US 12,014,141 B2
Systems and methods for improved transaction categorization using natural language processing
Tanmay Shah, Mountain View, CA (US); Pavanakumar Manakatti, Mountain View, CA (US); Suchit Kumar Biswal, Mountain View, CA (US); and Neha Kumari, Mountain View, CA (US)
Assigned to INTUIT INC., Mountain View, CA (US)
Filed by Intuit Inc., Mountain View, CA (US)
Filed on Feb. 3, 2021, as Appl. No. 17/167,052.
Prior Publication US 2022/0245342 A1, Aug. 4, 2022
Int. Cl. G06F 40/284 (2020.01); G06F 16/35 (2019.01); G06Q 40/12 (2023.01)
CPC G06F 40/284 (2020.01) [G06F 16/35 (2019.01); G06Q 40/12 (2013.12)] 18 Claims
OG exemplary drawing
 
1. A computer implemented method for categorizing a transaction description, said method comprising:
receiving a description associated with a transaction;
extracting a plurality of candidate keywords from the description, wherein extracting the plurality of candidate keywords from the description comprises:
tokenizing the description into a plurality of tokens;
determining a frequency count of each token; and
if a token is alphanumeric and an associated frequency count is above a count threshold, extracting the token as a candidate keyword;
generating a plurality of n-grams from the plurality of candidate keywords, each n-gram comprising a set of one or more candidate keywords;
calculating a score for each n-gram based on the one or more associated candidate keywords;
determining the n-gram with a highest score; and
generating a modified description with the determined n-gram.