1. Learning to Question: Leveraging User Preferences for Shopping Advice
Date : 2013/12/11
Author : Mahashweta Das, Aristides Gionis,
Gianmarco De Francisci Morales,
and Ingmar Weber
Source : KDD’13
Advisor : Jia-ling Koh
Speaker : Yi-hsuan Yeh

2. Outline
Introduction
Method
Experiments
Conclusion

3. Introduction
Motivation
Customers shop online, from their homes, without any human interaction involved.
Catalogs of online shops are so big and with so many continuous updates that no human, however expert, can effectively comprehend the space of available products.
Use a flowchart asks the shopper a question, and the sequence of answers leads the shopper to the suggested shopping option.

4. Introduction
SHOPPINGADVISOR is a novel recommender system that helps users in shopping for technical products.
car

5. Introduction
SHOPPINGADVISOR generates a tree-shaped flowchart, in which the internal nodes of the tree contain questions involve only attributes from the user space.
non-expert users can understand easily.

6. Introduction Find the best user attribute to ask at each node.
How to learn the structure of the tree, i.e., which questions to ask at each node.
Find the best user attribute to ask at each node.
This paper focus on identifying the attribute of interest, and not on the task of formulating the question in a human interpretable way.
How to produce a suitable ranking at each node.
Learning-to-rank approach

7. Outline Introduction Method Experiments Conclusion
LEARNSATREE algorithm
Experiments
Conclusion

8. LEARNSATREE algorithm
Table U (user)
attributes
users
Table P (product)
Table R (review)

9. User attributes Car (from Yahoo! Autos) Camera (form Flickr)
Ex：fuel economy, comfortable interior, stylish exterior
Camera (form Flickr)
Photo’s tag topic
Ex：food topic (tags：fruit, market)

10. Problem definition Build tree Rank products node 𝑞 A user attribute 𝛼
Top-k list of product recommendations

11. Learning product rankings
RANKSVM
Goal：Learn a weight vector 𝑤= 𝑤 1 , …, 𝑤 𝑚𝑝 for the 𝑚 𝑝 technical attributes of the products 𝑃
A > B
B > C
B > D .
RANKSVM
model A B D C
features
Product’s technical attributes

12. a1 a2 a3 a4 a5
Product A 1
Product B
𝑤= 0.2, 0.1, 0.5, 0.1, 0.1
rank(A) = =0.9
rank(B) = =0.3

13. Learning the tree structure
Goal：determine the best user attribute “𝛼” to split 𝑈 𝑞 at node 𝑞
𝑠𝑢𝑚

14. 𝑝 1 𝑝 2 𝑝 3 𝑝 1 𝑝 3 𝑝 2 Example： Correctly-rank： 𝑝 1 > 𝑝 2 >𝑝 3
System result
System result
eval(rank) = 2∗3 3∗(3−1) =1
eval(rank) = 2∗2 3∗(3−1)
=0.66
𝑝 1
𝑝 2
𝑝 3
𝑝 1
𝑝 3
𝑝 2
( 𝑝 1 , 𝑝 2 ), ( 𝑝 1 , 𝑝 3 ), ( 𝑝 2 , 𝑝 3 )
( 𝑝 1 , 𝑝 3 ), ( 𝑝 1 , 𝑝 2 ), ( 𝑝 3 , 𝑝 2 )

15. user
attribute 𝛼
𝑈 𝑞 𝛼
𝑈 𝑞 𝛼
Review table 𝑅
node 𝑞
split
user
𝑈 𝑞

16. Rank list RANKSVM Count payoff
product A B D C . F E
Rank list
RANKSVM
Count payoff
Consider all possible user attributes 𝛼, and choose as splitter the one that maximizes the pay-off.

17. Stopping criterion Grow the tree to its “entirety” Post-pruning
If a node’s child node is split by the “near-synonomous” tag trim the child node
Example:
travel
vacation
Employ pruning rules on the validation set.

18. Outline
Introduction
Method
Experiments
Conclusion

19. Datasets Car datasets Camera datasets Synthetic datasets
Yahoo! Autos
606 cars, 60 attributes
2180 reviews
2180 user, 15 tags (as attributes)
Ex：fuel economy, comfortable interior, stylish exterior
Camera datasets
Flickr tags
645 cameras (CNET)
11468 reviews
5647 user, 25 topic tags (as attributes)
Ex：food topic (tags：fruit, market)
Synthetic datasets
200 products, 4000 comments, 1000 users

20. Experiment setup SHOPPINGADVISOR RANKSVM k-NN SA.k-NN Author’s method
The ranked list returned by SHOPPINGADVISOR at the root
k-NN
k-nearest neighbors algorithm
SA.k-NN
Features are selected from SHOPPINGADVISOR

21. Quality evaluation A B D 25 topics 12 topics.
System result ranking list
average MRR
If user prefer “B”
 1 𝑟𝑎𝑛𝑘 𝑖 = 1 2

22. Performance evaluation

23. Outline
Introduction
Method
Experiments
Conclusion

24. Conclusion
Proposed a novel recommender system, SHOPPINGADVISOR, that helps users to shop for technical products.
SHOPPINGADVISOR leverages both user preferences and technical product attributes in order to generate its suggestions.
At each node, SHOPPINGADVISOR suggests a ranking of products matching the preferences of the user.
Compared with a baseline, and demonstrated the effectiveness of the approach.