Abstract
A novel genetic approach to optimisation has been devised and applied to information retrieval. In BTGP, each query can be expressed as a Boolean decision tree. This method, based on genetic programming techniques, first generates a population of query trees. These are evaluated for a set of training cases previously chosen and altered at each generation by the genetic operators of mutation and crossover. The precision and recall rates at extracting the relevant documents and rejecting the non-relevant ones is the measure of performance. At each generation the best solutions are those that when evaluated for a different set of test documents for which the system has not been trained, give the best fitness results. Degrees of precision and recall of 90% and above are obtained.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have