Training method of reply message generating model and reply message generating method and device

A technology for replying information and generating models, which is applied in the field of recurrent neural networks, and can solve problems such as low accuracy of replying information

Active Publication Date: 2018-06-12
ZHONGKE DINGFU BEIJING TECH DEV
View PDF4 Cites 18 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

However, the existing intelligent question answering system only relies on the input information of the current round to generate reply information, so the accuracy of the reply information fed back to the user is low

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
  • Training method of reply message generating model and reply message generating method and device
  • Training method of reply message generating model and reply message generating method and device
  • Training method of reply message generating model and reply message generating method and device

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0064] The embodiments of the present application are described in detail below.

[0065] Previously, in order to solve the problem of "low accuracy of the reply information generated only by relying on the input information of the current round", a reply information generation method based on the time sliding window was developed, that is, a time sliding window was set to extract the current session past Conversation records within a certain time range before, extract features from the conversation records and add them to the current round of conversation information, and then generate the current round of reply information according to the completed conversation information in the current round. However, there are several problems in this method: first, the length of the time sliding window is difficult to determine; second, there are great difficulties in extracting features from session records and completing the session information of the current round. This is because co...

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

An embodiment of the invention discloses a training method of a reply message generating model and a reply message generating method. The training method includes: acquiring first input sequence; using a first decoding hidden state vector as the initial value of the hidden layer of a first encoder, and using the first encoder to encode the first input sequence to obtain a second encoding hidden state vector; using a decoder to decode the second encoding hidden state vector to obtain first output sequence; calculating the error between standard output sequence and the first output sequence; ifthe error is higher than a preset termination threshold, updating the parameters of the first encoder and the decoder according to the error. When the reply message generating model obtained by the training method is used to generate reply messages, the dialogue information of a previous round or several previous rounds can be introduced into the generation process of the reply messages of the next-round dialogue, and accordingly accurate reply messages can be generated, and the generated reply messages conform to the natural language interaction habits of users.

Description

technical field [0001] The invention relates to the technical field of cyclic neural networks, in particular to a training method and training device for a reply information generation model. In addition, the present application also relates to a reply information generating method and generating device. Background technique [0002] Intelligent question answering is a technology that provides information services to users in the form of one question and one answer through human-computer interaction. According to the feedback mechanism of answer generation, intelligent question answering systems can be divided into retrieval-based question answering systems and generative-based question answering systems. The answer feedback mechanism based on generation mainly uses a large number of interactive data pairs (question-answer pairs) to train a reply information generation model, and then inputs the user input information into the trained reply information generation model, and...

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): G06F17/30G06F17/27
CPCG06F16/3329G06F16/36G06F40/284
Inventor 蒋宏飞王萌萌李健铨崔培君晋耀红杨凯程
Owner ZHONGKE DINGFU BEIJING TECH DEV
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