 0inefficient navigation

13
Conclusions/Future Work Semantic basis for metric Analyze McEneaney data –Create software tools –Correlate user success with Huffman metric Framework for “hunting” –Collaboration with McEneaney –Hypertext ’02

Similar presentations

© 2017 SlidePlayer.com Inc.

All rights reserved.

Ads by Google