Collaborative Filtering Alejandro Bellogn 1 , Jun Wang 2 , and Pablo - - PowerPoint PPT Presentation

collaborative filtering
SMART_READER_LITE
LIVE PREVIEW

Collaborative Filtering Alejandro Bellogn 1 , Jun Wang 2 , and Pablo - - PowerPoint PPT Presentation

Text Retrieval Methods for Item Ranking in Collaborative Filtering Alejandro Bellogn 1 , Jun Wang 2 , and Pablo Castells 1 1 Escuela Politcnica Superior, Universidad Autnoma de Madrid 2 Department of Computer Science, University College


slide-1
SLIDE 1

European Conference on Information Retrieval 2011 April 20, Dublin, Ireland

Text Retrieval Methods for Item Ranking in Collaborative Filtering

Alejandro Bellogín1, Jun Wang2, and Pablo Castells1

1Escuela Politécnica Superior, Universidad Autónoma de Madrid 2Department of Computer Science, University College London

@abellogin alejandro.bellogin@uam.es

slide-2
SLIDE 2

European Conference on Information Retrieval 2011 April 20, Dublin, Ireland

Collaborative Filtering – in a glimpse

  • Input: a rating matrix
slide-3
SLIDE 3

European Conference on Information Retrieval 2011 April 20, Dublin, Ireland

Collaborative Filtering – in a glimpse

  • The users
slide-4
SLIDE 4

European Conference on Information Retrieval 2011 April 20, Dublin, Ireland

Collaborative Filtering – in a glimpse

  • The items
slide-5
SLIDE 5

European Conference on Information Retrieval 2011 April 20, Dublin, Ireland

Collaborative Filtering – in a glimpse

  • A rating: from a user towards an item
slide-6
SLIDE 6

European Conference on Information Retrieval 2011 April 20, Dublin, Ireland

Collaborative Filtering – in a glimpse

  • User profile
slide-7
SLIDE 7

European Conference on Information Retrieval 2011 April 20, Dublin, Ireland

Collaborative Filtering – in a glimpse

  • Item profile
slide-8
SLIDE 8

European Conference on Information Retrieval 2011 April 20, Dublin, Ireland

Collaborative Filtering – in a glimpse

  • Unknown rating
slide-9
SLIDE 9

European Conference on Information Retrieval 2011 April 20, Dublin, Ireland

Collaborative Filtering – in a glimpse

  • Goal

If user Boris watched Love Actually, how would he rate it?

slide-10
SLIDE 10

European Conference on Information Retrieval 2011 April 20, Dublin, Ireland

Collaborative Filtering – in a glimpse

  • Prediction: how Boris rated similar items
slide-11
SLIDE 11

European Conference on Information Retrieval 2011 April 20, Dublin, Ireland

Text Retrieval

Query Process Text Retrieval Engine Output Inverted Index Term

  • ccurrences

(term-doc matrix) Query

slide-12
SLIDE 12

European Conference on Information Retrieval 2011 April 20, Dublin, Ireland

Collaborative Filtering?

User Profile Process Item Similarity Text Retrieval Engine Output Inverted Index User Profiles (User-item matrix) User profile (as query)

slide-13
SLIDE 13

European Conference on Information Retrieval 2011 April 20, Dublin, Ireland

In this work

  • A first attempt
  • Item ranking
  • Item-based CF
  • Good results
  • Improvements – current work
  • More models
  • Rating prediction
  • Now… algorithmic details
slide-14
SLIDE 14

European Conference on Information Retrieval 2011 April 20, Dublin, Ireland

Text Retrieval

  • In (Metzler & Zaragoza, 2009)
  • In particular: factored form

   

 

, , ,

t g q

s q d s q d t

 

     

1 2

, , , , s q d t w q t w d t 

slide-15
SLIDE 15

European Conference on Information Retrieval 2011 April 20, Dublin, Ireland

Text Retrieval

  • Examples
  • TF:
  • TF-IDF:
  • BM25:

