US 11,941,033 B2
Method and system for syntactic searching
Qi Zhu, Toronto (CA); Sean Weisman, Toronto (CA); Gregory Ward, Toronto (CA); Alishan Ladhani, Toronto (CA); and Jeff Li, Toronto (CA)
Assigned to Affirm, Inc., San Francisco, CA (US)
Appl. No. 16/767,017
Filed by Affirm, Inc., San Francisco, CA (US)
PCT Filed Nov. 27, 2018, PCT No. PCT/CA2018/051501
§ 371(c)(1), (2) Date May 26, 2020,
PCT Pub. No. WO2019/100167, PCT Pub. Date May 31, 2019.
Claims priority of provisional application 62/590,748, filed on Nov. 27, 2017.
Claims priority of provisional application 62/590,740, filed on Nov. 27, 2017.
Prior Publication US 2020/0320113 A1, Oct. 8, 2020
Int. Cl. G06F 16/33 (2019.01); G06F 16/31 (2019.01); G06F 16/338 (2019.01); G06F 16/35 (2019.01); G06F 40/205 (2020.01); G06F 40/40 (2020.01)
CPC G06F 16/3344 (2019.01) [G06F 16/31 (2019.01); G06F 16/338 (2019.01); G06F 16/353 (2019.01); G06F 40/205 (2020.01); G06F 40/40 (2020.01)] 18 Claims
OG exemplary drawing
 
1. A method for syntactic searching for products available from merchants, the method comprising:
receiving search term comprising multiple individual words from user input at a user device;
parsing the individual words in the search term using natural language processing (NLP);
comparing the parsed individual words with words stored in a NLP database including a plurality of previously searched words from prior searches, and pre-populated words into the NLP database based on merchant product information, wherein the previously searched words and the pre-populated words each include relationship information associating selected ones of the previously searched words and the pre-populated words with each other and categorization information that categorizes each of the previously searched words and the pre-populated words;
causing display, at the user device, of search results based on comparison of words stored in the NLP database;
receiving, from the user device, a word change request from the user for a selected word among the individual words;
retrieving one or more alternate words associated with the selected word in a semantic tree;
displaying, at the user device, both the selected word and the one or more alternate words simultaneously and proximate to the selected word in a list of options for selection as a selected alternate word, the one or more alternate words including only words having a root word that is non-overlapping with respect to the selected word;
receiving an indication of the selected alternate word from among the one or more alternate words; and
updating the search results based on inclusion of the selected alternate word in the search term by employing machine learning to update the relationship information and the categorization information,
wherein parsing the individual words comprises categorizing the individual words as a selected one of group consisting of a root noun, a symbol, an adjective, a number, and a preposition, and
wherein retrieving the one or more alternate words comprises retrieving only words in a same category as the selected word.