Summary: The IIT Intranet Mediator: An Overview
D. Grossman, S. Beitzel, E. Jensen, and O. Frieder
Information Retrieval Laboratory
Illinois Institute of Technology
A mediator sits between a user and a variety of data sources. It acts on behalf of the user to
provide "one stop shopping" to an organization's data. An intranet, or digital library, is an
excellent example of the need for a mediator because of the myriad of data types that exist in the
digital library. A user query such as "What are the three best restaurants in Chicago?" should
not search for the word three. A mediator will recognize that this query needs to access
structured information and will pose the query as a conventional SQL query to a data warehouse
of structured information. The mediator may submit a request to unstructured sources as well.
The result will be both a suggested answer to the user's question as well as a set of related links to
other potentially useful information. We note that our key goal is simply to answer natural
language questions that cross both structured and unstructured data. Presently, work was done on
the TREC Question Answering track that focuses on unstructured sources [Hara00]. Our work
builds on this as we wish to include structured sources as well.
Integrating Structured Data and Text
Our group has worked in the area of integrating structured data and text since the early 1990's.