AFFECT-DRIVEN DIALOG GENERATION

Number: 20200202887
Published: 2020-06-25
Filed: 2018-12-19
Assignee: Disney Enterprises, Inc.
Inventors: Modi; Ashutosh, Kapadia; Mubbasir, Fidaleo; Douglas A., Kennedy; James R., Witon; Wojciech, Colombo; Pierre
CPC: G06F40/35; G06N3/044; G06N3/04; G10L15/063; G06N3/0442; G10L15/28; G06N3/0455; G06N3/09; G06N3/088; G10L25/63; G06N3/0464; G10L15/22; G06N3/045
Verdict: Low Notable software
Source: Google Patents · FreePatentsOnline

The keeper's note

Affect-driven ML dialog generation for interactive characters.

Abstract

According to one implementation, an affect-driven dialog generation system includes a computing platform having a hardware processor and a system memory storing a software code including a sequence-to-sequence (seq2seq) architecture trained using a loss function having an affective regularizer term based on a difference in emotional content between a target dialog response and a dialog sequence determined by the seq2seq architecture during training. The hardware processor executes the software code to receive an input dialog sequence, and to use the seq2seq architecture to generate emotionally diverse dialog responses based on the input dialog sequence and a predetermined target emotion. The hardware processor further executes the software code to determine, using the seq2seq architecture, a final dialog sequence responsive to the input dialog sequence based on an emotional relevance of each of the emotionally diverse dialog responses, and to provide the final dialog sequence as an output.

Background

BACKGROUND

Advances in deep learning techniques have had a significant impact on end-to-end conversational systems. Most of the present research in this area focuses primarily on the functional aspects of conversational systems, such as keyword extraction, natural language understanding, and the logical pertinence of generated responses. As a result, conventional conversational systems are typically designed and trained to generate grammatically correct, logically coherent responses relative to input prompts. Although proper grammar and logical coherency are necessary qualities for dialog generated by a conversational system, most existing systems fail to express social intelligence. Consequently, there is a need in the art for affect-driven dialog generation solutions for producing responses expressing emotion in a controlled manner, without sacrificing grammatical correctness or logical coherence.SUMMARY

There are provided systems and methods for performing affect-driven dialog generation, substantially as shown in and/or described in connection with at least one of the figures, and as set forth more completely in the claims.

Claims

1. An affect-driven dialog generation system comprising: a computing platform including a hardware processor and a system memory; a software code stored in the system memory, the software code including a sequence-to-sequence (seq2seq) architecture trained using a loss function having an affective regularizer term based on a difference in an emotional content between a target dialog response and a dialog sequence determined by the seq2seq architecture during training; the hardware processor configured to execute the software code to: receive an input dialog sequence; generate, using the seq2seq architecture, a plurality of emotionally diverse dialog responses to the input dialog sequence based on the input dialog sequence and a predetermined target emotion; determine, using the seq2seq architecture, a final dialog sequence responsive to the input dialog sequence based on an emotional relevance of each of the plurality of emotionally diverse dialog responses to the input dialog sequence; and provide the final dialog sequence as an output for responding to the input dialog sequence. 8. A method for use by an affect-driven dialog generation system including a computing platform having a hardware processor and a system memory storing a software code including a sequence-to-sequence (seq2seq) architecture trained using a loss function having an affective regularizer term based on a difference in an emotional content between a target dialog response and a dialog sequence determined by the seq2seq architecture during training, the method comprising: receiving, using the hardware processor, an input dialog sequence; generating, using the hardware processor and the seq2seq architecture, a plurality of emotionally diverse dialog responses to the input dialog sequence based on the input dialog sequence and a predetermined target emotion; determining, using the hardware processor and the seq2seq architecture, a final dialog sequence responsive to the input dialog sequence based on an emotional relevance of each of the plurality of emotionally diverse dialog responses to the input dialog sequence; and providing, using the hardware processor, the final dialog sequence as an output for responding to the input dialog sequence. 15. A computer-readable non-transitory medium having stored thereon a software code including instructions, which when executed by a hardware processor, instantiate a method comprising: receiving an input dialog sequence; generating a plurality of emotionally diverse dialog responses to the input dialog sequence based on the input dialog sequence and a predetermined target emotion using a sequence-to-sequence (seq2seq) architecture trained using a loss function having an affective regularizer term based on a difference in an emotional content between a target dialog response and a dialog sequence determined by the seq2seq architecture during training; determining, using the seq2seq architecture, a final dialog sequence responsive to the input dialog sequence based on an emotional relevance of each of the plurality of emotionally diverse dialog responses to the input dialog sequence; and providing the final dialog sequence as an output for responding to the input dialog sequence.