Digital Government: A Language Modeling Approach to Metadata for Cross-Database Linkage and Search

University Of Massachusetts Amherst, Amherst MA

Investigators

Abstract

EIA-9983215 Croft, W. Bruce University of Massachusetts Ahmerst Digital Government: A Language Modeling Approach to Metadata for Cross-Databse Linkage and Search A crucial problem in government on-line services is that related information is often maintained in different systems across different agencies, thus making it difficult to integrate that information. Most attempts at "one-stop shopping" sites in the government fail for that reason; they must be maintained manually which does not work over time. On the other hand, data warehousing, where the data is cleaned, made conformant and housed in a central location, will not work due to the distributed nature of government-held information. One promising avenue to approach this problem would be through the use of automated systems which generate metadata (data about data) for various datasets, allowing them to be integrated more easily, while the original data remains in place at its home site. The work of this proposal will address this problem in a new way - language modeling. The government collaborators are the US Geological Survey, the US Department of Commerce, the General Services Administration and the US Library of Congress.