Data model 1202 may be specified by input or may be learned via a machine-learning technique. For example, the technique may identify general structures of composite identifiers (e.g., one or more characters having a character length within a range). Upon detecting such a structure with character identities not matching any defined composite, the technique may generate a new composite or may monitor to determine whether a threshold number of occurrences of the character(s) are detected. In some instances, a clustering or neural-network technique can be used to determine whether two different composite representations are to be merged as representing the same type of data.
Protocol generation system 1200 further includes a calendar 1208. Calendar 1208 may relate each of one, more or all dates to a day of the week, whether the date is a holiday, whether the date is a date associated with reporting significance (e.g., an end of a fiscal year), and soon. Calendar 1208 may correspond to stored data, may be determined based upon an algorithm or may be based on received communications including calendar data.