Students Selected to Present Research at DHS Summit

4/28/2009

Posters selected by DHS from nationwide competition

Written by

Two University of Illinois computer science graduate students, Yizhou Sun and Quang Do, are presenting posters of their research at the Third Annual U.S. Department of Homeland Security (DHS) University Network Summit on March 16-18, 2009. Both students are sponsored by the Multimodal Information Access and Synthesis (MIAS) Center directed by Professor Dan Roth. Their posters were selected by DHS from a nationwide competition that included hundreds of graduate students from across the country. The DHS is fully funding their trip to Washington, D.C. and their posters will be recognized at the Student Days event of the DHS University Network Summit.

Yizhou Sun
Yizhou Sun will present her work on RankClus: Integrating Clustering with Ranking for Heterogeneous Information Network Analysis. In many applications, there exists a large number of individual agents or components interacting with a specific set of components, and together they form large, interconnected, and sophisticated networks. We refer to such interconnected networks as information networks, with examples including the Internet, terrorist cells, research collaboration networks, public health systems, social networks, and so on. Clearly, information networks are ubiquitous and form a critical component of modern information infrastructure. Among them, the heterogeneous network is a special type of network that contains objects of multiple types.

Yizhou addressed the problem of generating clusters for specified types of objects, as well as ranking information for all types of objects based on these clusters in a multi-typed (i.e., heterogeneous) information network. A novel clustering framework called RankClus has been developed that directly generates clusters integrated with ranking. Her experimental results demonstrate that RankClus can generate more accurate clusters and in a more efficient way than the state-of-the-art link-based clustering methods. Moreover, the clustering results with their associated ranks can provide more informative views of data, as compared with traditional clustering.

Quang Do
Quang Do will present his work on Relation Detection Towards Textual Inference. Many natural language processing applications require semantic inference which, in turn, demands a common framework for applied semantics. Recently, textual entailment has shown considerable promise for textual inference. In this project, Quang develops a novel approach that uses Wikipedia as the background knowledge to detect the relations between entities in real-time. He positions the problem of relation detection in the context of the textual entailment task. This system can be considered as a black box that accepts two input entities and returns the relation between the entities along with their possible classes. The approach can discover whether two input entities pose an equivalence relation (synonym), alternation relation (sibling), forward entailment (child), backward entailment (parent), or independence (no relation). The goal of this project is to provide an effective solution for detecting relations between pairs of given entities at run-time. Although Quang's approach is designed to be applied directly to the textual entailment task, it can also be applied to many other problems in natural language processing, such as information extraction, question answering, and so forth.


Share this story

This story was published April 28, 2009.