Information Technology and Federal Statistics
Organizer: Cathryn Dippo
(dippoc@ore.psb.bls.gov)
Bureau of Labor Statistics

Description:
The vast amount of statistical information compiled by Federal, state, and local governments, academic research institutions, and even some private enterprises which put their data into the public domain could be very useful if it were easy to find and understand. Data-based decision-making will not become ubiquitous in our work, school, and home environments until extensive multidisciplinary research is conducted on the design and development of useful, usable interfaces that support information seeking, retrieval, extraction, integration, and display of statistical information. Moreover, with quantitative literacy being so low in this country, significant efforts are needed to build data-based knowledge and learning tools that provide the expertise needed to convert statistical information into knowledge and enhance user understanding in order to aid intelligent decision-making.

As a first step towards creating a national statistical information infrastructure, a portal to Federal statistics www.fedstats.gov has been developed under the auspices of the Interagency Council on Statistical Policy with funding from the 14 member agencies. Currently, this site is just a gateway to the information available on the various individual agency sites. To extend the usefulness of the site, extensive R&D efforts are needed to apply cutting-edge information technologies in the statistical information domain and to invest in multidisciplinary research that tackles the complex issues surrounding the collection and use of statistical information.

Some efforts are currently underway with funding from the National Science Foundation's Digital Government Program. These projects include automated ontology building for the purpose of data integration, modernization of the authoring environment for computer-assisted survey instruments, development and testing of new table presentation paradigms specifically designed for web browsing (i.e., tables which incorporate the metadata needed to understand the numbers in the table), and query-history-based access to respondent-level information that insures confidentiality protection. In addition, through the membership of several Federal statistics agencies in the National Science Foundation Federal Consortium, the PACI partnerships are exploring areas where their information technology infrastructure research can aide in the development of a national statistical information infrastructure.

Format:
This session will emphasize the collaboration between computer and information scientists with statistical information providers. The presentations will spotlight work in progress and describe the challenges yet to be addressed, accenting the multidisciplary aspects of the research.

Participants:
Amarnath Gupta, Chaitanya Baru, Richard Marciano, and llya Zaslavsky, San Diego Supercomputer Center, University of California, San Diego (presentation, Wrapping and Mediating Survey Data)

Carol Hert (presentation, Statistical Information Seeking and System Design)
Carol Hert who is a faculty member in the School of Information Studies at Syracuse University. Professor of Library and Information Sciences, University of North Carolina. Her research focus is on user behavior--the interaction of users, systems, and the organizations that provide and support those systems. For several years, she has conducted user studies for several Federal government websites, including the FedStats, BLS, and Department of Education sites.

Ed Hovy and Judith Klavans (presentation, The Role of Ontologies in Statistical Information Seeking)
Ed Hovy heads the Natural Language Group at Information Sciences Institute-University of Southern California. He is responsible for the development of a robust automated text summarization system and of large concept taxonomies/ontologies. Judith Klavans is the Director of the Center for Research on Information Access at Columbia University. She is working on ways to automatically determine lexical information via statistical means over parsed texts. As recipients of an NSF Digital Government grant, they will be working on automated ontology building using natural language processing methods. The focus area for this project is gasoline, a subject for which the Energy Information Agency, the Bureau of Labor Statistics and the Census Bureau disseminate statistical information. In this talk, they will discuss the role of statistics in language processing and information retrieval, in general, and the status of their Digital Government research project.

Carol Hert (poster, Statistical Information Seeking and System Design)

Ed Hovy (poster, The Role of Ontologies in Statistical Information Seeking)

Peter Joftis (poster, The Data Documentation Initiative: Current Status of an Attempt to Specify an XML DTD for Empirical Social Science Documentation)
University of Michigan