Talk give by Ke Tao at #ICWE2012 titled Leveraging User Modeling on the Social Web with Linked Data

1. Leveraging User Modeling on the Social
Web with Linked Data
#ICWE2012 Berlin, Germany, July 27th 2012
Fabian Abel, Claudia Hauff, Geert-Jan Houben, Ke Tao
Web Information Systems, TU Delft, the Netherlands
Delft
University of
Technology

2. What we do: Science and Engineering
for the Personal Web
domains: news social media cultural heritage public data e-learning
Personalized Personalized
Adaptive Systems
Recommendations Search
recommending
points of interest Analysis and
User Modeling
Semantic Enrichment,
Linkage and Alignment
user/usage data
Social Web
Leveraging User Modeling on the Social Web with Linked Data 2

3. Kyle
Hometown: South
Park, Colorado
Leveraging User Modeling on the Social Web with Linked Data 3

4. Kyle recently uploaded photos to Flickr
tags: delft, vermeer with
tags: girl
geo: The Hague earring
pearl
geo: The Hague
During his trip to the Netherlands,
he uploaded pictures to Flickr.
Leveraging User Modeling on the Social Web with Linked Data 4

5. Kyle tweets about his upcoming trip
Looking forward to visit
Paris next week!
Leveraging User Modeling on the Social Web with Linked Data 5

6. Interests of Kyle?
• Given Kyle’s Flickr and Twitter
activities, can we infer Kyle’s
interests? tags: delft, vermeer with
tags: girl
geo: The Hague earring
pearl
• Knowing that Kyle will visit Paris, geo: The Hague
France, can we recommend him
places that might be interesting
for him?
Looking forward to visit
Paris next week!
Leveraging User Modeling on the Social Web with Linked Data 6

7. Challenges
• How to create a meaningful profile that supports the given
application?
à how to bridge between the Social Web chatter of a user
and the candidate items of a recommender system?
c9
User
Profile
c6
Applica0on
c1
concept
weight
c5
that
demands
user
interest
profile
regarding
?
c1
-­‐concepts
c3
d4
d3
d1
d5
Social Web c2
d2
d3
d4
Candidate Items Recommendations
Leveraging User Modeling on the Social Web with Linked Data 7

8. Challenge of Recommending Points of
Interests (POIs) to Kyle
The astronomer
Johannes Vermeer The la
cema
ker
Looking forward to dbpedia:Louvre
visit Paris next week!
dbpedia:Paris
Leveraging User Modeling on the Social Web with Linked Data 8

9. LOD-based User Modeling
concepts
that
can
be
extracted
User
Profile
from
the
user
data
weigh0ng
strategies
concept
weight
c2
c1
c4
c9
0.4
c6
c2
c1
cy
0.1
c5
c3
0.2
cx
…
…
c3
Applica0on
background
knowledge
that
demands
user
user
data
(graph
structures)
interest
profile
regarding
Social
Web
-­‐concepts
Linked
Data
Leveraging User Modeling on the Social Web with Linked Data 9

10. User Modeling Building Blocks
Leveraging User Modeling on the Social Web with Linked Data 10

11. Strategies for exploiting the RDF-based
background knowledge graph
Indirect Mention
Indirect Mention
Direct Mention
@RDF_statement
@RDF_graph
dbpedia:Louvre
dbpp
rop:
tags: girl with lo catio
pearl earring n
geo: The Hague
dbpedia:The_Hague A
B
C
…
rd f:type Artifact
dbpedia:Girl_with_pearl_earring foaf:maker
The The
foaf:maker lacemaker astronomer
Johannes Vermeer
Leveraging User Modeling on the Social Web with Linked Data 11

12. Weighting Scheme
User
Profile
concept
weight
c1
374
?
c2
152
?
c3
73
?
…
…
Weighting scheme: count the
number of occurrences of a given
graph pattern.
Leveraging User Modeling on the Social Web with Linked Data 12

13. Source of User Data
• Twitter
• Flickr
Leveraging User Modeling on the Social Web with Linked Data 13

14. Mining the Geographic Origins of User Data
• Tweets or Flickr images posted by a GPS-enabled device;
• Images geo-tagged manually on Flickr world map
• Otherwise : exploit the title and the tags the users assign
to their images
Leveraging User Modeling on the Social Web with Linked Data 14

15. Evaluation
Leveraging User Modeling on the Social Web with Linked Data 15

16. Research Questions
1. How does the source of user data influence the quality in
deducing user preferences for POIs?
2. How does the consideration of background knowledge from
the Linked Open Data Cloud impact the quality of the user
modeling?
3. What (combination of) user modeling strategies allows for
the best quality?
Leveraging User Modeling on the Social Web with Linked Data 16

17. Dataset
users 394
duration 11 months
tweets 2,489,088
location 11%
pictures 833,441
location 70.6%(within 10km)
Leveraging User Modeling on the Social Web with Linked Data 17

18. Dataset Characteristics: Profile Sizes
ratio between Flickr-based profile size
and Twitter-based profile size 100
size of Flickr-based profile is greater than
10 size of Twitter-based profile
81.4% of the users,
1 Twitter-based profiles are
bigger than Flickr-based profiles
0.1
size of Twitter-based profile is greater than
0.01
size of Flickr-based profile Flickr-based
profile is
0.001 empty
0
0% 20% 40% 60% 80% 100%
user percentiles
Leveraging User Modeling on the Social Web with Linked Data 18

