Top-N Recommendations
from Implicit Feedback
leveraging Linked Open Data
Vito Claudio Ostuni, Tommaso Di Noia, Eugenio Di Sciascio, Roberto Mirizzi
ostuni@deemail.poliba.it, t.dinoia@poliba.it, disciascio@poliba.it, mirizzi@deemail.poliba.it

Polytechnic University of Bari - Bari (ITALY)

RecSys 2013 – 7th ACM Conference on Recommender Systems
October 12-16, 2013 Hong Kong, China
Outline
 Introduction and motivation
 SPrank: Semantic Path-based ranking
 Data model and Problem formulation
 Path-based features
 Learning the ranking function

 Experimental Evaluation
 Contributions and Conclusion

RecSys 2013 – 7th ACM Conference on Recommender Systems
October 12-16, 2013 Hong Kong, China
Linked Open Data
• Initiative for publishing and connecting data on the Web using
Semantic Web technologies;
• >30 billion of RDF triples from hundreds of data sources;
• Semantic Web done right [ http://www.w3.org/2008/Talks/0617-lod-tbl/#(3) ]

RecSys 2013 – 7th ACM Conference on Recommender Systems
October 12-16, 2013 Hong Kong, China
Linked Open Data
• Initiative for publishing and connecting data on the Web using
Semantic Web technologies;
• >30 billion of RDF triples from hundreds of data sources;
• Semantic Web done right [ http://www.w3.org/2008/Talks/0617-lod-tbl/#(3) ]

RecSys 2013 – 7th ACM Conference on Recommender Systems
October 12-16, 2013 Hong Kong, China
Hong Kong in DBpedia
db:Hong_Kong
db:thumbnail

subject

predicate

object

8134 triples

RecSys 2013 – 7th ACM Conference on Recommender Systems
October 12-16, 2013 Hong Kong, China
Hong Kong in DBpedia
db:Hong_Kong
db:thumbnail

Skyscrapers over 350 meters in Hong Kong?
select * where {
?s dbpedia-owl:location <http://dbpedia.org/resource/Hong_Kong>.
?s dcterms:subject
category:Skyscrapers_over_350_meters. }

RecSys 2013 – 7th ACM Conference on Recommender Systems
October 12-16, 2013 Hong Kong, China
Hong Kong in DBpedia
db:Hong_Kong
db:thumbnail

db:location

db:International_Commerce_centre
db:thumbnail

db:Central_Plaza_(Hong_Kong)
dcterms:subject
db:thumbnail

db:category:Skyscrapers_over_350_meters)

RecSys 2013 – 7th ACM Conference on Recommender Systems
October 12-16, 2013 Hong Kong, China
Motivation
Traditional Ontological/Semantic Recommender Systems:
• make use of limited domain ontologies;
• rely on explicit feedback data;
• address the rating prediction task.

RecSys 2013 – 7th ACM Conference on Recommender Systems
October 12-16, 2013 Hong Kong, China
Motivation
Traditional Ontological/Semantic Recommender Systems:
• make use of limited domain ontologies;
• rely on explicit feedback data;
• address the rating prediction task.

But…
• a lot of structured semantic data on the Web;
• Implicit feedback are easier to collect;
• Top-N Recommendations is a more realistic task.

RecSys 2013 – 7th ACM Conference on Recommender Systems
October 12-16, 2013 Hong Kong, China
Motivation
Traditional Ontological/Semantic Recommender Systems:
• make use of limited domain ontologies;
• rely on explicit feedback data;
• address the rating prediction task.

But…
• a lot of structured semantic data on the Web;
• Implicit feedback are easier to collect;
• Top-N Recommendations is a more realistic task.

Challenge:
• compute Top-N Item Recommendations from implicit feedback
exploiting the Web of Data.
RecSys 2013 – 7th ACM Conference on Recommender Systems
October 12-16, 2013 Hong Kong, China
Our approach
• Usage of structured semantic data freely available on the Web
(Linked Open Data) to describe items
DBpedia ontology

RecSys 2013 – 7th ACM Conference on Recommender Systems
October 12-16, 2013 Hong Kong, China
Our approach
• Analysis of complex relations between the user preferences and
the target item (extraction of path-based features)

RecSys 2013 – 7th ACM Conference on Recommender Systems
October 12-16, 2013 Hong Kong, China
Our approach
• Analysis of complex relations between the user preferences and
the target item (extraction of path-based features)

