Looking for breakthrough ideas for innovation challenges? Try Patsnap Eureka!

A chatting method for robots based on word vectors and recurrent neural networks

A technology of cyclic neural network and word vector, which is applied in the field of robot chat, can solve the problem of low accuracy of robot reply, and achieve the effect of improving reply accuracy, accuracy and performance

Active Publication Date: 2020-07-21
SOUTHWEST JIAOTONG UNIV
View PDF8 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0005] In view of the above-mentioned deficiencies in the prior art, a robot chat method based on word vectors and recurrent neural networks provided by the present invention solves the problem of low accuracy of traditional generative chat robot replies

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
  • A chatting method for robots based on word vectors and recurrent neural networks
  • A chatting method for robots based on word vectors and recurrent neural networks
  • A chatting method for robots based on word vectors and recurrent neural networks

Examples

Experimental program
Comparison scheme
Effect test

Embodiment approach

[0051] S203. Utilize the continuous bag-of-words model CBOW to carry out word vector training according to the word dictionary of the question-and-answer sentence, and obtain the word vector file QA.conv, and its implementation method is as follows:

[0052] S2031. Randomly generate a word vector matrix embedding according to the word dictionary, and initialize the word vector matrix [-1, 1]. The size of the word vector matrix embedding is (vocabulary_size, embedding_size), where vocabulary_size represents the word dictionary size , embedding_size represents the word vector dimension;

[0053] S2032. Extract the word vectors of the surrounding words of a certain central word according to the word vector matrix embedding, and solve the mean value vector of the word vectors of the surrounding words. The expression of the mean value vector avr of the word vectors of the surrounding words is as follows:

[0054]

[0055] Among them, n represents the number of unilateral words o...

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 provides a robot chat method based on word vectors and cyclic neural networks. The invention adopts the continuous bag-of-words model CBOW to train autonomous word vector files, calculates the similarity of sentences according to the word vectors, and retrieves the similarity with test questions The highest question sentence is compared with the preset sentence similarity threshold. If the threshold is lower than the threshold, the model that is iteratively trained by the multi-layer bidirectional network model LSTM and attention mechanism attention will be used for generative reply. If it is higher than the threshold, the maximum similarity will be output. Respond to the corresponding answer to the question. The invention solves the problem of low reply accuracy of traditional generative chat robots. The present invention can significantly improve the accuracy of the chat robot's reply, reduce the grammatical and semantic error rate of the generative chat robot's reply, and has higher interpretability compared with a single generative chat robot, and is of great importance to the research of the question answering system in the field of chatting meaning.

Description

technical field [0001] The invention belongs to the technical field of robot chatting, in particular to a robot chatting method based on a word vector and a recurrent neural network. Background technique [0002] With the rapid development of artificial intelligence technology, research on intelligent chat robot technology has attracted widespread attention. At present, the main applied technologies are divided into retrieval and generation. The retrieval chatbot analyzes the grammar and semantics of the sentence and searches the database for the most suitable answer sentence. Therefore, the retrieval method is more suitable for some service industries ( Such as: medicine, industry, bank), obviously, the focus of the retrieval method is on the quality of the corpus, and the quality of the corpus has become a major factor limiting its performance; Training, the most popular model at present is the seq2seq model. As long as the knowledge field involved in the corpus is wide e...

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 Patents(China)
IPC IPC(8): G06N3/00G06N3/04G06F16/33G06F16/332
CPCG06N3/008G06F16/3329G06F16/3344G06N3/045
Inventor 苟先太康立烨张葛祥胡梦陶明江
Owner SOUTHWEST JIAOTONG UNIV
Who we serve
  • R&D Engineer
  • R&D Manager
  • IP Professional
Why Patsnap Eureka
  • Industry Leading Data Capabilities
  • Powerful AI technology
  • Patent DNA Extraction
Social media
Patsnap Eureka Blog
Learn More
PatSnap group products

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

© 2024 PatSnap. All rights reserved.Legal|Privacy policy|Modern Slavery Act Transparency Statement|Sitemap|About US| Contact US: help@patsnap.com