19. Experimental Setup
• Task:
= Recommending POIs
= Predicting POIs which a user will visit
• Ground truth:
• split data into training data (= first nine months) and test data (= last
two months)
• POIs that the user visited in the last two months are considered as
relevant
• Metrics:
• Precision@k, Recall@k and F-Measure@k: precision, recall and f-
measure within the top k of the ranking of recommended items
Leveraging User Modeling on the Social Web with Linked Data 19

20. Results: Impact of User Data Source
Selection
Combination of Twitter
!#+"
and Flickr user data
!"#$%&%'()*+#$,--*,(.*/01#,&2"#*
!#*" allows for best
!#)" performance
!#("
!#'" 78$!"
!#&" 98$!"
!#%"
,8$!"
!#$"
!"
,-./01" 23.451" ,-./01"6"
23.451"
Twitter seems to be
more valuable for the
given application
Leveraging User Modeling on the Social Web with Linked Data 20

21. Results: Impact of Strategies for Exploiting
RDF-based Background Knowledge
The more background
!#," information the better
!"#$%&%'()*+#$,--*,(.*/01#,&2"#*
!#+"
the user modeling
!#*"
!#)" performance
!#("
89$!"
!#'"
!#&" :9$!"
!#%" ;9$!"
!#$"
!"
-./012" .4-./012" .4-./012"
304564" 304564"7" 304564"77"
Leveraging User Modeling on the Social Web with Linked Data 21

22. Leveraging User Modeling on the Social Web with Linked Data
Leveraging User Modeling on the Social Web with Linked Data 13
13
Results: Combining different Strategies
Table 2. Overview on the di erent strategies for integrating background knowledge.
Table 2. Overview on the di erent strategies for integrating background knowledge.
Twitter and Flickr are exploited as user data source.
Twitter and Flickr are exploited as user data source.
Leveraging User Modeling on the Social Web with Linked Data 13
strategy
strategy P@10 R@10 F@10
P@10 R@10 F@10 P@20 R@20 F@20
P@20 R@20 F@20
Table 2. Overview on the di erent strategies for integrating background knowledge.
core strategies:
core strategies:
Twitter and Flickr are exploited as user data source.
direct mentions
direct mentions 0.715 0.260 0.412
0.715 0.260 0.412 0.580 0.179 0.298
0.580 0.179 0.298
strategy
indirect mentions I P@10 R@10 F@10
0.699 0.268 0.426 P@20 R@20 F@20
0.566 0.185 0.308
indirect mentions I 0.699 0.268 0.426 0.566 0.185 0.308
indirect mentions II core strategies: 0.475
0.820 0.312 0.727 0.436 0.569
indirect mentions II 0.820 0.312 0.475 0.727 0.436 0.569
direct mentions 0.715 0.260 0.412
combined strategies: 0.580 0.179 0.298
combined strategies:
indirect mentions mentions I
direct & indirect I 0.699 0.268 0.426
0.733 0.216 0.360 0.566 0.185 0.308
0.608 0.287 0.416
indirect mentions mentions II
direct & indirect II 0.820 0.312 0.475
0.836 0.333 0.4975 0.727 0.436 0.569
0.747 0.466 0.596
combined strategies:
indirect mentions I + II 0.830 0.325 0.489 0.739 0.456 0.587
direct & indirect mentions I 0.733 0.216 0.360 0.608 0.287 0.416
direct & indirect mentions I + II 0.839 0.337 0.502 0.751 0.473 0.603
direct & indirect mentions II 0.836 0.333 0.4975 0.747 0.466 0.596
indirect mentions I + II 0.830 0.325 0.489 0.739 0.456 0.587
direct & indirect mentions RDF statements from the Linked Open Data Cloud,
which does not exploit I + II 0.839 0.337 0.502 0.751 0.473 0.603
the F@20 performance is more than doubled. The consideration of background
knowledge obtained from the Linked Open Data Cloud therefore clearly improves
Combining all background
the performance of user modeling on the Social Web which answers our main
which does not exploit RDF statements from the Linked Open Data Cloud,
research question.
exploitation strategies improves the
the F@20 performance is more than doubled. The consideration of background
Furthermore, we can answer the specific research questions raised at the
knowledge obtained from the Linked Open Data Cloud therefore clearly improves
beginning of this section as follows. For the task of recommending points of
user modeling performance clearly
the performance of user modeling on the Social Web which answers our main
interests, it turns out that (1) the aggregation of Twitter and Flickr user data
research question.
allows for the best user modeling performance and that (2) the user modeling
Furthermore, we can answer the specific research questions raised at the
quality increases the more background information we consider from the Linked
beginning of this section as follows. For the task of recommending points of
Open Data Cloud. Finally, (3) the best performance is achieved by combining
interests, it turns out that (1) the aggregation of Twitter and Flickr user data
the di erent graph patterns for acquiring background information and inferring
allows for the best user modeling performance and that (2) the user modeling
user preferences.
Leveraging User Modeling on the Social Web with Linked Data 22
quality increases the more background information we consider from the Linked
Open Data Cloud. Finally, (3) the best performance is achieved by combining
6 Conclusions

23. Conclusions
What we did:
• LOD-based User Modeling on the Social Web
• Different strategies for exploiting RDF-based background
knowledge
Findings:
1. Combination of different user data sources (Flickr & Twitter) is
beneficial for the user modeling performance
2. User modeling quality increases the more background
knowledge one considers
3. Combination of strategies achieves the best performance
Future work:
• Investigate weighting schemes that weight the different RDF
graph patterns for acquiring background knowledge differently
Leveraging User Modeling on the Social Web with Linked Data 23

24. Thank you!
Email: K.Tao@tudelft.nl
Twitter: @wisdelft @taubau
http://persweb.org
Leveraging User Modeling on the Social Web with Linked Data 24