pragmatic evaluation of folksonomies

19

Click here to load reader

Upload: markus-strohmaier

Post on 08-May-2015

1.375 views

Category:

Technology


0 download

TRANSCRIPT

Page 1: Pragmatic evaluation of folksonomies

Knowledge Management Institute

Pragmatic Evaluation of Folksonomies

20th International World Wide Web Conference (WWW2011)Hyderabad, India

D. Helic, M. Strohmaier, C. Trattner, M. Muhr, K. Lerman

Markus StrohmaierAssistant Professor, Graz University of Technology, Austria

Visiting Scientist, (XEROX) PARC, USA

1

Markus Strohmaier 2011

Page 2: Pragmatic evaluation of folksonomies

Knowledge Management Institute

Taxonomies: Categorization by Experts

Taxonomy: Usually produced and maintained byfew (e g dozens of) domain expertsfew (e.g. dozens of) domain experts.

Alternative: Folk-generated taxonomies( F lk i “)(„Folksonomies“)

But how useful are such hierarchical structures? How can they be evaluated?

2

Markus Strohmaier 2011

Page 3: Pragmatic evaluation of folksonomies

Knowledge Management Institute

Outline of this talk

1. FolksonomiesC t ti d E l tiConstruction and Evaluation

2 Decentralized Search2. Decentralized SearchJ. Kleinberg‘s algorithm

3. Pragmatic Evaluation FrameworkPresentation and discussion

4. Results & Findings

3

Markus Strohmaier 2011

Page 4: Pragmatic evaluation of folksonomies

Knowledge Management Institute

Outline of this talk

1. FolksonomiesC t ti d E l tiConstruction and Evaluation

2 Decentralized Search2. Decentralized SearchJ. Kleinberg‘s algorithm

3. Pragmatic Evaluation FrameworkPresentation and discussion

4. Results & Findings

4

Markus Strohmaier 2011

Page 5: Pragmatic evaluation of folksonomies

Knowledge Management Institute

Tagging: Social classification by users

ResourcesUsers label and categorize

resources with concepts (tags)

U

Tags

is a tuple V:= (U, T, R, Y) whereth th di j i t fi it t U T R d t user 1

Users

• the three disjoint, finite sets U, T, R correspond to– a set of persons or users u ∈ U – a set of tags t ∈ T and

user 1

– a set of resources or objects r ∈ R

• Y ⊆ U ×T ×R, called set of tag assignmentstag 1 res. 1

Tag similarity based on

5

Markus Strohmaier 2011

users and resources

Page 6: Pragmatic evaluation of folksonomies

Knowledge Management Institute

Construction of FolksonomiesF t t lit t t litFrom tag centrality to tag generality:high tag centrality:

more abstract

low tag centrality:more specific

Other existing folksonomy algorithms: k-means, affinity propagation, …

6

Markus Strohmaier 2011

[Heyman and Garcia-Molina 2006]

Page 7: Pragmatic evaluation of folksonomies

Knowledge Management Institute

Semantic Evaluation of FolksonomiesSemantic Evaluation of Folksonomies

Emerging Hierarchy Expert Hierarchyg g y(Emergent)via e.g. hierarchical clustering

p y(Golden Standard)WordNet: a lexical DB for English

computers

programmingProgramming

Map-ping

Synset Hierarchy

distance d = 1 distance

Design Python

languages

distance d1 = 1 distance d2 = 2

gpatterns

g g

java python

Semanticgrounding

abs. difference |d1 - d2| a simple proxy for the quality

8

Markus Strohmaier 2011

j pythonp p y q yof emergent semantics

Page 8: Pragmatic evaluation of folksonomies

Knowledge Management Institute

Outline of this talk

1. FolksonomiesC t ti d E l tiConstruction and Evaluation

2 Decentralized Search2. Decentralized SearchJ. Kleinberg‘s algorithm

3. Pragmatic Evaluation FrameworkPresentation and discussion

4. Results & Findings

9

Markus Strohmaier 2011

Page 9: Pragmatic evaluation of folksonomies

Knowledge Management Institute

Decentralized SearchDecentralized Search

Background knowledge:

Idea: use folksonomies as background knowledge Then, the performance of decentralized search

Shortest path to targetBackground knowledge:(a tag hierarchy)

g gpdepends on the suitability of folksonomies.

In other words, we can evaluate the suitability of

Folksonomy1

Folksonomy...

Folksonomyn

folksonomies for decentralized search through simulations.

shortest path found with l l k l d 4A (tag-tag) network:

Goal: Navigate from START to TARGETusing local and background knowledge

local knowledge pLK = 4

Δ = pLK-pGK

start target

using local and background knowledge only

shortest path with candidates

10

Markus Strohmaier 2011J. Kleinberg. The small-world phenomenon: An algorithmic perspective. Proc. 32nd ACM Symposium on Theory of Computing, 2000. Also appears as Cornell Computer Science Technical Report 99-1776 (October 1999)

pglobal knowledge pGK = 3

Page 10: Pragmatic evaluation of folksonomies

Knowledge Management Institute

Outline of this talk

1. FolksonomiesC t ti d E l tiConstruction and Evaluation

2 Decentralized Search2. Decentralized SearchJ. Kleinberg‘s algorithm

3. Pragmatic Evaluation FrameworkPresentation and discussion

4. Results & Findings

11

Markus Strohmaier 2011

Page 11: Pragmatic evaluation of folksonomies

Knowledge Management Institute

Pragmatic Evaluation Framework

General idea:• Use the OUTPUT produced by folksonomy algorithms

(hi hi l t t ) INPUT (b k d(hierachical structures) as INPUT (background knowledge) for decentralized search.

