=Paper=
{{Paper
|id=Vol-1165/paper3
|storemode=property
|title=Ricochet: Context and Complementarity-Aware, Ontology-based POIs Recommender System
|pdfUrl=https://ceur-ws.org/Vol-1165/salad2014-3.pdf
|volume=Vol-1165
|dblpUrl=https://dblp.org/rec/conf/esws/LuLS14
}}
==Ricochet: Context and Complementarity-Aware, Ontology-based POIs Recommender System==
https://ceur-ws.org/Vol-1165/salad2014-3.pdf
Ricochet: Context and Complementarity-Aware,
Ontology-based POIs Recommender System
Chun Lu1, Philippe Laublet2, Milan Stankovic1,2
1
Sépage S.A.S, 96 bis boulevard Raspail, 75006 Paris, France
2 STIH, UniversitéParis-Sorbonne, 28 rue Serpente, 75006 Paris, France
chun@sepage.fr, philippe.laublet@paris-sorbonne.fr,
milstan@sepage.fr
Abstract. In this paper we propose a new approach for improving the personaliza-
tion of POIs recommender system. Existing context-aware POIs recommender sys-
tems usually take into account only peripheral contextual variables. We present
Ricochet, an ontology-based system that refines the recommendation results by im-
plementing an inter-POI parameter that we call the “complementarity”. We show
how this new parameter can generate more effective recommendations. Our exper-
iments are grounded using data from the location-based social network (LBSN)
Yelp.com.
Keywords: POI, Recommender system, Context, Complementarity, Ontology
1 Introduction
Recommender systems have changed the way people find products, information and even
other people. They provide personalized recommendations and predictions over a large
amount of information.
Places of interest, also called points of interest (POIs), are geographical marks that
represent a certain importance for people because they play a specific role in the city. For
example, places where we eat (restaurant), where we sleep (hotel), where we spend a good
moment (bar) or where we participate in cultural activities (museum, theater).
With the rapid growth of location-based social networks (LBSNs), POIs recommender
systems are becoming increasingly popular. Various types of approaches can be found
both in academic literature (such as context-aware approach [1, 3, 10, 12]), and in
commercialized applications and Web sites (Foursquare1, Yelp2, Facebook places3) that
mostly rely on collaborative filtering. However, most of these systems do not take into
account the dynamic nature of the user’s preferences, and assume that the user is likely to
accept recommendations in the same way in any situation, regardless of the POI he is
currently in, or that he just visited.
1 https://www.foursquare.com/
2 http://www.yelp.com/
3 https://www.facebook.com/
SALAD 2014 Workshop – Services and Applications over Linked APIs and Data 10
The research questions driving our work are: why do people go from one POI to an-
other? Is there a link between these two POIs?
The contributions of this paper are two-fold:
A set of recommendation criteria for constructing a relevant POIs recommender sys-
tem, derived from a qualitative user study, based on interviews of 12 users of LBSN
applications such as Foursquare and Yelp.
An ontology of POIs compatible with Yelp taxonomy of place categories. A Semantic
Web-based approach that is capable of re-ranking Yelp’s recommendations by taking
into account the complementarity parameter.
The rest of the paper is organized as follows. In section 2, we present the state of the
art. In section 3, we detail the “Ricochet” system. In section 4, we present the evaluation
methodology and we report the results. In section 5, we summarize the outcome of this
work and we mention some future work.
2 State of the art
The POIs recommender systems are a recent but important research domain that attracts
contributions from academic research institutions and companies constructing novel user
applications.
In [10], the authors present a location-based POIs recommender system which infers a
user’s preferences by mining this person’s social network profile and by considering the
physical constraints delimited by the location and the form of transportation. The system
also takes into account how the user is feeling at the moment. We drove this notion of
feeling further by studying the impact of the feeling that POIs provoke to the choice and
the complementarity of future POIs. In [1, 3, 12], the authors propose context-aware POIs
recommender systems. In [1] we find the thoroughest set of contextual variables: distance
to POI, temperature, weather, season, companion, time day, weekday, crowdedness, fa-
miliarity, mood, budget, travel length, means of transport, travel goal. But for us, these
variables are peripheral. Our work enriches the set of contextual variables with the inter-
POIs parameter “complementarity”. Another difference is that these systems do not use
Semantic Web technologies.
