US 12,456,019 B2
Response generation from multi-intent input
Krithika Viswanathan, Chennai (IN); Amit Mishra, Chennai (IN); Pravin Kumar Sankari Bhagavathiappan, Kanyakumari District (IN); and Niharika Marisetti, West Godavari (IN)
Assigned to Bank of America Corporation, Charlotte, NC (US)
Filed by Bank of America Corporation, Charlotte, NC (US)
Filed on Jun. 9, 2023, as Appl. No. 18/332,005.
Prior Publication US 2024/0412005 A1, Dec. 12, 2024
Int. Cl. G06F 40/205 (2020.01); G06F 40/157 (2020.01); G06F 40/284 (2020.01); G06F 40/40 (2020.01); G10L 15/22 (2006.01)
CPC G06F 40/40 (2020.01) [G06F 40/157 (2020.01); G06F 40/205 (2020.01); G06F 40/284 (2020.01)] 20 Claims
OG exemplary drawing
 
1. A system for generating a response from a multi-topic input, comprising:
a memory configured to store a table of conversational software algorithms and a plurality of topics, wherein the table comprises:
a first entry that indicates that a first conversational software algorithm is associated with a first topic; and
a second entry that indicates that a second conversational software algorithm is associated with a second topic; and
a processor, operably coupled to the memory, and configured to:
receive an input from a user, wherein the input indicates a request to perform one or more tasks;
extract a first set of features from the input, wherein the first set of features represents at least a sentiment of the input or a position of each word within the input;
generate a first vector that represents the extracted first set of features in a vector space;
access a training dataset that comprises a set of vectors, wherein each vector from among the set of vectors is associated with a given historical conversation;
compare the first vector with a second vector from the training dataset by determine a distance between the first vector and the second vector in the vector space;
determine that the distance is less than a threshold distance;
in response to determining that the distance is less than the threshold distance:
split the input into a plurality of utterances according to historical utterances associated with the second vector, wherein:
the plurality of utterances comprises a first utterance and a second utterance;
the first utterance indicates a first operation;
the second utterance indicates a second operation;
the first operation and the second operation are to be performed in order to perform the one or more tasks; and
determine a sequence of the first operation and the second operation to be performed in order to perform the one or more tasks;
identify, based at least in part upon the table, the first conversational software algorithm that is associated with the first utterance, wherein the first conversational software algorithm is configured to process the first utterance;
identify, based at least in part upon the table, the second conversational software algorithm that is associated with the second utterance, wherein the second conversational software algorithm is configured to process the second utterance;
communicate the first utterance to the first conversational software algorithm;
receive a first response from the first conversational software algorithm in response to communicating the first utterance to the first conversational software algorithm;
communicate the second utterance to the second conversational software algorithm;
receive a second response from the second conversational software algorithm in response to communicating the second utterance to the second conversational software algorithm;
determine whether each of the first response and the second response is valid; and
in response to determining that each of the first response and the second response is valid:
generate a final response based at least in part upon the first response and the second response; and
output the final response to the user.