       

1 2

, qf , tf , w q t t w d t t d  

         

1 2

, qf , tf , log df w q t t N w d t t d t          

                     

 

 

3 1 3 1 2 1

1 qf , qf df 0.5 1 tf , , log df 0.5 1 dl / dl tf , k t w q t k t N t k t d w d t t k b b d t d                    

slide-16
SLIDE 16

European Conference on Information Retrieval 2011 April 20, Dublin, Ireland

Collaborative Filtering

  • Standard item-based formulation (Adomavicius & Tuzhilin 2005)

       

sim , rat , rat , sim ,

u u

j I j I

i j u i u j i j

 

  

slide-17
SLIDE 17

European Conference on Information Retrieval 2011 April 20, Dublin, Ireland

Collaborative Filtering

  • Standard item-based formulation (Adomavicius & Tuzhilin 2005)

       

sim , rat , rat , sim ,

u u

j I j I

i j u i u j i j

 

  

slide-18
SLIDE 18

European Conference on Information Retrieval 2011 April 20, Dublin, Ireland

Collaborative Filtering

  • Standard item-based formulation
  • More general

   

 

   

 

1 2

rat , , , , ,

j g u j g u

u i f u i j f u j f i j

 

 

 

       

sim , rat , rat , sim ,

u u

j I j I

i j u i u j i j

 

  

slide-19
SLIDE 19

European Conference on Information Retrieval 2011 April 20, Dublin, Ireland

Text Retrieval for Collaborative Filtering

  • In item-based Collaborative Filtering
  • Apply different models
  • With different normalizations and norms: sqd, L1 and L2

       

tf , sim , qf rat , t d i j t u j   t j d i q u    sqd

Document No norm Norm ( /|D|) Query No norm

s00 s01

Norm ( /|Q|)

s10 s11

slide-20
SLIDE 20

European Conference on Information Retrieval 2011 April 20, Dublin, Ireland

Text Retrieval for Collaborative Filtering

  • TF L1 s01 equivalent to item-based CF

       

sim , rat , rat , sim ,

u u

j I j I

i j u i u j i j

 

 

     

 

     

   

1 2

tf , , , , qf tf ,

t g q t g q t g q

t d s q d w q t w d t t t d

  

 

  

       

tf , sim , qf rat , t d i j t u j  

slide-21
SLIDE 21

European Conference on Information Retrieval 2011 April 20, Dublin, Ireland

Results

  • Movielens 1M
  • Movielens100k: equivalent results
  • TF L1 s01 equivalent to item-based CF (baseline)

0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35 0.40 TF L1 s01 TF-IDF L1 s01 TF-IDF L2 s11 BM25 L2 s11 TF L1 s10 BM25 L1 s01 TF-IDF L1 s10 BM25 L1 s00 TF-IDF L2 s10 TF-IDF L1 s00 TF L2 s10 BM25 L2 s10 TF L1 s00 BM25 L1 s11 BM25 L1 s10 BM25 L2 s01 TF L2 s11 TF-IDF L2 s01 TF L2 s01

nDCG

slide-22
SLIDE 22

European Conference on Information Retrieval 2011 April 20, Dublin, Ireland

Conclusions

  • It is possible to use Text Retrieval methods in rating-based

Collaborative Filtering

  • Our methods outperform classic Collaborative Filtering

methods

  • …Questions?
slide-23
SLIDE 23

European Conference on Information Retrieval 2011 April 20, Dublin, Ireland

References

  • Adomavicius, G., Tuzhilin, A.: Toward the next generation of

recommender systems: a survey of the state-of-the-art and possible

  • extensions. IEEE TKDE 17(6), 734-749 (2005)
  • Metlzer, D., Zaragoza, H.: Semi-parametric and non-parametric term

weighting for information retrieval. LNCS, vol. 5766, pp. 42-53. Springer (2009)