Semantic Tagging Using Genetic Algorithm

A genetic algorithm with adaptive evaluation function is presented to deal with data sparseness problem in automatic semantic tagging. Taking advantage of the hierarchy structure of Synonymy Thesaurus, semantic induction is used to improve the quality in estimating the parameters of the function in genetic algorithm. Based on the definitions of two fundamental concepts, the principle of semantic induction is described. Restrictive selection policy is applied to reverse the decline of model's discernment caused by the induction. Finally, the genetic algorithm is implemented and testing results show that the algorithm is feasible to different training data sizes.