A method for processing a user's spoken communication that is received by the user's electronic communications device and converted to text thereby, to assess relationships between words and phrases in the spoken communication, where necessary, to generate a directive to the user on how he/she might modify future spoken communications to more clearly convey a semantic content intended, for purposes of self-help and self-training in the user's spoken communications. The method also generates unsolicited text based on preprogrammed instructions. The user can configure the method in such a way that the user, not the software application program operating in the electronic communications device, contribute to the final directive. A version modifying original spoken text with complete sentences, and another, called user-mediated, that only identifies keywords or phrases and makes suggestions from word phrase pools to the user with modified keywords and phrases, to produce a directive are described.