In [2, 4, 7, 8, 9], the authors use Semantic Web technologies in different recommender
systems: adaptive hypermedia systems, hotel search, POIs recommender system for driv-
ers etc. The data are grounded with the ontology that provides a very well semantic sup-
port for developing and improving personalized functionalities such as recommendations.
In our work, we also use these technologies because of the power and facility in the
knowledge representation and the inference.
In [13] and [14], the authors use sentiment analysis techniques to study user’s com-
ments of a venue. [13] concentrates on the general polarity and [14] on the different ap-
preciations about different items at a venue. These approaches can contribute to a better
user preference profile. But they need detailed comments and cannot help the decision of
the immediate recommendation. [5] studies the temporal effects for the location recom-
SALAD 2014 Workshop – Services and Applications over Linked APIs and Data 11
mendation, more precisely on correlations between a user’s check-in time and the corre-
sponding check-in preferences. In our work, we study the immediate effect caused by the
check-in activity and thus recommend complementary POIs that best respond to this ef-
fect.
Except for academic papers, there are several commercialized applications like Four-
square, Yelp, Facebook places etc. They usually allow users to do check-ins. Recommen-
dations in these LBSNs usually combine the collaborative filtering and the context. On
Foursquare, some recommendations are based on the user’s and the user’s friends’ check-
in history. For example, “You haven’t been here yet”, “Your friends have checked-in
here” etc. Foursquare has released a new recommender system recently. Our work was
conducted before. These proprietary applications do not use Semantic Web technologies
and do not consider the complementarity parameter between POIs. In this paper we
demonstrate the importance and the advantage of doing so.
3 The Ricochet system
In this section, we describe the Ricochet system. The presentation consists of the follow-
ing parts: criteria of recommendation, construction of OntoPOI, recommendation engine.
3.1 Criteria of recommendation
We conducted user interviews in order to understand the important elements of POI
choice. According to the responses of our interviewees, we found 3 types of criteria that
people take into account when choosing a POI: contextual criteria, intrinsic criteria of
POIs and criteria of complementarity between POIs.
Contextual criteria indicate the weather, the moment of the day etc. Intrinsic criteria of
POIs indicate characteristics of POIs, for example, the opening hours, the popular hours,
the price, the atmosphere, the comfort, the decoration, the location, the type of food, the
quality of food, the quality of the service, the smells etc. Criteria of complementarity in-
dicate relations between POIs and reasons why we go from one POI to another. In order
to represent the knowledge about POIs and to do intelligent inferences, we decided to
construct an ontology.
3.2 Construction of OntoPOI
Like all ontologies, OntoPOI contains two basic components: classes and properties. The
class "Thing" has four sub-classes: "Place", "Context", "Characteristic" and "Comple-
mentarity". "Context", "Characteristic" and "Complementarity" correspond with the three
types of criteria of recommendation. The sub-classes of "Place" are the taxonomy of en-
tertainment POIs. As we wanted to realize some experiments with Yelp’s data, we de-
cided to use the same taxonomy as Yelp: "Active Life", " Arts and Entertainment ", "
Beauty and Spas ", "Food", "Nightlife" and "Restaurant". OntoPOI is thus compatible
SALAD 2014 Workshop – Services and Applications over Linked APIs and Data 12
with Yelp’s data and can reorganize them in an intelligent way thanks to the good infer-
ring capacity of the logic-based ontology. OntoPOI can be downloaded at: http://sep-
age.com/ontology/OntoPOI_Ricochet.owl
Some precisions on the representation of criteria of complementarity
In [10], the authors made a filtering based on the feeling of the user. They used the
same POIs categorization as Foursquare. For them, each POI category was mapped to a
particular feeling:
Arts & Entertainment= "feeling artsy" College & Education="feeling nerdy"
Nightlife="feeling like a party animal" Great Outdoors="feeling outdoorsy"
Shops="feeling shopaholic" Food="feeling hungry"
Home / Work / Other="feeling workaholic"
Our work is aligned to this idea and takes it further. We go to a POI because we have
a certain feeling at a given moment. This feeling can be caused by the POI that we just
visited and the POI where we decide to go can satisfy this feeling. This feeling can be
also interpreted as a need or a sensation. These feelings/needs/sensations can be physio-
logical, as the hunger, the thirst and the elimination. They can also be physical or intel-
lectual. The complementarity can be seen as the link between the POI that causes a need
and the POI that satisfies this need. For example, the hunger can be caused by a POI
where we make physical efforts and can be satisfied by a POI where we eat. The relation
of complementarity between two POIs can be interpreted in the following way:
POI 1 causes Feeling satisfies POI 2
Figure 1. Principle of criteria of complementarity
In [11], the authors showed that physical activities possessed a specific intensity and
this can be assessed. We considered that every entertainment activity possessed an inten-
sity of expressiveness at several levels: cognitive, emotional and physical. This notion
concretizes the feeling discussed above. Empirically, we classified our POIs in four in-
tensities of expressiveness: high, moderately high, moderately low and low. We created
four sub-classes of "Place": "High Intensity Expressiveness place", "Moderately High In-
tensity Expressiveness place", "Moderately Low Intensity Expressiveness place", "Low
Intensity Expressiveness place". Then, we classified each of the POIs classes according
to their intensity of expressiveness. The result of our interviews showed that daily activi-
ties required an alternation of different rhythms and intensities. An "X Intensity Expres-
siveness Place" causes X intensity. X intensity needs to be alternated by Y intensity other
than X. Y intensity is provoked by a "Y Intensity Expressiveness place". We can simplify
the deduction like this:
X Intensity Expressiveness place
X Intensity
Y(≠X) Intensity Expressiveness place
Figure 2. Deduction of the complementarity for its representation in OntoPOI
SALAD 2014 Workshop – Services and Applications over Linked APIs and Data 13
To represent this in the ontology, we used OWL Full4. We created two Object Proper-
ties “is caused by” and “is satisfied by” which have the class “Complementarity” as do-
main and rdfs:Class as range. The class "Complementarity" has four instances: "high in-
tensity", "moderately high intensity", "moderately low intensity", "low intensity". We de-
fined, for each instance, their values of the two properties by applying the process repre-
sented on Figure 2. We created some other Object Properties: "is the exposition of" (inside
or outside) and "is the time for". We created some Datatype Properties: "has address",
"has city", "has phone", "has postal code" etc which are basic information about POIs.
3.3 Recommendation engine
We used several tools: Yelp API, Jena API and Google Maps API. Firstly, the Java pro-
gram gets data by accessing to Yelp API that returns 50 POIs near the current location of
the user. Secondly, the information of these POIs are translated into RDF triples and
stored in Jena RDF repository with OntoPOI. Thirdly, according to the context and the
check-in information of the user, we generate SPARQL queries, in order to determine the
adequacy of each POI with regards to the user’s situation, according to 3 criteria. The
total adequacy is calculated according to the following formula:
Total point = Point of the weather criterion + Point of the moment of day criterion
+3 * Point of the complementarity criterion.
Ricochet recommends the 10 highest-rated POIs in descending order. The recommended
POIs are marked on a local map by using Google Maps API. This process is represented
on Figure 3.
Translation of
Location Yelp’s data into
RDF Triple
Context/
Check-in
Storage of data in
Rating POIs Query RDF repository
Recommendation Processing with OntoPOI
GoogleMAP Inference
Figure 3. Diagram of the recommendation engine
4 Evaluation
In order to evaluate the effectiveness of our approach, we compared the recommendations
produced by two variants of Ricochet. The first (R1) is that described above, the second
(R2) is not complementarity-aware. Using these two variants, we could measure if the
proposed complementarity-aware system can improve the user perceived relevance of the
recommendations with the same data. The evaluations took place in Paris in France with
4 http://www.w3.org/TR/owl-ref/#OWLFull
SALAD 2014 Workshop – Services and Applications over Linked APIs and Data 14
the participation of 10 persons. We varied the day time (morning, afternoon, night) and
the place (not only downtown). We asked the evaluators to choose a POI nearby in our
database and to imagine that they just visited it. We obtained the recommendations made
by the two variants. Then, we invited our evaluators to rate each recommended POI by
referring to the following scale: 0: I won’t go there (not relevant); 1: I hesitate (partially
relevant); 2: I’ll go there without hesitation (definitely relevant). To gauge the relevance,
we used the following metrics:
Precision was used to evaluate the quality of the recommendations. It is the number
of relevant recommended POIs divided by the total number of recommended POIs. The
POI having the score 2 counts for 1 relevant POI, 1 counts for 0.5, 0 doesn’t count.
