By Jennifer Zaino November 19, 2008

With a three-year, $550,000 grant from the National Science Foundation, UT Dallas researchers will be delving into issues around the scalability of the semantic web, entity resolutions, and policy specification and reasoning.

The project, whose funding was announced in October, joins other semantic web research efforts taking place at UT Dallas funded by agencies such as the Intelligence Advanced Research Projects Activity, as well as collaborative efforts taking place with Raytheon Co. on creating semantic web technology that can find and analyze visual information.

The principal investigator for the project is Dr. Bhavani Thuraisingham, a professor of computer science in the Erik Jonsson School of Engineering and Computer Science at UT Dallas. UT Dallas has been working with HP Labs in developing the JENA RDF engine. “but one problem is managing large, large graphs,” explains Thuraisingham.

“It’s very well to talk about the semantic web doing this, that, and the other thing,” but representing all this information requires very large graphs, he said. That’s where UT Dallas’ expertise in data mining, one area of specialization for Thuraisingham’s colleague Latifur Khan, will come in handy.

The entity resolution problem is one that is capturing the attention of many researchers. When the same word has different interpretations, ontologies are required to sort out what the particular reference is to. She uses her own name as an example of potential confusion: Bhavani is the name of a ferocious goddess in India, a river, and a city or town, as well as her own name. That ambiguity needs to be resolved. While she can’t provide details at this point, she notes that UT Dallas is looking at some clustering techniques as a unique approach to dealing with these issues.

One of the reasons the government is interested in the semantic web is related to the issues of data sharing, springing from the 9/11 tragedy where information that might have helped target the terrorists in advance of the attacks had not been shared among different agencies. The semantic web opens the door to making the data that users want to get more available, useful and relevant, but what should be shared and policies around how that data should be shared will be an issue, whether it’s within government agencies or among parties such as health care practitioners, insurance providers, and patients, or even within various social networks.

对于第2点，我觉得工作有点类似University Washington Turing Center提出的Open Information Extraction以及Machine Reading，其中有一块就是要解决Web disambiguation。

对于第3点，因为自己不是做reasoning, rule和policy的，而且上面描述不是很详细，所以暂时没有什么感受和想法。:)