Towards Deeper Analysis of User Generated Content: Adding Semantics and Topics to Profiles, Links and Communities

Speaker:	Dr. Sofus A. Macskassy
		Director of Fetch Labs
		Fetch Technologies

Title:		"Towards Deeper Analysis of User Generated Content:
		 Adding Semantics and Topics to Profiles, Links and
		 Communities"

Date:		Wednesday, 16 Feb 2011

Time:		2:00pm - 3:00pm

Venue:		Rm3301A (via lifts 17/18), HKUST


Abstract:

The amount of data available from user generated content, such as social
media content produced by Twitter and various Blog sites, is growing at a
phenomenal rate. This dynamic environment presents a large number of
unique and very difficult challenges in the areas of information
management, extraction, aggregation, integration and analysis, and all
related sub-fields. Some of the key challenges in this environment center
around the problem of understanding the dynamics of the information in
terms of users and communities: what are the hot topics, what is new, what
is of interest to a user or a community, what are the communities, who are
key people or key information providers, what web-sites cater to what
demographic, ... I argue that the majority of these tasks require that
one first needs to identify communities and understand the specific
interests of users and communities. I further argue that links created in
social media are created for a variety of reasons and one cannot identity
a community by only looking at the links nor by treating all links the
same. I will in this talk explore two tasks: categorizing users and
identifying communities. Users generates lot of content and we can learn
a lot about users based on this content. This can help identify
communities by shared interest, regardless of linking behavior. Secondly,
we can use this content to understand the context of specific links,
thereby helping us to more clearly differentiate communities by taking
this context into account. I will show we can leverage this information to
gain much better insight into social media and how this can be used in a
variety of applications.


******************
Biography:

Dr. Sofus A. Macskassy is the Director of Fetch Labs at Fetch Technologies
and an Adjunct Professor at the University of Southern California. He was
previously a Research Scientist in the Information Sciences department at
Stern School of Business, New York University, where he worked in domains
such as financial news and counter-terrorism. His main research areas
include statistical relational learning, data mining, text mining and
social network analysis. Dr. Macskassy is the developer of the open-source
Network Learning Toolkit (NetKit-SRL), a machine learning toolkit for
networked data. He received his PhD in Computer Science from Rutgers
University in January 2003.