1. CINEMAPPY: A CONTEXT-AWARE MOBILE APP FOR
MOVIE RECOMMENDATIONS BOOSTED BY DBPEDIA
Vito Claudio Ostuni1, Tommaso Di Noia1, Roberto Mirizzi2,
Davide Romito1, Eugenio Di Sciascio1
ostuni@deemail.poliba.it, t.dinoia@poliba.it, roberto.mirizzi@hp.com, romito@deemail.poliba.it, disciascio@poliba.it
1Politecnico di Bari 2HP Labs
Via Orabona, 4 1501 Page Mill Road
70125 Bari (ITALY) Palo Alto, CA (US) 94304
SeRSy 2012 – International Workshop on Semantic Technologies
meet Recommender Systems & Big Data
ISWC 2012 November 11, 2012 Boston, USA

2. Outline
 What are Recommender Systems?
 Collaborative filtering & Content-based algorithms
 Taking into account user’s context: Context Aware RSs
 Why should we use LOD to feed them?
 Cinemappy: a mobile context-aware and content-based
Recommender System for movie-cinema bundles
 Architecture
 Contextual factors
 Recommender Engine
 Conclusion
SeRSy 2012 – International Workshop on Semantic Technologies
meet Recommender Systems & Big Data
ISWC 2012 November 11, 2012 Boston, USA

3. Recommender Systems
A definition
Recommender Systems (RSs) are software tools and techniques
providing suggestions for items to be of use to a user.
[F. Ricci, L. Rokach, B. Shapira, and P. B. Kantor, editors. Recommender Systems Handbook. Springer, 2011.]
Input Data:
A set of users U={u1, …, uN}
A set of items I={i1, …, iM}
The rating matrix R=[ru,i]
Problem Definition:
Given user u and target item i
Predict the rating ru,i
SeRSy 2012 – International Workshop on Semantic Technologies
meet Recommender Systems & Big Data
ISWC 2012 November 11, 2012 Boston, USA

4. Two main approaches
Collaborative filtering
The problem of collaborative filtering is to predict how well a user will like an item
that he has not rated yet, given a set of historical preference judgments for a
community of users.
5
?
current user
?
1 3
4
5
SeRSy 2012 – International Workshop on Semantic Technologies
meet Recommender Systems & Big Data
ISWC 2012 November 11, 2012 Boston, USA

5. Two main approaches
Content-based
Content-based RSs recommend items to a user based on their description and on
the profile of the user’s interests
Movie’s attributes
1
• Cast
4 • Genres
• Director
….
SeRSy 2012 – International Workshop on Semantic Technologies
meet Recommender Systems & Big Data
ISWC 2012 November 11, 2012 Boston, USA

6. Taking into account user’s context:
Context Aware RSs
What do we mean by context?
- any information that can be used to characterize the situation of an entity
- any information or conditions that can influence the user perception
- any information that can characterize the interaction between a user and
the application
External factors:
Time, weather, season,….
User factors:
Mood, activity, company, location,..
SeRSy 2012 – International Workshop on Semantic Technologies
meet Recommender Systems & Big Data
ISWC 2012 November 11, 2012 Boston, USA

7. Using Linked Data in Recommender Systems
Content Based
Mitigation of Limited Content Analysis (huge amount of available knowledge)
No Content Analyzer required (structured data available)
Collaborative filtering
Mitigation of cold start via hybridization
Cross domain recommendations
Knowlegeable explanations
SeRSy 2012 – International Workshop on Semantic Technologies
meet Recommender Systems & Big Data
ISWC 2012 November 11, 2012 Boston, USA

8. Cinemappy: a mobile context-aware and
content-based RS of movie-cinema bundles
Suggest to the user movies to be watched in theaters taking into
account user preferences and user context (position, companion,..)
Like/dislike
Explicit
Company
Context
Movie, Cinema bundle
SeRSy 2012 – International Workshop on Semantic Technologies
meet Recommender Systems & Big Data
ISWC 2012 November 11, 2012 Boston, USA

