The purpose of this thesis is to apply Inductive Logic Programming (ILP) to the segmentation of ambiguous Thai words. The ILP system which has been chosen is FOIL. Another learning system, which is used to be compared with FOIL, is a propositional learning system, RIPPER. First, FOIL and RIPPER are used to learn features of ambiguous Thai words. The outputs of FOIL and RIPPER are a set of first-order Horn clauses and a set of propositional rules respectively, each of which defines the features of the ambiguous Thai words. Experimental results show that the clauses learned by FOIL are more accurate than rules of RIPPER. Then, these clauses from FOIL are used to help for segmentation of ambiguous Thai words. Experimental results show that the usage of the clauses improves the accuracy of word segmentation which uses trigram model.