US 11,782,964 B2
Method to recommend intents based on a weighted ranked hierarchical graph
Roberto Coutinho, Porto Alegre (BR); Roberto Silveira, Porto Alegre (BR); Guilherme Gomes, Porto Alegre (BR); and Carlos Nascimento, Porto Alegre (BR)
Assigned to ADP, INC., Roseland, NJ (US)
Filed by ADP, Inc., Roseland, NJ (US)
Filed on Oct. 8, 2021, as Appl. No. 17/450,380.
Prior Publication US 2023/0112843 A1, Apr. 13, 2023
Int. Cl. G06F 16/33 (2019.01); G06F 40/279 (2020.01); G06F 16/34 (2019.01); G06F 16/338 (2019.01)
CPC G06F 16/3344 (2019.01) [G06F 16/338 (2019.01); G06F 16/34 (2019.01); G06F 40/279 (2020.01)] 20 Claims
OG exemplary drawing
 
1. A method comprising:
executing, via a processor of a host computer system communicating across a network with a user computer system, a chatbot;
receiving, at the processor of the host computer system via the chatbot, an initial intent as a request by a user of the chatbot;
receiving, at the host computer system from a database, a plurality of hierarchical intents, each intent in the plurality of hierarchical intents comprising at least one text-based word;
mapping, via the processor of the host computer system, the plurality of hierarchical intents to a set of specific canonical actions, each canonical action in the set of specific canonical actions identifying an executable command which the host computer system is configured to execute, resulting in a map of intents to canonical actions, wherein a first canonical action of a first set of specific canonical actions is mapped to both a first hierarchical intent of the plurality of hierarchical intents and a second hierarchical intent of the plurality of hierarchical intents;
transforming, via the processor, the map of intents to canonical actions into an electronic graph representation;
receiving, at the processor via the chatbot, a ranked count of executed canonical actions, the ranked count of executed canonical actions identifying a number of times each canonical action within the electronic graph representation has been executed by the processor;
identifying, via the processor, a plurality of possible intents based on the initial intent, the plurality of possible intents represented within the electronic graph representation as connected to a common higher node;
removing, via the processor, at least one intent from the plurality of possible intents based on a level of authorization associated with the user, resulting in modified possible intents; and
selecting, via the processor, a recommended canonical action from within the set of specific canonical actions, the recommended canonical action being based on the ranked count of executed canonical actions and the modified possible intents.