Relevant recommended POIs
Precision =
Total recommended POIs
Recall was used to evaluate the quantity of POIs extracted. We modified the traditional
recall. We showed not all the database but only 10 recommended POIs. The evaluators
judged the relevance of these shown POIs. We cannot know their appreciation about the
non-shown POIs. The modified recall is the number of relevant recommended POIs di-
vided by the total number of relevant recommended POIs of the two variants.
Relevant recommended POIs
Recall (modified) =
Total relevant recommended POIs of the 2 variants
normalized Discounted Cumulative Gain (nDCG) was used to evaluate the quality
of the ranked list. The nDCG value of a ranking list at position n is calculated in the
following way:
n
2r(j) − 1, j = 1
N(n) ≡ Zn ∑ {2r(j) − 1
j=1 ,j > 1
log(j)
where 𝑟(𝑗) is the rating of the j-th document in the ranking list, and the normalization
constant Zn is chosen so that the perfect list gets a nDCG score of 1 ([5]).
The results are shown on Figure 4.
Figure 4. Results of the precision, the recall and the nDCG
R1 is the first variant, R2 the second. 1-10 are the numbers of the ten evaluations. We can
see clearly that the performance of R1 is generally better than that of R2 in terms of the
quality, the quantity and the ranking. For 1 (a park), the two performances are the same.
It was the dinner time. During the mealtime, the rank of the food POIs is largely elevated.
The rest of the evaluations were done outside the mealtime where Ricochet privileged the
SALAD 2014 Workshop – Services and Applications over Linked APIs and Data 15
complementarity parameter. The evaluators were pleased to be recommended comple-
mentary POIs. For 2 (a coffee), 8 (a massage center), we are well rested and often want a
more intense activity, R2 recommended a beauty spa, a day spa, a massage, a tea room
which have a similar intensity, R1 recommended only more intense POIs like a park, a
dance studio, a museum etc. For 4 (a cinema), R2 recommended another cinema. For 5
(a gym club), 7 (a swimming pool), 10 (a tennis court), we normally feel tired, but R2
recommended still wearying POIs like a swimming pool, a gym club, a bike rental, a
museum, R1 recommended more relaxing POIs, like a juice bar, a cinema, a cosmetic
beauty supply, a coffee-tea. For 3 (a museum) and 6 (a hair salon), 9 (a musical venue),
the performances are close. As before, Ricochet recommended complementary POIs that
have a different intensity. But it turned out that the proposed intensity was not always
suitable. People often need to change rhythm between two activities. But every individual
has his own rhythm. The change of the rhythm seems to be submitted to a modulation
according to the psyche of the individual. For example, some people like only the restful
activities, for them, we would do better to eliminate the tiring activities. Others like all
the activities, light as intense. For the latter, the range of the recommendations would
deserve to be more refined. In our current system, there is only one rule on the change of
the rhythm. However the appropriateness of a POI with regards to the rhythm is important
and we intend to further personalize the taking into account of the rhythm, potentially by
combining our approach with a machine learning approach.
To conclude this evaluation, even though the number of evaluators was limited, the
results showed evidence that considering the complementarity can improve the relevance
of the recommended POIs. In our future work we intend to conduct a more complete
evaluation with more users and more specific situations.
5 Conclusion and Future work
In this paper, we presented Ricochet an ontology-based POIs recommender system and
illustrated the advantage of this approach and the importance of implementing the com-
plementarity parameter when recommending POIs.
In spite of good critics of our evaluators, there are still several weak points in our sys-
tem to be improved in a future work.
Firstly, being dependent to Yelp API, we cannot have access to all recommendable
POIs but to a pre-selection made by Yelp. This is also the reason why we didn’t compare
our results against the baseline ordering from Yelp. This influences on the application of
the complementarity parameter and thus on the quality of the results. Secondly, we iden-
tified but did not include yet the intrinsic criteria in our system. And nevertheless, these
criteria can potentially be useful for refining the results. Thirdly, it could be possible to
cover a more complete set of contextual information like in [1]. Fourthly, the notion of
complementarity could also be further refined to improve the personalization of the rec-
ommendation. The improvement and the personalization require a better knowledge on
the user, and more exactly, on the scale of the user’s acceptable activities. Fifthly, we can
use the technologies installed on today’s mobile devices like the wireless accelerometers,
the heart rate monitor and the sensor of calories. We can measure the exact intensity of
SALAD 2014 Workshop – Services and Applications over Linked APIs and Data 16
the activity or the number of calories spent. Having these data, we can better recommend
appropriated POIs that lead people to a balanced and healthy life. With more and more
strongly typed user data available, the use of Semantic Web knowledge structures to make
relevant, context-aware recommendations makes more and more sense.
