Automating Google Workspace (GWS) & more with Apps Script
Generating Resource Profiles by Exploiting the Context of Social Annotations
1. Generating Resource Profiles by Exploiting
the Context of Social Annotations
ISWC, Bonn, Germany, Oct 27th 2011
Ricardo Kawase1, George Papadakis2, Fabian Abel3
1L3S Research Center, Leibniz University Hannover, Germany
2 ICCS, National Technical University of Athens, Greece
3Web Information Systems, TU Delft
Delft
University of
Technology
2. Social Annotations Folksonomies
armstrong
baker,cool
Users
Tags
armstrong, baker, dizzy, cool, dizzy, jazz armstrong
jazzmusic, jazz, trumpet Resources
• Folksonomy:
• set of tag assignments
• Formal model [Hotho et al. „07]: tag
F = (U, T, R, Y) user resource
tag assignment
Generating Resource Profiles by Exploitingthe Context of Social Annotations 2
3. Generating resource profiles
cool Profile? Applications
concept weight
weight that operate on
?
resource profiles
baker 1 (e.g. search,
cool 2 content-based
recommender)
baker, cool
• Resource profile = representation of a resource
= set of weighted concepts
• Straightforward approach = occurrence frequency of tags:
SELECT tag, count(distinct user) FROM tas WHERE resource = XY GROUP BY tag
Applications rely on good resource profiles!
Generating Resource Profiles by Exploitingthe Context of Social Annotations 3
4. Problems of traditional folksonomies
descriptive vs.
subjective tags no tags
armstrong
baker, cool
Tags
armstrong, baker, dizzy, cool, dizzy, jazz armstrong
jazzmusic, jazz, trumpet ambiguity
synonyms of tags
Generating valuable resource profiles becomes difficult
Generating Resource Profiles by Exploitingthe Context of Social Annotations 4
5. Exploiting Context in Folksonomies
music
Jazz (noun) is a
jazz style of music that… Resource Y
created: 1979-12-06
context context context creator: …
User X
Age: 30 years
Education: … context
tag
user resource
jazz
tag assignment User X
• Context-enabled folksonomy: TAS XY
created: 2011-04-19
Fc = (U, T, R, Y,C, Z) meaning: dbpedia:Jazz
- Cis the actual metadata information
- Z Y xCis the set of context assignments
Generating Resource Profiles by Exploitingthe Context of Social Annotations 5
6. Context in Social Tagging Systems
• TagMe! – Tagging and
exploration front-end for
Flickrpictures that allows to attach
three types of contexts to tag
assignments:
• Spatial information (assign tags to
certain areas)
• Categories (= tagging of tag
assignments, e.g. buildings)
• DBpediaURIs(meaning of tag
assignments, dbpedia:Opera)
• BibSonomy – Social resource sharing system for
bookmarks and publications
• Context of resources = BibTeX information of publications
Generating Resource Profiles by Exploitingthe Context of Social Annotations 6
7. context
cool Profile?
concept weight
?
context
baker, cool
How can we exploit the context
of social annotations to
generate resource profiles?
Generating Resource Profiles by Exploitingthe Context of Social Annotations 7
8. Standard Resource Profiling
tag X Resource
u1 Profile R1
R1 concept weight
tag X 0.67
tag Y tag Y 0.33
tag X
u2 tag-based resource
profile
tag assignments performed on resource R1
• Exploiting the tags that have been directly assigned to the
resource
Generating Resource Profiles by Exploitingthe Context of Social Annotations 8
9. Context Profiling
tag A
u1 R1 Context
Profile C1
context
tag A C1 concept weight
u2 R2 tag A 0.75
tag B0.25
tag A
u3 R3 tag-based context
tag B profile
u3 R3
tag assignments in context folksonomy that
refer to context C1
• Aggregation of (context) profiles is possible ( again by
means of mixture approach)
Generating Resource Profiles by Exploitingthe Context of Social Annotations 9
10. Generating context-based resource profiles
• Generic strategy for generating context-based resource
profiles:
Context-based Resource Context
=α + (1-α)
Resource Profile Profile
Profile
concept weight concept weight concept weight
tx 0.7 tx 0.85 tx 0.55
ty 0.3 ty 0.15 ty 0.45
context
Generating Resource Profiles by Exploitingthe Context of Social Annotations 10
11. context
cool Profile?
concept weight
?
context
baker, cool
Weighting Strategies
Generating Resource Profiles by Exploitingthe Context of Social Annotations 11
12. Context-based Weighting Strategies (1)
1. User-based co-occurrence:
• Hypothesis: users tend to annotate similar resources tags a
user assigns to other resources are also relevant for the resource
profile that should be constructed
jazz Context-based
u1 Resource
Profile R1
trumpet R1 concept weight
jazz 0.67
trumpet 0.33
R2
Generating Resource Profiles by Exploitingthe Context of Social Annotations 12
13. Context-based Weighting Strategies (2)
2. Category-based co-occurrence:
• Hypothesis: resources that occur in tag assignments that are
classified in the same category are similar “tags of that
category” are also relevant for the resource
jazz Category: Context-based
u1 R1 music Resource
Profile R1
concept weight
jazz 0.67
trumpet 0.33
trumpet
u3 R3
Generating Resource Profiles by Exploitingthe Context of Social Annotations 13
14. Context-based Weighting Strategies (3)
3. Semantic Meaning – URI-based co-occurrence:
• Hypothesis: tags that have the same meaning complement the
tag-based resource profile positively
jazz Context-based
dbpedia:Jazz Resource
u1 R1
Profile R1
concept weight
jazz 0.67
jazzmusic 0.33
jazzmusic
u3 R3
Generating Resource Profiles by Exploitingthe Context of Social Annotations 14
15. Context-based Weighting Strategies (4)
4. Semantic Meaning – “binary”:
• Hypothesis: tags that can be mapped to a DBpedia resource are
more important than other tags
jazz dbpedia:Jazz Context-based
u1 R1 Resource
Profile R1
concept weight
jazz 1
cb1981 0
cb1981
?
