Unlock instant, AI-driven research and patent intelligence for your innovation.

Universal English man-machine conversation generation method and system based on balanced distribution

A technology of man-machine dialogue and English, applied in neural learning methods, semantic analysis, computer components, etc.

Active Publication Date: 2020-04-21
BEIHANG UNIV
View PDF4 Cites 3 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0005] The technical problem to be solved by the present invention is: to overcome the problem that the existing dialogue generators always generate common answers, and to provide a method for generating general English dialogues based on balanced distribution. quality

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
  • Universal English man-machine conversation generation method and system based on balanced distribution
  • Universal English man-machine conversation generation method and system based on balanced distribution
  • Universal English man-machine conversation generation method and system based on balanced distribution

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0090] The present invention will be further described below with reference to the drawings and examples.

[0091] The present invention is a general English human-machine dialogue generation method based on balanced distribution, which makes full use of the dialogue behavior categories of the user's response to the model generation, and applies these feedback information to the generation model by affecting the output distribution of the decoder.

[0092] The present invention first trains to obtain the feature-aware weight network or feature-aware recurrent network, and then performs the step of generating, mapping the above-mentioned vocabulary to the corresponding number, and using the global vector (GloVe) word vector to initialize the words contained in the context Next, the encoder encodes the context into a hidden state vector, and the decoder generates each step output according to the context vector. Generally, the words generated here are often words with a high probabili...

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 English man-machine conversation generation method and system based on balanced distribution. The method comprises the steps: firstly proposing a feature perception weight network, and enabling the network to take the generated response as the input, so as to learn the proper weight related to different conversation information (such as conversation behaviors or emotion); thereafter, the weights are assigned to different words in the response generated by the generator, which are used as offsets to modify the original loss function. Finally, the modified loss function is used to train a generator to generate responses related to various external information. The contribution of the invention mainly comprises two aspects: (1) a method is provided, which introducesexternal weights to almost all generators so as to better adapt to various external information; (2) providing a feature perception weight network, and generating different weights for each word according to the relationship between each word and conversation external information; compared with other basic models, the method has the advantage that the generation effect is obviously improved.

Description

Technical field [0001] The invention belongs to the technical field of dialog generation and natural language processing research, and improves the generation quality of the generator by adding a feature perception weight network to the generator. The information used in the method is the dialogue information including dialogue behavior and dialogue emotion. Through this information, the present invention can adjust the distribution of the original generated words of the generator, so that the generated response can meet the needs of the context. Background technique [0002] With the rise of computers, humans began to conceive of artificial intelligence, and the core of artificial intelligence should be the ability of robots to communicate with humans. Therefore, Turing defined the Turing test as an evaluation criterion for artificial intelligence. Specifically, The Turing test is to allow humans to judge whether the other party is artificial intelligence based on only text answ...

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): G06F16/332G06F40/30G06K9/62G06N3/04G06N3/08
CPCG06F16/3329G06N3/08G06N3/048G06N3/044G06N3/045G06F18/24
Inventor 荣文戈杨浩东欧阳元新熊璋
Owner BEIHANG UNIV