• Formalization of the Top-N Item recommendation problem from
implicit feedback in a Learning To Rank setting

RecSys 2013 – 7th ACM Conference on Recommender Systems
October 12-16, 2013 Hong Kong, China
Data model
Implicit Feedback Matrix
^

S

I1

i2

i3

i4

u1

1

1

0

0

u2

1

0

1

0

u3

0

1

1

0

u4

0

1

0

Knowledge Graph

1

RecSys 2013 – 7th ACM Conference on Recommender Systems
October 12-16, 2013 Hong Kong, China
Data model
Implicit Feedback Matrix
^

S

I1

i2

i3

i4

u1

1

1

0

0

u2

1

0

1

0

u3

0

1

1

0

u4

0

1

0

Knowledge Graph

1

RecSys 2013 – 7th ACM Conference on Recommender Systems
October 12-16, 2013 Hong Kong, China
Data model
Implicit Feedback Matrix
^

S

I1

i2

i3

i4

u1

1

1

0

0

u2

1

0

1

0

u3

0

1

1

0

u4

0

1

0

Knowledge Graph

1

RecSys 2013 – 7th ACM Conference on Recommender Systems
October 12-16, 2013 Hong Kong, China
Problem formulation

u

^


u

^

I  {i  I | s ui  1}

Set of relevant items for u

I  {i  I | s ui  0}

Set of irrelevant items for u



Iu *  Iu

Sample of irrelevant items for u

xui 

Feature vector

D



^





 xui , s ui  i  ( I u  I u * )

TR 
u

Training Set

RecSys 2013 – 7th ACM Conference on Recommender Systems
October 12-16, 2013 Hong Kong, China
Path-based features
path acyclic sequence of relations ( s , .. rl , .. rL )
u3 s i2 p2 e1 p1 i1
xui ( j ) 

 (s, p2 , p1)

# pathui ( j )
D

 # path
d 1

ui

(d )

Frequency of pathj in the sub-graph
related to u ad i

• The more the paths, the more the item is relevant.
• Different paths have different meaning.
• Not all types of paths are relevant.

RecSys 2013 – 7th ACM Conference on Recommender Systems
October 12-16, 2013 Hong Kong, China
Path-based features
i1

xu3i1 ?

e1

u1

e3
u2

i2
e2

u3
i3

u4
i4
RecSys 2013 – 7th ACM Conference on Recommender Systems
October 12-16, 2013 Hong Kong, China

e5

e4
Path-based features
path1 (s, s, s) : 1

i1

e1

u1

e3
u2

i2
e2

u3
i3

u4
i4
RecSys 2013 – 7th ACM Conference on Recommender Systems
October 12-16, 2013 Hong Kong, China

e5

e4
Path-based features
path1 (s, s, s) : 2

i1

e1

u1

e3
u2

i2
e2

u3
i3

u4
i4
RecSys 2013 – 7th ACM Conference on Recommender Systems
October 12-16, 2013 Hong Kong, China

e5

e4
Path-based features
path1 (s, s, s) : 2
path2 (s, p2, p1) : 1

i1

e1

u1

e3
u2

i2
e2

u3
i3

u4
i4
RecSys 2013 – 7th ACM Conference on Recommender Systems
October 12-16, 2013 Hong Kong, China

e5

e4
Path-based features
path1 (s, s, s) : 2
path2 (s, p2, p1) : 2

i1

e1

u1

e3
u2

i2
e2

u3
i3

u4
i4
RecSys 2013 – 7th ACM Conference on Recommender Systems
October 12-16, 2013 Hong Kong, China

e5

e4
Path-based features
path1 (s, s, s) : 2
path2 (s, p2, p1) : 2
path3 (s, p2, p3, p1) : 1

i1

e1

u1

e3
u2

i2
e2

u3
i3

u4
i4
RecSys 2013 – 7th ACM Conference on Recommender Systems
October 12-16, 2013 Hong Kong, China

e5

e4
Path-based features
path1 (s, s, s) : 2
path2 (s, p2, p1) : 2
path3 (s, p2, p3, p1) : 1

2
xu3i1 (1) 
5
2
xu3i1 (2) 
5
1
xu3i1 (3) 
5

i1

e1

u1

e3
u2

i2
e2

u3
i3

u4
i4
RecSys 2013 – 7th ACM Conference on Recommender Systems
October 12-16, 2013 Hong Kong, China

e5

e4
Learning the ranking function
Point-wise Learning To Rank
Learn a prediction function f :

