Campus Event Calendar
Event Entry
New for: D1, D2, D3, D4, D5
What and Who
The Benefit of Adaptation in Knowledge Management Systems
Dr. Guillaume Bouchard
Talk
Dr. Guillaume Bouchard is a senior research scientist in data mining in
machine learning at Xerox Research Centre Europe. He joined XRCE in 2004
after receiving a PhD in Applied Mathematics from Institut National de
Recherche en Informatique et en Automatique (INRIA) and Université Joseph
Fourier. Since then, his work mainly focused on development of statistical
models for large data problems and the support to crowdsourcing platforms
for knowledge management.
AG 1, AG 2, AG 3, AG 4, AG 5, SWS, RG1, MMCI
Public Audience
English
Note: We use this to send email in the morning.
Date, Time and Location
Thursday, 2 February 2012
16:00
120 Minutes
E1 7 - cluster building
0.01
Saarbrücken
Abstract
> In Human-Computer Interaction, systems implementing Mixed Initiative are
> referring to interfaces which integrate human and automated reasoning to
> take advantage of their complementary reasoning styles and computational
> strengths. Thanks to the rapid growth of machine learning and information
> retrieval techniques which allow a quick adaptation of the quality of the
> content recommended to the users, the Mixed Initiative principle brings a
> real benefit in knowledge management applications. In the first part of the
> talk, two crowdsourcing systems implementing this principle will be
> presented: Mail2Wiki, a system that enables easy contribution and initial
> curation of content from the personal space of email to the shared
> repository of a wiki, and Innovation Cockpit, a Idea Management System that
> enables reviewers to easily cluster and score many ideas or suggestions
> written by the crowd (i.e. employees or citizens).
>
>
> In the second part of the talk, some recent advances in recommender systems
> will be presented: an heteroscedastic matrix factorization model which takes
> into account a per-user and per-item noise level, enabling the efficient
> identification of users or objects for which recommendation is easier than
> others, and a generic model to extend the concept of matrix and tensor
> factorization to any relational database. We show empirically that compared
> to classical recommender systems, models that take into account the
> relational structure of the data significantly improves the quality of
> recommendations in some tagging and rating experiments.
> referring to interfaces which integrate human and automated reasoning to
> take advantage of their complementary reasoning styles and computational
> strengths. Thanks to the rapid growth of machine learning and information
> retrieval techniques which allow a quick adaptation of the quality of the
> content recommended to the users, the Mixed Initiative principle brings a
> real benefit in knowledge management applications. In the first part of the
> talk, two crowdsourcing systems implementing this principle will be
> presented: Mail2Wiki, a system that enables easy contribution and initial
> curation of content from the personal space of email to the shared
> repository of a wiki, and Innovation Cockpit, a Idea Management System that
> enables reviewers to easily cluster and score many ideas or suggestions
> written by the crowd (i.e. employees or citizens).
>
>
> In the second part of the talk, some recent advances in recommender systems
> will be presented: an heteroscedastic matrix factorization model which takes
> into account a per-user and per-item noise level, enabling the efficient
> identification of users or objects for which recommendation is easier than
> others, and a generic model to extend the concept of matrix and tensor
> factorization to any relational database. We show empirically that compared
> to classical recommender systems, models that take into account the
> relational structure of the data significantly improves the quality of
> recommendations in some tagging and rating experiments.
Contact
Dr. Ivan Titov
--email hidden
passcode not visible
logged in users only
Mona Linn, 02/02/2012 10:59 -- Created document.