UIUC

Abstract:

With the advent of Web 2.0, more and more data is being produced and consumed online. Individuals, especially from the younger generation, are getting informed and building their opinions based on what they watch, read, and discuss on the Web. At the same time, it is also easy for anyone, even nefarious agents, to generate (dis)information and make it widely available. In this scheme of things, can you trust all the content that is available online?
In this talk, I will present some of the challenges in computing trustworthiness of free text claims and argue the need to go beyond structured, extraction-centric approaches to unstructured, textual evidence-driven trust models. Building trustworthiness models for textual claims involves understanding the different ways in which claims can be expressed in free text, aggregating weak signals based on the quality of evidence, addressing differential trust levels of sources (experts vs. laymen, commercial vs. governmental websites, etc.), and exploring the use of community knowledge (expressed in forums) to enable Web users decide credibility of sources and claims. In addition, it is also instructive to study how users perceive the information presented to them; and how to build systems that encourage users to get an unbiased opinion about controversial topics. I will present some of my work in these directions.

Speaker bio:
Vinod is a doctoral candidate at the University of Illinois at Urbana-Champaign, working with Prof.ChengXiang Zhai and Prof.Dan Roth. His research interests include text informatics, natural language processing, machine learning, and information extraction.