| CPC G06N 3/08 (2013.01) [G06F 40/20 (2020.01); G06N 3/047 (2023.01); G06N 5/041 (2013.01); G10L 15/063 (2013.01); G10L 15/1815 (2013.01)] | 15 Claims |
|
1. A method of using a neural network based open-domain dialogue model for open-domain dialogue response generation, the method comprising:
receiving an input utterance from a device having a conversation with a trained neural network based open-domain dialogue model;
obtaining, from the trained neural network based open-domain dialogue model, a response, wherein the trained neural network based open-domain dialogue model is trained based on quality scores of candidate replies using a highest quality score and a quality score of a randomly selected candidate reply among the candidate replies, and wherein obtaining the response comprises:
obtaining at least one candidate reply to the input utterance;
obtaining at least one quality score corresponding to the at least one candidate reply is based on a plurality of reference-free discriminators, and wherein the plurality of the reference-free discriminators evaluate at least one of:
a fluency of the at least one candidate reply by determining a perplexity of candidate replies based on contextual information,
a specificity of the at least one candidate reply by determining normalized inverse function document frequencies of the candidate replies, and
a consistency of the at least one candidate reply by calculating a probability that the at least one candidate reply contradict previous candidate replies output by the trained neural network based open-domain dialogue model during the conversation; and
determining the response based on the at least one quality score corresponding to the at least one candidate reply.
|