A data extraction system for extracting a unique identifier from a plurality of different types of tokens, the data extraction system comprising a central processing system arranged to receive a data string representing an image of the token, the central processing system comprises: a data store of predetermined data records relating to the plurality of different types of token, each predetermined data record including a plurality of discrete features derived from an image of each type of token; a token type recognition module for identifying a type of token which the representation relates to, the recognition module comprising: a discrete feature identifier for iteratively identifying discrete features present in the representation; and a matching engine for iteratively comparing each of the identified features with each of the discrete features of the plurality of different types of token stored in the data store and registering each matched feature; wherein the recognition module is arranged to compare a current number of registered matched features of the representation with a predetermined number of minimum matched features and when the current number of registered matched features is at least equal to the at least the predetermined number of minimum matched features to determine which type of token the representation relates to; a data extractor for extracting at least one unique identifier of the token from the representation, wherein the data extractor is arranged to use the type of token identified by the token type recognition module to locate a region of the representation where the unique identifier is provided and to focus a data extraction process at that region.