D



^

s.t. f ( xui )  sui

Assumption: if f is accurate, then the ranking induced by f should
be close to the desired ranking
• Simplest LTR technique
• Very effective in practice (Yahoo! Learning to Rank Challenge best
solution was extremely randomized trees in a standard regression setting)

RecSys 2013 – 7th ACM Conference on Recommender Systems
October 12-16, 2013 Hong Kong, China
BagBoo
BagBoo: a scalable hybrid bagging-the-boosting model
[D. Pavlov, A. Gorodilov, C. Brunk CIKM2010]

• Combination of Random Forest (Bagging) and Gradient Boosted
Regression Trees (Boosting)
• Combines the high accuracy of gradient boosting with the resistance
to overfitting of random forests

For b=1 to B:
Tb  TR
fb  learn GBRT from Tb
1 B
f   fb
B b 1
RecSys 2013 – 7th ACM Conference on Recommender Systems
October 12-16, 2013 Hong Kong, China
Evaluation Methodology
• Top-N Item recommendation task
• Evaluation methodology similar to:
[Cremonesi, Koren and Turrin, RecSys 2010]

• Evaluation with different user profile size:
given 5
given 10

User
profile

5

User profile

Test Set

10
……

given All

User profile

Test Set

10
Test Set

RecSys 2013 – 7th ACM Conference on Recommender Systems
October 12-16, 2013 Hong Kong, China
Datasets
Subset of Movielens mapped to DBpedia
3,792 users
2,795 movies
104,351 entities

Subset of Last.fm mapped to DBpedia
852 users
6,256 artists
150,925 entities

Mappings
http://sisinflab.poliba.it/mappingdatasets2dbpedia.zip

RecSys 2013 – 7th ACM Conference on Recommender Systems
October 12-16, 2013 Hong Kong, China
Evaluation of different ranking functions
Movielens
0,6

0,5

recall@5

0,4

BagBoo

0,3

GBRT
Sum

0,2

0,1

0
given 5

given 10

given 20

given 30

given 50

given All

user profile size

RecSys 2013 – 7th ACM Conference on Recommender Systems
October 12-16, 2013 Hong Kong, China
Evaluation of different ranking functions
Last.fm
0,6

0,5

recall@5

0,4

BagBoo

0,3

GBRT
Sum

0,2

0,1

0
given 5

given 10

given 20

given All

user profile size

RecSys 2013 – 7th ACM Conference on Recommender Systems
October 12-16, 2013 Hong Kong, China
Comparative approaches
MyMediaLite
• BPRMF, Bayesian Personalized Ranking for Matrix Factorization
• BPRLin, Linear Model optimized for BPR (Hybrid alg.)
• SLIM, Sparse Linear Methods for Top-N Recommender Systems

• SMRMF, Soft Margin Ranking Matrix Factorization

RecSys 2013 – 7th ACM Conference on Recommender Systems
October 12-16, 2013 Hong Kong, China
Comparison with other approaches
Movielens
0,6

0,5

recall@5

0,4
SPrank
BPRMF

0,3

SLIM
BPRLin

0,2

SMRMF

0,1

0
given 5

given 10

given 20

given 30

given 50

user profile size

RecSys 2013 – 7th ACM Conference on Recommender Systems
October 12-16, 2013 Hong Kong, China

given All
Comparison with other approaches
Last.fm
0,6

0,5

recall@5

0,4
SPrank
BPRMF

0,3

SLIM
BPRLin

0,2

SMRMF

0,1

0
given 5

given 10

given 20

given All

user profile size

RecSys 2013 – 7th ACM Conference on Recommender Systems
October 12-16, 2013 Hong Kong, China
Contributions
SPrank: Semantic Path-based ranking
 Combination of semantic item descriptions from the Web
of Data and implicit feedback
 Mining of the semantic graph using path-based features
 Learning To Rank setting

Future Work:
 Deeper analysis of the path-based features
 Usage of other Learning To Rank approaches

RecSys 2013 – 7th ACM Conference on Recommender Systems
October 12-16, 2013 Hong Kong, China
Q&A
A Little Semantics Goes a Long Way.
Hendler Hypothesis

RecSys 2013 – 7th ACM Conference on Recommender Systems
October 12-16, 2013 Hong Kong, China

