MIME-Version: 1.0 Server: CERN/3.0pre6 Date: Wednesday, 20-Nov-96 19:31:55 GMT Content-Type: text/html Content-Length: 3673 Last-Modified: Thursday, 14-Mar-96 20:02:11 GMT
I work on the CSTR project, an ARPA sponsored effort to make computer science technical reports more easily accessible. As part of that work, I designed a distributed technical report server which is now running at many universities.
I am interested in Corporate (or Group) Memory, meaning electronic systems for capturing and accessing the knowledge used and produced by (the workers of) an institution, in order to increase the quality of or reduce the time required to do future work. Corporate memory includes not just the intellectual products of the institution (e.g. an engineer's designs, a lawyer's contracts, an author's screenplays) but also knowledge about the process that produced the product: knowledge of dead-ends explored, tools used, and justifications supporting the final decisions.
I have also begun a project (with Dan Huttenlocher) in developing corporate memory through shared annotation of structured documents. This project investigates how people can share information by reading and writing annotations in electronic documents shared by the group. An initial prototype implementation is being used by Cornell class CS212. Here, the shared documents are problem sets and course notes, and a nnotations might be requests for clarifications by students, technical questions, or corrections made by staff. The question is whether this will be a useful means for students to obtain answers, whether students will find each other's questions a useful source for learning, whether students will often be able to answer each other's questions (correctly), and whether the course staff will find this a useful means for feedback in improving the course. So far, the evidence is that they do.
I'm also interested in natural language generation and the design of computational proxies ("agents") which can safely and reliably carry out remote computations on foreign machines without risk to either you or the owner of the remote machine.