hybrid semantic and fuzzy approaches to context-aware personalisation

24
HYbrid semantic and fuzzy approaches to context-aware PERsonalisation Valentin Grouès Supported by the National Research Fund, Luxembourg 1

Upload: kiss-tudor

Post on 31-Oct-2014

979 views

Category:

Education


0 download

DESCRIPTION

Presentation by Valentin Groues Semantic Technologies Seminar, Tudor Research Centre, Luxembourg, 21/03/2011

TRANSCRIPT

Page 1: HYbrid semantic and fuzzy approaches to context-aware PERsonalisation

HYbrid semantic and fuzzy approaches to context-aware PERsonalisation

Valentin Grouès

Supported by the National Research Fund, Luxembourg1

Page 2: HYbrid semantic and fuzzy approaches to context-aware PERsonalisation

2

Title: HYbrid semantic and fuzzy approaches to context-aware PERsonalisation

Supervisors: Dr Yannick Naudet (CRPHT) - Ph.Dr Odej Kao (TuB)

Hypothesis: The perceived results of personalisation systems can be improved by combining the reasoning capabilities given by Semantic Web technologies and the representation of human imprecisions through fuzzy theory.

HyPer

Page 3: HYbrid semantic and fuzzy approaches to context-aware PERsonalisation

Recommender Systems

How to look for a needle in a haystack?

Just use the appropriate tool

3

How to filter and find the needed information in a perpetually growing amount of data? Recommender systems aim at providing personalised suggestions about items, actions or content considered of interest to the user

Page 4: HYbrid semantic and fuzzy approaches to context-aware PERsonalisation

Recommender Systems

Content-based recommender sytems:

– recommend items similar to those the user has previously liked/experienced

Limitations:- over-specialisation

- new user problem

- requires good description of items

Limitations:- over-specialisation

- new user problem

- requires good description of items

Advantages:- no cold start for new items

- doesn’t require many users, can work in a one user environment.

- can provide explanations

Advantages:- no cold start for new items

- doesn’t require many users, can work in a one user environment.

- can provide explanations

4

Page 5: HYbrid semantic and fuzzy approaches to context-aware PERsonalisation

Recommender Systems

Collaborative Filtering (Amazon, Netflix, etc.) 1. Look for users who share a similar rating pattern to that of the active user

2. Use the ratings from like-minded users found in step 1 to calculate a prediction for a given item.

Limitations:- new user and new item problem

(cold start)- sparsity problem- grey sheep problem- non diversity problem - not suitable for items sold only

once

Limitations:- new user and new item problem

(cold start)- sparsity problem- grey sheep problem- non diversity problem - not suitable for items sold only

once

Advantages:- no need for item description

- almost solves the over-specialisation problem of CBF

- good precision

- low-cost capture of complex taste mechanisms

Advantages:- no need for item description

- almost solves the over-specialisation problem of CBF

- good precision

- low-cost capture of complex taste mechanisms

5

Page 6: HYbrid semantic and fuzzy approaches to context-aware PERsonalisation

Recommender Systems

Hybrid Systems Demographic Filtering (DMF):

– Categorizes the user based on his/her profile to provide recommendations based on demographic clusters.

– The user will be recommended items similar to the ones other members of the same demographic characteristics liked.

Knowledge-based recommender:–Use a priori domain knowledge to match user requirements

with the properties of items. This approach uses explicit models of both the users and the products being recommended.

6

Page 7: HYbrid semantic and fuzzy approaches to context-aware PERsonalisation

How to improve Recommender Systems?

1. Better methods for representing user behavior and information about items

2. Focusing on generating an accurate list of recommendation rather than a list full of individually accurate recommendations

3. Incorporation of contextual information into recommendation process

4. Development of less intrusive and more flexible recommendation methods, explanations

5. Development of recommender system effectiveness measures

=> Semantics + Context-awareness + Fuzzy SetsAdomavicius, G. and Tuzhilin, A. Toward the next generation of recommender systems: a survey of the state-of-the-art and possible extensions. IEEE Transactions on Knowledge and Data Engineering 17, 6 (2005), 734-749.Lops, G. , Gemmis, M., Semeraro, G. Content-based Recommender Systems: State of the Art and Trends, in Recommender Systems Handbook, 2010

7

Page 8: HYbrid semantic and fuzzy approaches to context-aware PERsonalisation

Need for Semantics

Semantic ambiguity:

User: u=(Indonesia=0.7;Java=0.9;island=0.2)

Items: d1=(Java=0.4;hotel=0.8), d2=(Java=0.4;software=0.8)

pref(d1,u)=pref(d2,u)=0.19

Distinction between the two concepts is essential for not producing undesirable recommendations

islandprogramming

language

Iván Cantador. "Exploiting the Conceptual Space in Hybrid Recommender Systems: a Semantic-based Approach". 2008, Madrid

8

Page 9: HYbrid semantic and fuzzy approaches to context-aware PERsonalisation

Need for Semantics

Assumption of terms independance:

User: u=(Indonesia=0.7;Java=0.9;island=0.2)

Items: d1=(Java=0.4;hotel=0.8), d2 =(Java=0.4;archipelago=0.8)

pref(d1,u)=pref(d2,u)=0.19

Iván Cantador. "Exploiting the Conceptual Space in Hybrid Recommender Systems: a Semantic-based Approach". 2008, Madrid

island island

Assumption of terms independance:

User: u=(Indonesia=0.7;Java=0.9;island=0.2)

Items: d1=(Java=0.4;hotel=0.8), d2 =(Java=0.4;archipelago=0.8)

Semantic relations between concepts have to be considered

9

Page 10: HYbrid semantic and fuzzy approaches to context-aware PERsonalisation

Context awareness

Mobile environment Different situations can correspond to different needs Geographical location, time of day, weather, etc.