9. Cinemappy Architecture
SeRSy 2012 – International Workshop on Semantic Technologies
meet Recommender Systems & Big Data
ISWC 2012 November 11, 2012 Boston, USA

10. Contextual factors
Explicit
•Companion : family, friends, partner, by myself is better, coworker, none
Implicit
•Time : all the movies scheduled before the current time, plus the time to get to
the theatre, have to be discarded
•Geographical relevance : depending on the current location of the user, the
system should be able to suggest movies to watch in cinemas
close/relevant to them and discard the farther ones even if they may
result more appealing with respect to the user preferences.
Some criteria of geographical relevance:
•Cluster: a multiplex could be more interesting than a normal cinema
•Co-location: a cinema close to a pub could be more useful
•Anchor point proximity: a cinema close home could be more easy to get to
SeRSy 2012 – International Workshop on Semantic Technologies
meet Recommender Systems & Big Data
ISWC 2012 November 11, 2012 Boston, USA

11. Recommender Engine
Contextual Pre-filtering
Content-based
Contextual Post-filtering
RECOMMENDER
Pre-filtering
CB RS
Post-filtering
SeRSy 2012 – International Workshop on Semantic Technologies
meet Recommender Systems & Big Data
ISWC 2012 November 11, 2012 Boston, USA

12. Contextual Pre-filtering
Time and distance contextual factors
For each user u, the set of movies Mu (set of movies that the system can
recommend to u) is defined as containing the movies scheduled in the next d
days in theaters in a range of k kilometers around the user position. The system
can recommend only movies in Mu.
Companion contextual factor
Regarding this context we use the micro-profiling approach. To the user u is
associated a set of different profiles each one related to a specific value cmp of
the companion context.
 
profile (u , cmp )  m j ,v j  v j = 1 if u likes m j with companion cmp, v j = -1 otherwise 
 
SeRSy 2012 – International Workshop on Semantic Technologies
meet Recommender Systems & Big Data
ISWC 2012 November 11, 2012 Boston, USA

13. Content-based Recommendations
with DBpedia (i)
We predict the rating using a Nearest Neighbor Classifier
 v j  sim(m j , mi ) ??
m j  profile ( u ,cmp )
rPreF (ucmp , mi ) 
profile(u, cmp )
But…Our movies are RDF resources…
how to compute similarities between RDF resources?
SeRSy 2012 – International Workshop on Semantic Technologies
meet Recommender Systems & Big Data
ISWC 2012 November 11, 2012 Boston, USA

14. Computing similarity in DBpedia
SeRSy 2012 – International Workshop on Semantic Technologies
meet Recommender Systems & Big Data
ISWC 2012 November 11, 2012 Boston, USA

15. Vector Space Model for DBpedia (i)
Righteous Kill
Heat
Righteous Kill
Al Pacino
Robert De Niro
Brian Dennehy
Heat
Robert De Niro starring
Al Pacino
Brian Dennehy
John Avnet
Serial killer films
Heist films
Crime films genre
subject/broader
Drama director
starring
Crime films
Heat
Brian Dennehy
Drama
John Avnet
Righteous Kill
Heist films
Robert De Niro
Al Pacino
Serial killer films
T.Di Noia, R. Mirizzi, V. C. Ostuni, D. Romito, and M. Zanker.
Linked open data to support content-based recommender systems.
In 8th International Conference on Semantic Systems (I-SEMANTICS 2012), ICP. ACM Press, 2012.
SeRSy 2012 – International Workshop on Semantic Technologies
meet Recommender Systems & Big Data
ISWC 2012 November 11, 2012 Boston, USA

16. Vector Space Model for DBpedia (ii)
Robert Brian
Al Pacino
STARRING De Niro Dennehy
(a1)
(a2) (a3)
Righteous
  
Kill (m1)
Heat (m2)   
Righteous Kill Heat
wactorx ,moviey  tf actorx ,moviey  idf actorx
Righteous Kill (m1) wa1,m1 wa2,m1 wa3,m1
Heat (m2) wa1,m2 wa2,m2 0
SeRSy 2012 – International Workshop on Semantic Technologies
meet Recommender Systems & Big Data
ISWC 2012 November 11, 2012 Boston, USA

