Abstract: Automatic categorization of a financial transaction based upon a mapping of useful characters from the transaction's description to a financial category. The description of the financial transaction is parsed to identify one or more useful strings of characters. A data file of business names is then searched for a match with the parsed string or strings from the transaction description. The data file is optimized to minimize both lookup times and the size of the data file by representing business-name-to-financial-category mappings using a serialized trie accessed via a memory mapped file. Nodes of the trie having children but no siblings are compressed into dangling nodes. A table of shared suffixes is also used. If a match is found in the business name data file, then the transaction is categorized according to the corresponding business name to financial category mapping. Otherwise, the transaction may be categorized based upon a match found in a database of financial transaction description keywords.