Framework Instantiation1. Generate n folksonomies K-means, Aff.Prop.,

DegCentrality, CloCentrality

2. Model navigational task exploratory navigation

3. Select evaluation metrics success rate, stretch

4 Sim late na igation decentralized search4. Simulate navigation decentralized search

5. Evaluate performance comparative evaluation

12

Markus Strohmaier 2011

Page 12: Pragmatic evaluation of folksonomies

Knowledge Management Institute

Simulating Exploratory NavigationTopically

tags

related tagsSTART TARGET

resources

Random R d

Topically related

resourcesRandom start

page: e.g. landing

page from

Random resource

resources

Usefulness of:

page from search engine

F lk F lk F lk

We generate 100.000 search pairs (start, target) for each dataset, and run simulations

13

Markus Strohmaier 2011

Folksonomy1

Folksonomy...

Folksonomyn

run simulations

Page 13: Pragmatic evaluation of folksonomies

Knowledge Management Institute

Outline of this talk

1. FolksonomiesC t ti d E l tiConstruction and Evaluation

2 Decentralized Search2. Decentralized SearchJ. Kleinberg‘s algorithm

3. Pragmatic Evaluation FrameworkPresentation and discussion

4. Results & Findings

14

Markus Strohmaier 2011

Page 14: Pragmatic evaluation of folksonomies

Knowledge Management Institute

Success Rates Across Different FolksonomiesTag generalityflickr dataset

k-means /

Tag generality approaches

Random

affinity propagation

Random folksonomy

Success rate: The number of times an agent is successful

All approaches outperform a random folksonomy

The number of times an agent is successful in finding a path using a particular folksonomy as background knowledge

y

Tag generality approaches outperform k-means / Aff. Propagation

max hops n: the maximal number of steps an agent is allowed to perform before stopping (a tunable

t t l f ll li k )

n

16

Markus Strohmaier 2011

Propagationparameter e.g., an agent only follows n links).

Page 15: Pragmatic evaluation of folksonomies

Knowledge Management Institute

Success Rates Across Different Datasets

Holds for all datasets(to diff

But how efficient are

those(to diff. extents)

those folksonomies

during search?

17

Markus Strohmaier 2011

search?

Page 16: Pragmatic evaluation of folksonomies

Knowledge Management Institute

Stretch Δ = p pStretch Δ = pLK-pGKShortest Paths found with Local Knowledge

Bib K M

Finds no path: Δ = infinite

Bibsonomy K-Means

Δ infiniteFinds paths that is +1 longer:Δ = 1

T litHolds for all

d t t Finds shortest possible path:Δ = 0

Tag generality approaches (d+e) find much shorter

paths!

datasets(to diff. extents)

paths!

18

Markus Strohmaier 2011

Page 17: Pragmatic evaluation of folksonomies

Knowledge Management Institute

Pragmatic Evaluation Framework

Framework Instantiation Alternatives1. Generate n folksonomies K-means, Aff.Prop.,

DegCentrality, other folksonomy

algorithms orCloCentrality expert taxonomies

2. Model navigational task exploratorynavigation

other tasks

3 Select evaluation metrics success rate, stretch other evaluation metrics3. Select evaluation metrics4. Simulate navigation decentralized search actual click data

5. Evaluate performance comparativeevaluation

other evaluationapproachesevaluation approaches

Pragmatic evaluation produces different results for different tasks and different assumed or observed navigation behavior.

The evaluation framework is completely general with regard to

19

Markus Strohmaier 2011

the task, data and evaluation metrics adopted.

Page 18: Pragmatic evaluation of folksonomies

Knowledge Management Institute

Results & Findings: Summary

1 F lk i f l b k d k l d f1. Folksonomies are useful background knowledge for navigation.

2. Existing folksonomy algorithms are more useful than a random baselinethan a random baseline.

3 Tag generality approaches perform better than3. Tag generality approaches perform better than existing hierarchical clustering approaches.

4. Pragmatic results support theoretical analysis (not presented in talk – included in paper).

20

Markus Strohmaier 2011

(not presented in talk included in paper).

Page 19: Pragmatic evaluation of folksonomies

Knowledge Management Institute

Th k YThank You.

Markus [email protected]

D. Helic, M. Strohmaier, C. Trattner, M. Muhr, K. LermanPragmatic Evaluation of Folksonomies

20th International World Wide Web Conference (WWW2011)Hyderabad, India, March 28 - April 1, ACM, 2011.

http://kmi.tugraz.at/staff/markus/documents/2011_WWW2011_Folksonomies.pdf

21

Markus Strohmaier 2011