A method for analyzing content of multimedia files is presented. The method comprises processing messages from mail servers, determining a type of the message, executing a convert module and obtaining text content of the message when said type is audio, extracting text from the text content or the message, creating a text version of components from the text, dividing the text version into different files, and analyzing the different files. Further, the messages can be captured, sent to a pending folder, retrieved from the pending folder. Creating the text version and dividing the text version can be performed by a mail engine. Natural language processing can be performed on the files. The text can be examined for compliance and, if appropriate, marked with quarantine violations.