17. Vector Space Model for DBpedia(iii)
wa1 ,m1  wa1 ,m2  wa2 ,m1  wa2 ,m2  wa3 ,m1  wa3 ,m2
simstarring (m1 , m2 ) 
wa1 ,m1  wa2 ,m1  wa3 ,m1  wa1 ,m2  wa2 ,m2  wa3 ,m2
2 2 2 2 2 2
 starring  simstarring (m1 , m2 )
+
 director  simdirector (m1 , m2 )
+
 subject  simsubject (m1 , m2 )
+
… =
sim(m1 , m2 )
SeRSy 2012 – International Workshop on Semantic Technologies
meet Recommender Systems & Big Data
ISWC 2012 November 11, 2012 Boston, USA

18. Content-based Recommendations
with DBpedia (ii)
We predict the rating using a Nearest Neighbor Classifier wherein the
similarity measure is a linear combination of property-dependent
similarities
  p  sim p (m j , mi )

m j  profile ( u ,cmp )
vj  p
P
rPreF (ucmp , mi ) 
profile(u, cmp)
What about computing alpha coefficients?
T.Di Noia, R. Mirizzi, V. C. Ostuni, D. Romito, and M. Zanker.
Linked open data to support content-based recommender systems.
In 8th International Conference on Semantic Systems (I-SEMANTICS 2012), ICP. ACM Press, 2012.
What about other content-based algorithms with LOD?
T. Di Noia, R. Mirizzi, V. C. Ostuni, and D. Romito.
Exploiting the web of data in model-based recommender systems.
In 6th ACM Conference on Recommender Systems (RecSys 2012). ACM, ACM Press, 2012.
SeRSy 2012 – International Workshop on Semantic Technologies
meet Recommender Systems & Big Data
ISWC 2012 November 11, 2012 Boston, USA

19. Contextual Post-filtering
Regarding the geographical criteria we apply post-filtering in order to re-rank the
recommendations considering these criteria. For each criterion we introduce a
binary variable which means the absence or presence of that criteria.
( h c cl  ar  ap )
r ( ucmp , mi )  1  rPreF (ucmp , mi )   2 
5
h (hierarchy): it is equal to 1 if the cinema is in the same city of the current
user position, 0 otherwise;
c (cluster ): it is equal to 1 if the cinema is part of a multiplex cinema, 0 otherwise;
cl (co-location): it is equal to 1 if the cinema is close to other POIs, 0 otherwise;
ar (association-rule): it is equal to 1 if the user knows the price of the ticket, 0
otherwise.
ap (anchor-point proximity): it is equal to 1 if the cinema is close to the user's
house or the user's office, 0 otherwise.
SeRSy 2012 – International Workshop on Semantic Technologies
meet Recommender Systems & Big Data
ISWC 2012 November 11, 2012 Boston, USA

20. Recommendations
What?
Where?
SeRSy 2012 – International Workshop on Semantic Technologies
meet Recommender Systems & Big Data
ISWC 2012 November 11, 2012 Boston, USA

21. Conclusion & Future directions
We have presented Cinemappy: a context-aware content-based
recommender system for movies and movie theaters suggestions.
The main features are:
 Android App
 The Content-based Recommender is boosted by DBpedia localized
graphs (IT and EN)
 Several contextual factors have a fundamental role in the
recommendations. Some of them are geographic criteria that go
beyond the simple geographic distance.
We are currently working on:
 Exploiting several implicit form of user feedbacks
 Improving the recommendation with a hybrid approach (content-
based and collaborative filtering)
SeRSy 2012 – International Workshop on Semantic Technologies
meet Recommender Systems & Big Data
ISWC 2012 November 11, 2012 Boston, USA

22. Q&A
Stay tuned! Soon Cinemappy
available on the Android Market
We acknowledge partial support of HP IRP 2012. Grant CW267313.
SeRSy 2012 – International Workshop on Semantic Technologies
meet Recommender Systems & Big Data
ISWC 2012 November 11, 2012 Boston, USA