Since we do not know how many speech units are there in the sentence to be recognized, the result of the recognition can be a sentence with less or more number of speech units than the correct number. Therefore the recognized sentence is matched with the correct sentence using dynamic programming techniques []. In this case three types of errors can occur; substitutions, insertions and deletions. After that we can calculate two performance indicators unlike only one indicator in the isolated case. One is called percentage correct and it is the percentage of correctly recognized speech units. The other one which is called accuracy gives the difference between percentage correct and percentage of deletions plus insertions.