Mining the Web for knowledge with sub-optimal mining algorithms

The Web provides a forum in which AI systems can be demonstrated and compared. This paper addresses a fuzzy method for context-sensitive textual matching. We are investigating two key approaches. Knowledge on the Web must be retrieved and structured to facilitate mining operations. Case-based filtering allows the algorithm to adapt dynamically to changes in content or efficiency of expression. Our approach is to design sub-optimal mining algorithms that sacrifice completeness for speed, tractability and breadth of coverage. The mined knowledge is fed back to serve as a heuristic filter.