Top-N Recommendations from Implicit Feedback leveraging Linked Open Data

  • 1.
    Top-N Recommendations from ImplicitFeedback leveraging Linked Open Data Vito Claudio Ostuni, Tommaso Di Noia, Eugenio Di Sciascio, Roberto Mirizzi ostuni@deemail.poliba.it, t.dinoia@poliba.it, disciascio@poliba.it, mirizzi@deemail.poliba.it Polytechnic University of Bari - Bari (ITALY) RecSys 2013 – 7th ACM Conference on Recommender Systems October 12-16, 2013 Hong Kong, China
  • 2.
    Outline  Introduction andmotivation  SPrank: Semantic Path-based ranking  Data model and Problem formulation  Path-based features  Learning the ranking function  Experimental Evaluation  Contributions and Conclusion RecSys 2013 – 7th ACM Conference on Recommender Systems October 12-16, 2013 Hong Kong, China
  • 3.
    Linked Open Data •Initiative for publishing and connecting data on the Web using Semantic Web technologies; • >30 billion of RDF triples from hundreds of data sources; • Semantic Web done right [ http://www.w3.org/2008/Talks/0617-lod-tbl/#(3) ] RecSys 2013 – 7th ACM Conference on Recommender Systems October 12-16, 2013 Hong Kong, China
  • 4.
    Linked Open Data •Initiative for publishing and connecting data on the Web using Semantic Web technologies; • >30 billion of RDF triples from hundreds of data sources; • Semantic Web done right [ http://www.w3.org/2008/Talks/0617-lod-tbl/#(3) ] RecSys 2013 – 7th ACM Conference on Recommender Systems October 12-16, 2013 Hong Kong, China
  • 5.
    Hong Kong inDBpedia db:Hong_Kong db:thumbnail subject predicate object 8134 triples RecSys 2013 – 7th ACM Conference on Recommender Systems October 12-16, 2013 Hong Kong, China
  • 6.
    Hong Kong inDBpedia db:Hong_Kong db:thumbnail Skyscrapers over 350 meters in Hong Kong? select * where { ?s dbpedia-owl:location <http://dbpedia.org/resource/Hong_Kong>. ?s dcterms:subject category:Skyscrapers_over_350_meters. } RecSys 2013 – 7th ACM Conference on Recommender Systems October 12-16, 2013 Hong Kong, China
  • 7.
    Hong Kong inDBpedia db:Hong_Kong db:thumbnail db:location db:International_Commerce_centre db:thumbnail db:Central_Plaza_(Hong_Kong) dcterms:subject db:thumbnail db:category:Skyscrapers_over_350_meters) RecSys 2013 – 7th ACM Conference on Recommender Systems October 12-16, 2013 Hong Kong, China
  • 8.
    Motivation Traditional Ontological/Semantic RecommenderSystems: • make use of limited domain ontologies; • rely on explicit feedback data; • address the rating prediction task. RecSys 2013 – 7th ACM Conference on Recommender Systems October 12-16, 2013 Hong Kong, China
  • 9.
    Motivation Traditional Ontological/Semantic RecommenderSystems: • make use of limited domain ontologies; • rely on explicit feedback data; • address the rating prediction task. But… • a lot of structured semantic data on the Web; • Implicit feedback are easier to collect; • Top-N Recommendations is a more realistic task. RecSys 2013 – 7th ACM Conference on Recommender Systems October 12-16, 2013 Hong Kong, China
  • 10.
    Motivation Traditional Ontological/Semantic RecommenderSystems: • make use of limited domain ontologies; • rely on explicit feedback data; • address the rating prediction task. But… • a lot of structured semantic data on the Web; • Implicit feedback are easier to collect; • Top-N Recommendations is a more realistic task. Challenge: • compute Top-N Item Recommendations from implicit feedback exploiting the Web of Data. RecSys 2013 – 7th ACM Conference on Recommender Systems October 12-16, 2013 Hong Kong, China
  • 11.
