Loading

• EVENTS
• Upcoming
• Calendar
• Colloquia
• TCSDLS
• Student Talks
• Research Seminars
• inDuke Seminars
• Search
• Archive
• 2013-2014 2012-2013 2011-2012 2010-2011 2009-2010 2008-2009 2007-2008 2006-2007 2005-2006 2004-2005 2003-2004 2002-2003 2001-2002 2000-2001 1999-2000 1998-1999 1997-1998 1996-1997 1995-1996 1994-1995

Claims of "fact'' are made from data constantly -- by journalists, politicians, lobbyists, public relations specialists, sports aficionados, etc. Database research has in the past focused on how to answer queries, but has not devoted much attention to discerning the quality of the resulting claims, or to formulating good queries from the outset. This work seeks to advance the understanding of what makes for a high- quality claim based on data, and how to find queries that lead there.

We choose public interest journalism as our main target application domain. Data-driven claim-checking and lead-finding are increasingly important journalistic tasks, as more and more data become publicly available. Taking advantage of data availability, we hope to reduce cost, increase effectiveness, and broaden participation for public interest journalism, by putting practical computational tools for these data-driven investigative tasks in the hands of journalists and citizens alike.

In this talk, I will present a framework for modeling parameterized claims and perturbation in claim parameters. I will show how we define various quality measures and two interesting claim-checking problems within the framework. I will also briefly summarize our work on lead-finding and propose several directions for future work.

Advisor(s): Jun Yang & Pankaj Agarwal

Ashwin Machanavajjhala, Cong Yu