u3 R1
Generating Resource Profiles by Exploitingthe Context of Social Annotations 15
16. Context-based Weighting Strategies (5)
5. Weighting based on Spatial context – area size:
• Hypothesis: the larger the area to which a tag is assigned to the
more important the tag for the resource
Context-based
chet baker Resource
Profile R1
R1 concept weight
jazz 0.67
trumpet 0.33
u1
trumpet R1
Generating Resource Profiles by Exploitingthe Context of Social Annotations 16
17. Context-based Weighting Strategies (6)
6. Weighting based on Spatial context – distance
from center:
• Hypothesis: the closer the (centroid of the) area to the center of
the picture the more important the tag for the resource
Context-based
tag1 Resource
Profile R5
concept weight
tag1 0.83
tag2 0.17
R5
u1 tag2
distance(tag1) <distance(tag2)
Generating Resource Profiles by Exploitingthe Context of Social Annotations 17
18. Context-based Weighting Strategies (7)
7. Journal-based co-occurrence:
• Hypothesis: tags that are assigned to publications that were
published in the same journal are also relevant for the resource
Context-based
Resource
SPARQL Journal: Web Profile R1
u1 R1 Semantics concept weight
SPARQL 0.67
semantics 0.33
semantics
u3 R3 R1
Generating Resource Profiles by Exploitingthe Context of Social Annotations 18
19. Context-based Weighting Strategies (8)
8. Journal-Year-based co-occurrence:
• Hypothesis: tags that are assigned to publications that were
published in the same journal AND in the same year are also
relevant for the resource
Context-based
year: 2007 Resource
SPARQL
u1 R1 Profile R1
concept weight
SPARQL 0.67
Journal: Web RDF store 0.33
RDF store Semantics
u3 R3
R1
trust year: 2009
u5 R16
Generating Resource Profiles by Exploitingthe Context of Social Annotations 19
20. Overview on Weighting Strategies
[TagMe!]
Based on:
Baseline: - Categories
Tag-based co-occurrence
- Spatial information
frequency
- Semantic meaning
User-based context
tag
user resource
tag assignment Resource-based:
- BibTeX properties
[BibSonomy]
Combining strategies more than 120 context-based profiling
strategies for TagMe!
Generating Resource Profiles by Exploitingthe Context of Social Annotations 20
21. Broken Slide
Generating Resource Profiles by Exploitingthe Context of Social Annotations 21
22. context
cool Profile?
concept weight
?
context
baker, cool
Which resource profiling strategy
generates the most valuable
profiles?
Generating Resource Profiles by Exploitingthe Context of Social Annotations 22
23. Experimental setup
• “Tag Prediction” task (leave-one-out cross validation):
• remove one tag from the resource
• create (context-based) resource profile
• use profile to create a ranking of tags hidden tag should be at the
top of the ranking
• Baseline:tag co-occurrence
• Metrics: Success@k = probability that the relevant tag
appear within the top k of the ranking
• Data sets: TagMe!
Tag Assignments (TAs) 1,288
TAs with Spatial Information 671 BibSonomy
TAs with Category Information 917 Resources 566,939
TAs with URI Information 1,050 Users 6,569
Tag Assignments (TAs) 2,622,423
TAs with all information 432
Generating Resource Profiles by Exploitingthe Context of Social Annotations 23
24. Results [TagMe!]
Context-based profiling strategies outperform Semantic
baseline (tag frequency) significantly. meaning and
spatial
information
allow for best
performance.
Area size more
valuable than
distance to center
no significant
difference w.r.t.
category- and user-
based strategy
Generating Resource Profiles by Exploitingthe Context of Social Annotations 24
25. Combining different types of context-
based profiling strategies
Mixture of
context-based
strategies
improve
performance (by
37%)
Context-based
strategies have
to be combined
intelligently in
order to increase
cumulative gain
in performance.
Generating Resource Profiles by Exploitingthe Context of Social Annotations 25
26. Results [BibSonomy]
Again: Context-based profiling strategies
outperform baseline (tag frequency) significantly.
The more
specific the
context, the
better the
performance
( reducing
noise)
Generating Resource Profiles by Exploitingthe Context of Social Annotations 26
27. Conclusions
• What we did: framework for generating resource profiles
by exploiting contextual information of social annotations
• Context-based folksonomy model
• Set of context-based resource profiling strategies (both generic and
application-specific strategies)
• Evaluation in two social tagging systems: TagMe! and BibSonomy
• Results:
• Context-based strategies outperform other strategies that do not
exploit contextual information
• Context of tag assignments (e.g. semantic meaning) allows for best performance
• Context of the user who performs the tag assignment is competitive
• Mixing context-based strategies improves quality but does not
necessarily result in a cumulative gain in performance (“over-
contextualization”) smart mixing performs best (>40% improvement)
Generating Resource Profiles by Exploitingthe Context of Social Annotations 27