    Our approach • Usageof structured semantic data freely available on the Web (Linked Open Data) to describe items DBpedia ontology RecSys 2013 – 7th ACM Conference on Recommender Systems October 12-16, 2013 Hong Kong, China
  • 12.
    Our approach • Analysisof complex relations between the user preferences and the target item (extraction of path-based features) RecSys 2013 – 7th ACM Conference on Recommender Systems October 12-16, 2013 Hong Kong, China
  • 13.
    Our approach • Analysisof complex relations between the user preferences and the target item (extraction of path-based features) • Formalization of the Top-N Item recommendation problem from implicit feedback in a Learning To Rank setting RecSys 2013 – 7th ACM Conference on Recommender Systems October 12-16, 2013 Hong Kong, China
  • 14.
    Data model Implicit FeedbackMatrix ^ S I1 i2 i3 i4 u1 1 1 0 0 u2 1 0 1 0 u3 0 1 1 0 u4 0 1 0 Knowledge Graph 1 RecSys 2013 – 7th ACM Conference on Recommender Systems October 12-16, 2013 Hong Kong, China
  • 15.
    Data model Implicit FeedbackMatrix ^ S I1 i2 i3 i4 u1 1 1 0 0 u2 1 0 1 0 u3 0 1 1 0 u4 0 1 0 Knowledge Graph 1 RecSys 2013 – 7th ACM Conference on Recommender Systems October 12-16, 2013 Hong Kong, China
  • 16.
    Data model Implicit FeedbackMatrix ^ S I1 i2 i3 i4 u1 1 1 0 0 u2 1 0 1 0 u3 0 1 1 0 u4 0 1 0 Knowledge Graph 1 RecSys 2013 – 7th ACM Conference on Recommender Systems October 12-16, 2013 Hong Kong, China
  • 17.
    Problem formulation  u ^  u ^ I {i  I | s ui  1} Set of relevant items for u I  {i  I | s ui  0} Set of irrelevant items for u   Iu *  Iu Sample of irrelevant items for u xui  Feature vector D  ^     xui , s ui  i  ( I u  I u * ) TR  u Training Set RecSys 2013 – 7th ACM Conference on Recommender Systems October 12-16, 2013 Hong Kong, China
  • 18.
    Path-based features path acyclicsequence of relations ( s , .. rl , .. rL ) u3 s i2 p2 e1 p1 i1 xui ( j )   (s, p2 , p1) # pathui ( j ) D  # path d 1 ui (d ) Frequency of pathj in the sub-graph related to u ad i • The more the paths, the more the item is relevant. • Different paths have different meaning. • Not all types of paths are relevant. RecSys 2013 – 7th ACM Conference on Recommender Systems October 12-16, 2013 Hong Kong, China
  • 19.
    Path-based features i1 xu3i1 ? e1 u1 e3 u2 i2 e2 u3 i3 u4 i4 RecSys2013 – 7th ACM Conference on Recommender Systems October 12-16, 2013 Hong Kong, China e5 e4
  • 20.
    Path-based features path1 (s,s, s) : 1 i1 e1 u1 e3 u2 i2 e2 u3 i3 u4 i4 RecSys 2013 – 7th ACM Conference on Recommender Systems October 12-16, 2013 Hong Kong, China e5 e4
  • 21.
    Path-based features path1 (s,s, s) : 2 i1 e1 u1 e3 u2 i2 e2 u3 i3 u4 i4 RecSys 2013 – 7th ACM Conference on Recommender Systems October 12-16, 2013 Hong Kong, China e5 e4
  • 22.
    Path-based features path1 (s,s, s) : 2 path2 (s, p2, p1) : 1 i1 e1 u1 e3 u2 i2 e2 u3 i3 u4 i4 RecSys 2013 – 7th ACM Conference on Recommender Systems October 12-16, 2013 Hong Kong, China e5 e4
  • 23.
    Path-based features path1 (s,s, s) : 2 path2 (s, p2, p1) : 2 i1 e1 u1 e3 u2 i2 e2 u3 i3 u4 i4 RecSys 2013 – 7th ACM Conference on Recommender Systems October 12-16, 2013 Hong Kong, China e5 e4
  • 24.
    Path-based features path1 (s,s, s) : 2 path2 (s, p2, p1) : 2 path3 (s, p2, p3, p1) : 1 i1 e1 u1 e3 u2 i2 e2 u3 i3 u4 i4 RecSys 2013 – 7th ACM Conference on Recommender Systems October 12-16, 2013 Hong Kong, China e5 e4
  • 25.
    Path-based features path1 (s,s, s) : 2 path2 (s, p2, p1) : 2 path3 (s, p2, p3, p1) : 1 2 xu3i1 (1)  5 2 xu3i1 (2)  5 1 xu3i1 (3)  5 i1 e1 u1 e3 u2 i2 e2 u3 i3 u4 i4 RecSys 2013 – 7th ACM Conference on Recommender Systems October 12-16, 2013 Hong Kong, China e5 e4
  • 26.
