3/18/99 
Attendees: Rajeev, Sebastien, Dick, Shinji, Yue

Sebastien and Yue presented some preliminary results from the system they
have implemented.  The system is tentatively called SMILE (Structure
MIning and Logic Extraction).  The objective of SMILE is to convert plain
text documents into XML documents.

For the current implementation, SMILE can only parse the input file into
two nested levels, one identified by key words and the second identified
by blank lines.  The input files of SMILE are currently plain text
resumes, and the outputs are the same documents represented in XML format.
The output XML documents are then displayed using IE5.0 browser.

We agreed that the results are interesting but still preliminary. More
heuristics may be used to help identifying nested structures, for example,
we may use bullets or commas. Because the key words used by SMILE are
results of a pre-processing step, the system is flexible enough to process
different types of documents with different keyword sets. 

We also discussed the problem of identifying nested structures in general.
The problem is described in http://itgserv/~yuez/reports/interesting.doc.
Rajeev suggested that the general problem is NP-hard, the approximate
problem may not be simpler, but we will think more about it.