visualization of music suggestions - status update
DESCRIPTION
Status update on the thesis visualization of music suggestions. For more information consult: http://soundsuggest.wordpress.comTRANSCRIPT
Visualization of music suggestions
Promotor: Erik Duval, Joris KlerkxSupervisor: Joris Klerkx
KULeuven - Visualization of music suggestions - Joris Schelfaut
2
Timeline
October: problem definition
November: literature study
December: prototype
January: evaluate and implement
...
KULeuven - Visualization of music suggestions - Joris Schelfaut
3
1 Problem definition
• The Herlocker paper: “Explaining Collaborative Filtering Recommendations”– Recommender systems– Collaborative filtering
Black boxExplanation system (white box)
KULeuven - Visualization of music suggestions - Joris Schelfaut
4
2 Literature study
• Many examples of (visual) explanation systems
KULeuven - Visualization of music suggestions - Joris Schelfaut
5
3 PrototypeFirst attempt at a paper prototype
Idea: items that occur together in one profile share a link, i.e. the profile
KULeuven - Visualization of music suggestions - Joris Schelfaut
6
4 Implementation
Experimenting with chrome extensions and D3.js
Visualization ofUser profiles
KULeuven - Visualization of music suggestions - Joris Schelfaut
7
Problem
• No real background on fundamental concepts Additional literature required
KULeuven - Visualization of music suggestions - Joris Schelfaut
8
Timeline
February: literature study part 2
March: re-evaluating and testing the prototype (paper / digital)
April: completing and evaluating implemented application
May: finishing thesis text
KULeuven - Visualization of music suggestions - Joris Schelfaut
9
5 Literature study part 2
• (Recommender systems)• Insight gaining• Visualization– Principles / techniques– Graph based
• (Evaluation techniques)
KULeuven - Visualization of music suggestions - Joris Schelfaut
10
6 Prototype > paper
Template used for theprototype screens
KULeuven - Visualization of music suggestions - Joris Schelfaut
11
6 Prototype > paper evaluation
• Main questions:1. Can a user figure out what the visualization
means?2. Can a user use the visualization to make
estimates about the reliability of suggestions?Tasks + think aloud protocol• SUS test
KULeuven - Visualization of music suggestions - Joris Schelfaut
12
6 Prototype > paper results
• Short answers:1. Yes;2. Yes.
• SUS:– 5 test users– 77 avg from 65, 85, 85, 75, 75– Would you use it? 2 say no, 3 say yes
• Remarks– Textual help
KULeuven - Visualization of music suggestions - Joris Schelfaut
13
6 Prototype > digital
Visualization D3.jsusing a static JSON as input
KULeuven - Visualization of music suggestions - Joris Schelfaut
14
7 Implementation
Last.fm/API Servlet / google appengine
JavaScript / SVGChrome extension
JSONJSON
KULeuven - Visualization of music suggestions - Joris Schelfaut
15
Info / References
Blog:http://soundsuggest.wordpress.com
GitHub:https://github.com/soundsuggest/