Abstract: A method for forming a file of messages for a computer program including the steps of first forming a first file portion as a message word table having entries corresponding to the message words, said entries ordered in the table by their frequency of occurrence in the messages. Then forming a second portion as a message token string table containing a token string for each message where each token string includes several tokens, each representing a word or variable in the message. The tokens are located in the string according to the location of the corresponding words in the message. Each token has a value that represents the location of its corresponding word in the message word table. The third step includes forming a third file portion as a message look-up table that contains a unique identifier for each message and a corresponding pointer that indicates the address of a token in the token string table.