Intelligent call-out dialogue method and system based on double dynamic memory networks

A dynamic and network technology, applied in the field of intelligent outbound dialogue methods and systems, to achieve the effect of precise answering

Pending Publication Date: 2022-02-18
SUNYARD SYST ENG CO LTD
View PDF0 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

However, end-to-end dialogue systems are usually difficult to interact with external knowledge bases, so an end-to-end dialogue system that can combine external knowledge bases is proposed

Method used

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
View more

Image

Smart Image Click on the blue labels to locate them in the text.
Viewing Examples
Smart Image
  • Intelligent call-out dialogue method and system based on double dynamic memory networks
  • Intelligent call-out dialogue method and system based on double dynamic memory networks

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0029] combined with figure 1 , the technical solution of the present invention is a kind of intelligent outbound dialogue method based on dual dynamic memory network, dual dynamic memory network, hereinafter referred to as DDMN, comprises the following steps:

[0030] Obtain dialogue data, which includes multi-round dialogue data and current round customer speech;

[0031] The multi-round dialogue data is processed through the BiGRU encoder to obtain the encoding vector H. The encoding vector H is used for the global management and iterative update of the dialogue history memory and the dialogue state memory in the dialogue memory manager. The dialogue state memory is addressed according to the encoding vector H. The operation finds the normalized weight a specified to be allocated to the memory slot t , to normalize the weight a t Read the context representation vector c in the dialogue history memory for guidance t , each round represents the vector c according to the co...

Embodiment 2

[0041] combined with figure 2 , the technical solution of the present invention is a kind of intelligent outbound dialogue system based on dual dynamic memory network, comprising:

[0042] Dialogue encoder, which processes multi-round dialogue data to obtain encoding vector information;

[0043] Decoder, after performing word embedding vectorization on the current round of customer speech, to obtain decoding vector information;

[0044] Dialogue memory manager, including dialogue history memory and dialogue state memory. The dialogue history memory stores the representation of the dialogue context during the entire dialogue process. The dialogue state memory is updated every round to track the flow of historical information and capture the encoding vector information of the current round. and decoding vector information to generate the first output distribution;

[0045] KB memory manager, including KB memory and KB memory pointer, KB memory uses an end-to-end memory networ...

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to view more

PUM

No PUM Login to view more

Abstract

The invention discloses an intelligent call-out dialogue method and system based on a dual-dynamic memory network, and the method comprises the steps: obtaining dialogue data which comprises multi-round dialogue data and a current-round customer verbal skill; processing the multi-turn dialogue data to obtain coding vector information which is used for global management and iterative updating of a dialogue history memory and a dialogue state memory in a dialogue memory manager; after word embedding vectorization is carried out on the current turn of customer verbal skill, obtaining decoding vector information and using the information for processing a conversation state memory of the memory manager and accessing a KB memory of the KB memory manager; the dialogue memory manager dynamically and sequentially expanding the dialogue memory, tracking the dialogue history through an updating mechanism, and generating first output distribution; dynamically accessing the KB memory manager to generate second output distribution; and generating a reply customer verbal skill according to the first and second output distributions. According to the invention, by learning man-machine historical dialogues and combining with the outbound knowledge base, smooth and accurate answering between the robot and the customer is realized.

Description

technical field [0001] The invention relates to the technical field of automatic dialogue generation, in particular to an intelligent outbound dialogue method and system based on a dual dynamic memory network. Background technique [0002] The traditional manual customer service personnel have high mobility, high labor cost, lack of standard service, low service efficiency and other problems, resulting in difficult recruitment, long training time, and increased related costs of the enterprise. Therefore, there is a demand for and development of intelligent customer service. Task-based dialogue robots play a key role in intelligent customer service. Task-oriented dialogue system (task-orienteddialogue TOD) is mainly divided into two categories: pipeline (pipeline) task-oriented dialogue system and end-to-end (end-to-end) task-oriented dialogue system, intelligent outbound dialogue system belongs to task-oriented Dialogue system, the traditional intelligent outbound system is...

Claims

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to view more

Application Information

Patent Timeline
no application Login to view more
Patent Type & Authority Applications(China)
IPC IPC(8): G06F9/50G06F16/332
CPCG06F9/5016G06F16/3329
Inventor 周鸿章宣明辉操玉琴刘卫东李宇豪
Owner SUNYARD SYST ENG CO LTD
Who we serve
  • R&D Engineer
  • R&D Manager
  • IP Professional
Why Eureka
  • Industry Leading Data Capabilities
  • Powerful AI technology
  • Patent DNA Extraction
Social media
Try Eureka
PatSnap group products

Browse by: Latest US Patents, China's latest patents, Technical Efficacy Thesaurus, Application Domain, Technology Topic.

© 2024 PatSnap. All rights reserved.Legal|Privacy policy|Modern Slavery Act Transparency Statement|Sitemap