Modelling pronounciation dictionary of Turkish by CART trees

Although Turkish is usually read as written, there are still differences and mismatches with spoken and written Turkish. In order to build reliable speech processing applications, or to determine linguistics standards, defining pronounciation rules in Turkish is a promising task. Objectively, the main purpose of this study is said to determine the transition rules between orthographic and phonetic transcriptions by referencing the "Language and Pronounciation Dictionary of Turkish". On the other hand, FESTIVAL is used to settle the letter-to-sound rules which is accepted as a milestone in a typical speech synthesis system. Thus, the experimental results are reported according to the final letter-to-sound rules. All the training and test words, which are used to determine the letter-to-sound rules statistically, are all selected from the "Language and Pronounciation Dictionary of Turkish".