Embodiments of the present invention disclose a method, computer program product, and system for recognizing patterns in log files with unknown grammar. A computer replaces one or more alphanumeric strings with a first alphanumeric character to generate a first resulting string. The computer then replaces one or more identical pairs of characters of the first resulting string with a second alphanumeric character to generate a second resulting string. The computer then replaces one or more consecutive instances of the second alphanumeric character, in the second resulting string, with one instance of the second alphanumeric character to generate a compressed string.