The POIs recommender systems are adopted by the modern society. An ongoing effort
should be made to help people discover more personalized and relevant POIs.
References
1. Baltrunas, L., Ludwig, B., Peer, S., Ricci, F.: Context-Aware Places of Interest Recommenda-
tions of Mobile Users. In: Human-Computer Interaction International, pp. 531-540. Heidel-
berg, Berlin (2011)
2. Brut, M., Sedes, F.: Modélisation basée sur ontologies pour développer des recommandations
personnalisées dans les systèmes hypermédia adaptatives. In: INFORSID, pp. 61-76. Marseille
(2010)
3. Costa, H., Furtado, B., Pires, D., Macedo, L., Cardoso, A.: Context and Intention-Awareness
in POIs Recommender Systems. In: Proc. 6th ACM Conf. on Recommender Systems, 4th
Workshop on Context-Aware Recommender Systems, RecSys’12, NY, USA (2012)
4. Debattista, J., Scerri, S., Rivera, I., Handschuh, S.: Ontology-based rules for recommender sys-
tems. In: Proceedings of the International Workshop on Semantic Technologies meet Recom-
mender Systems & Big Data, pp. 49-60. (2012)
5. Gao, Huiji, Tang, Jiliang, Hu, Xia, Liu, Huan.: Exploring temporal effects for location recom-
mendation on location-based social networks. In: Proceedings of the 7th ACM conference on
Recommender systems, pp. 93-100. ACM, New York, USA (2013)
6. Liu, T.-Y., Xu, J., Qin, T., Xiong, W., Li, H.: Letor: Benchmark dataset for research on learning
to rank for information retrieval. In: Proceedings of SIGIR 2008 Workshop on Learning to
Rank for Information Retrieval. (2007)
7. Ngamsrithepparit, S. et al.: Rule Management System for Ontology-based Recommendation
System. In: The Joint International Symposium on Natural Language Processing and Agricul-
tural Ontology Service 2011, Bangkok, Thailand, Feb 9-10, (2012)
8. Niemann, M., Mochol, M., Tolksdorf, R.: Improving online hotel search - what do we need
semantics for? In: Proc. of the Semantics (2006)
9. Parundekar, R., Oguchi, K.: Driver Recommendations of POIs using a Semantic Content-
based Approach. In: SeRSy, pp. 13-24. (2012)
10. Savage, N.S. et al.: I’m feeling LoCo: A Location Based Context Aware Recommendation
System. In: Advances in Location-Based Services, pp. 37-54. Heidelberg, Berlin (2012)
11. Tapia, E.M. et al.: Real-Time recognition of physical activities and their intensities using wire-
less accelerometers and a heart rate monitor. In: Proceedings of the 2007 11th IEEE Interna-
tional Symposium on Wearable Computers, pp. 1-4. (2007)
12. Waga, K., Tabarcea, A., Fränti, P.: Recommendation of Points of Interest from User Generated
Data Collection. In: CollaborateCom 2012, Pittsburgh, USA, October (2012)
13. Yang, D., Zhang, D., Yu Z., Wang, Z.: A sentiment-enhanced personalized location recom-
mendation system. In: Proceedings of the 24th ACM Conference on Hypertext and Social Me-
dia (HT '13), pp. 119-128. ACM, New York, NY, USA (2013)
14. Yang, D., Zhang, D., Yu, Z., Yu, Z.: Fine-grained preference-aware location search leveraging
crowdsourced digital footprints from LBSNs. In: Proceedings of the 2013 ACM international
joint conference on Pervasive and ubiquitous computing (UbiComp '13), pp. 479-488. ACM,
New York, NY, USA (2013)
SALAD 2014 Workshop – Services and Applications over Linked APIs and Data 17