Abstract: Disclosed are various embodiments for the identification and assignment of classifications for items. One of a plurality of classifications are identified in at least one computing device based at least in part on a plurality of tokens associated with the item and a data dictionary. An identified classification is thereby produced, and the identified classification is associated with a confidence score. The identified classification is automatically assigned to the item when the confidence score meets a threshold. A manual confirmation of the identified classification is obtained when the confidence score is below the threshold. The manual confirmation indicates whether the identified classification is erroneous.