St. Bernard: the File Retrieving Softbot

There is a need to empirically validate theoretical Machine Learning research. This can be di cult because the algorithms being tested often require large amounts of training data. Softbots provide a unique solution to this problem. Softbots or software robots are intelligent agents that interact with software environments. By attaching the agent directly to the training environment, it is possible for softbots to gather their own training data. St. Bernard is a softbot that locates les within the UNIX 1 le system. It is based on the theory of satis cing search. The training data used by satis cing search is gathered directly from its environment. St. Bernard demonstrates the usefulness of satis cing search and addresses some of the practical issues of implementing this