    Learning the rankingfunction Point-wise Learning To Rank Learn a prediction function f : D  ^ s.t. f ( xui )  sui Assumption: if f is accurate, then the ranking induced by f should be close to the desired ranking • Simplest LTR technique • Very effective in practice (Yahoo! Learning to Rank Challenge best solution was extremely randomized trees in a standard regression setting) RecSys 2013 – 7th ACM Conference on Recommender Systems October 12-16, 2013 Hong Kong, China
  • 27.
    BagBoo BagBoo: a scalablehybrid bagging-the-boosting model [D. Pavlov, A. Gorodilov, C. Brunk CIKM2010] • Combination of Random Forest (Bagging) and Gradient Boosted Regression Trees (Boosting) • Combines the high accuracy of gradient boosting with the resistance to overfitting of random forests For b=1 to B: Tb  TR fb  learn GBRT from Tb 1 B f   fb B b 1 RecSys 2013 – 7th ACM Conference on Recommender Systems October 12-16, 2013 Hong Kong, China
  • 28.
    Evaluation Methodology • Top-NItem recommendation task • Evaluation methodology similar to: [Cremonesi, Koren and Turrin, RecSys 2010] • Evaluation with different user profile size: given 5 given 10 User profile 5 User profile Test Set 10 …… given All User profile Test Set 10 Test Set RecSys 2013 – 7th ACM Conference on Recommender Systems October 12-16, 2013 Hong Kong, China
  • 29.
    Datasets Subset of Movielensmapped to DBpedia 3,792 users 2,795 movies 104,351 entities Subset of Last.fm mapped to DBpedia 852 users 6,256 artists 150,925 entities Mappings http://sisinflab.poliba.it/mappingdatasets2dbpedia.zip RecSys 2013 – 7th ACM Conference on Recommender Systems October 12-16, 2013 Hong Kong, China
  • 30.
    Evaluation of differentranking functions Movielens 0,6 0,5 recall@5 0,4 BagBoo 0,3 GBRT Sum 0,2 0,1 0 given 5 given 10 given 20 given 30 given 50 given All user profile size RecSys 2013 – 7th ACM Conference on Recommender Systems October 12-16, 2013 Hong Kong, China
  • 31.
    Evaluation of differentranking functions Last.fm 0,6 0,5 recall@5 0,4 BagBoo 0,3 GBRT Sum 0,2 0,1 0 given 5 given 10 given 20 given All user profile size RecSys 2013 – 7th ACM Conference on Recommender Systems October 12-16, 2013 Hong Kong, China
  • 32.
    Comparative approaches MyMediaLite • BPRMF,Bayesian Personalized Ranking for Matrix Factorization • BPRLin, Linear Model optimized for BPR (Hybrid alg.) • SLIM, Sparse Linear Methods for Top-N Recommender Systems • SMRMF, Soft Margin Ranking Matrix Factorization RecSys 2013 – 7th ACM Conference on Recommender Systems October 12-16, 2013 Hong Kong, China
  • 33.
    Comparison with otherapproaches Movielens 0,6 0,5 recall@5 0,4 SPrank BPRMF 0,3 SLIM BPRLin 0,2 SMRMF 0,1 0 given 5 given 10 given 20 given 30 given 50 user profile size RecSys 2013 – 7th ACM Conference on Recommender Systems October 12-16, 2013 Hong Kong, China given All
  • 34.
    Comparison with otherapproaches Last.fm 0,6 0,5 recall@5 0,4 SPrank BPRMF 0,3 SLIM BPRLin 0,2 SMRMF 0,1 0 given 5 given 10 given 20 given All user profile size RecSys 2013 – 7th ACM Conference on Recommender Systems October 12-16, 2013 Hong Kong, China
  • 35.
    Contributions SPrank: Semantic Path-basedranking  Combination of semantic item descriptions from the Web of Data and implicit feedback  Mining of the semantic graph using path-based features  Learning To Rank setting Future Work:  Deeper analysis of the path-based features  Usage of other Learning To Rank approaches RecSys 2013 – 7th ACM Conference on Recommender Systems October 12-16, 2013 Hong Kong, China
  • 36.
    Q&A A Little SemanticsGoes a Long Way. Hendler Hypothesis RecSys 2013 – 7th ACM Conference on Recommender Systems October 12-16, 2013 Hong Kong, China