B552 Project Idea Exchange Page

1. Code Transformation Project.
   • Contact: Aaron H.
   • Overview: Development of a case-based reasoning system for the transformation of s-expressions to other s-expressions. The problem domain will include the transformation of Scheme code from one implementation dialect to another, and may include a component dealing with natural language. Major problems include adaptation, efficient case retreival, and extracting relevant and useful pattern templates from the provided examples.
   • Background: Currently, Descot is a decentralized code distribution framework designed to facilitate the disemination and retreival of library metadata. It provides the underlying communication framework that enables communities without central authority to coöperate and integrate individual tools for library management together with multiple repositories. It also enables inter-repository synchronization and reference. It utilizes technology from the Semantic Web to represent and query library metadata.

     However, in a community like Scheme, many dialects exist which may require subtle, non obvious program transformations that do not lend themselves to easy rule-based programming. This project, therefore, intends to extend Descot with a component for library translation between dialects.

   • Possible additions: It is possible that some additional work on this will be done to test the system by using natural language. This will likely not be a component that falls under the surpervision of this class.
2. Web Mining, Semantic Web project.
   • Contact: Aaron H. (he doesn't have time to pursue this but would be happy to be a resource)
   • Overview: Develop a web/Usenet crawler that identifies libraries of Scheme code, collects and derives a semantic descriptions of these using Descot's metalanguage [1] and collects them for insertion into a library store.
   • Background: Descot is a semantic framework for communicating and interfacing between and to library stores (databases). It uses an RDF store as the basic server and enables shared interaction for library distribution in a decentralized, tool, and server independent manner. However, people current must provide library metadata by hand. Since libraries are often posted to Usenet or on Web sites in standard formats, it makes sense to collect these libraries and try to derive the semantic information from context.