10

Page 11: HYbrid semantic and fuzzy approaches to context-aware PERsonalisation

Fuzzy Sets and Fuzzy Logic

To represent imprecise information inherent to the human way of thinking

Humans have a tendency to use imprecise concepts for claiming tastes: “cheap restaurant”, “long movie”, “young actor”, etc.

Limitations of crisp systems:– For a user willing to find a restaurant with a cost up to 20€ the system

will equally discard a restaurant costing 21€ as a restaurant costing 300€.

a user would prefer having an answer proportional to the distance between his ideal preference and the recommended content

11

Page 12: HYbrid semantic and fuzzy approaches to context-aware PERsonalisation

What are Fuzzy Sets?

12

Page 13: HYbrid semantic and fuzzy approaches to context-aware PERsonalisation

Our previous research

Million Dollar Baby recommended Unforgiven and The Good, the Bad and the Ugly discarded (westerns)

Naudet, Y., Aghasaryan, A., Toms, Y., & Senot, C. (2008). An Ontology-Based Profiling and Recommending System for Mobile TV. 2008 Third International Workshop on Semantic Media Adaptation and Personalization (pp. 94-99). IEEE.Mignon, S., Groues, V., and Naudet, Y. Advanced Personalisation by Ontologies: Audiovisual Content Filtering on Mobile Devices. Journées Francophones des Ontologies, (2008).Naudet, Y., Mignon, S., Lecaque, L., Hazotte, C., and Groues, V. Ontology-Based Matchmaking Approach for Context-Aware Recommendations. AXMEDIS, (2008).

13

Page 14: HYbrid semantic and fuzzy approaches to context-aware PERsonalisation

Semantic similarity measures

How to compare two instances? Medor and Felix have some similarities:

Common parent, both Mammals Similar properties, both 4 legs and same owner

sim(Medor,Felix)=?

14

Page 15: HYbrid semantic and fuzzy approaches to context-aware PERsonalisation

Integrating fuzzy sets within ontologies

FuSOR: A model for representing fuzzy sets and linguistic values within ontologies (Y. Naudet, V. Grouès, M. Foulonneau,

Introduction to Fuzzy-Ontological Context-Aware Recommendations in Mobile Environments, APRESW 2010)

15

Page 16: HYbrid semantic and fuzzy approaches to context-aware PERsonalisation

FuSor: Characteristics of the approach

Can be used as an extension of an ontology without requiring any modifications, OWL DL compliant

Allows using fuzzy sets and their membership functions for any datatype property

Supports context and domain dependency

16

Page 17: HYbrid semantic and fuzzy approaches to context-aware PERsonalisation

Ex: Describing interest boundaries

Membership functions can be used to define the way a user interest deviates from an “ideal” value.

Ex: “I am looking for a restaurant with prices up to 20€ but I could accept up to 25€ even if I would be less satisfied”.

17

Page 18: HYbrid semantic and fuzzy approaches to context-aware PERsonalisation

eFoaf

Cover demographic and basic user information

Context aware (e.g. not only one contact address)

Simple and complex interests associated with a context of validity

Open to external RDF datasets

Skills, abilities and handicaps

18

Page 19: HYbrid semantic and fuzzy approaches to context-aware PERsonalisation

An application to transport

• Personalisation of carpooling solutions:

– Match carpoolers based on their profiles, their expectations:

• music tastes• child seat• animals• smoking allowed?

19

Clairvoyant

Clairvoyant

Page 20: HYbrid semantic and fuzzy approaches to context-aware PERsonalisation

An application to transport

• Personalisation of itineraries based on:

– Preferences between means of transportation– User priorities:

• Cost• Time

• CO2 footprint

• Touristic interest

20

Clairvoyant

Clairvoyant

Page 21: HYbrid semantic and fuzzy approaches to context-aware PERsonalisation

An application to transport

• Recommendation in case of an unforeseen event:

– Find an alternative itinerary– Recommendations based on user profiles:

• Hotels• Restaurants• Museum

21

Clairvoyant

Clairvoyant

Page 22: HYbrid semantic and fuzzy approaches to context-aware PERsonalisation

What’s done?

FuSOR: an approach to extend existing model to use fuzzy sets eFoaf: an extension of foaf to represent rich user profiles and

preferences Prototype of recommender system making use of semantics and

fuzzy sets

22

Page 23: HYbrid semantic and fuzzy approaches to context-aware PERsonalisation

What’s next?

Explore other uses of fuzzy sets and fuzzy logic for recommendations: – fuzzy sets for item description (this movie belongs to the action

genre with a degree of 0.7, this movie is long) Use the list of items liked by the user, history of consumption Further development of a prototype applied to a particular use

case (job ads, movies, restaurants) Performance optimisation: distributed computing, caching

mechanisms and different semantic web libraries Evaluations

23

Page 24: HYbrid semantic and fuzzy approaches to context-aware PERsonalisation

Thank you for your attention!

24

Any questions ?