- Oggetto:
- Oggetto:
Trust modeling and user modeling to address social network overload through message recommendationTrust modeling and user modeling to address social network overload through message recommendation
- Oggetto:
Academic year 2014/2015
- Course ID
- SEM-TMUMA
- Year
- 1° anno 2° anno 3° anno
- Type
- Seminario
- Credits/Recognition
- 0
- Course disciplinary sector (SSD)
- INF/01 - informatica
- Delivery
- Tradizionale
- Language
- Inglese
- Attendance
- Facoltativa
- Type of examination
- Non prevista
- Oggetto:
Sommario del corso
- Oggetto:
Program
Abstract
In this talk I present two projects that develop artificial intelligence
approaches for assisting users to derive positive experiences in
today's current online communities. In particular, our online social
networks increasingly have massive numbers of peers and hence very
large numbers of potential messages to view. We demonstrate how it
is possible to present to users those messages with highest predicted
benefit. This is achieved through a combination of techniques
from the artificial intelligence subfields of trust modeling and
user modeling. The first project focuses on using an approach
motivated by POMDPs to learn, in a principled manner, those
messages that bring the highest utility to a user,
in environments where peers offer ratings of those messages.
This is achieved through observations of a message's rating,
the similarity of the message rater to the user and the
credibility of the message rater. We validate our approach
both through simulations and against realworld datasets from
Reddit and Epinions. The second project was a hands-on study
of a set of existing social networks (Coursera, Reddit
and Health Forums) which led to the development of a
generalized recommendation algorithm that can operate
regardless of social network. This is achieved by careful modeling
of the message, the author, the recipient and the network.
Our conclusion is that user modeling and trust modeling are
both important for effective management of social network information
and that artificial intelligence approaches show some promise
in assisting users in these online environments.Suggested readings and bibliography
- Oggetto:
Note
The seminar will be held by
Robin Cohen
David R. Cheriton School of Computer Science
University of Waterloo
Waterloo, Ontario, Canada- Oggetto: