Deep neural network and reinforcement learning-based generative machine reading comprehension method

A deep neural network and reinforcement learning technology, applied in the field of natural language processing, can solve problems such as inability to use effective information fragments at the same time, achieve more flexibility in optimization goals, and simplify the training process

Active Publication Date: 2018-08-17
SOUTH CHINA UNIV OF TECH
View PDF4 Cites 88 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

The S-NET model proposed in the paper "ChuanqiT, et al. S-Net: From Answer Extraction to Answer Generation for Machine Reading Comprehension[J]. arXiv preprint arXiv: 1706.04815, 2017" adopts the method of first extraction and then synthesis, which can better can generate answers that conform to the form of natural language, but its extraction method has shortcomings, and it cannot use multiple effective information fragments in the text at the same time
The model proposed in the paper "Rajarshee Mitra.An Abstractiveapproach to Question Answering[J].arXiv preprint arXiv:1711.06238,2017" directly adopts the generation method, the model is more simplified, but it loses the extractive model mark to highlight the effective information in the original text The advantages

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
  • Deep neural network and reinforcement learning-based generative machine reading comprehension method
  • Deep neural network and reinforcement learning-based generative machine reading comprehension method
  • Deep neural network and reinforcement learning-based generative machine reading comprehension method

Examples

Experimental program
Comparison scheme
Effect test

Embodiment

[0027] This embodiment describes a generative machine reading comprehension model based on deep neural network and reinforcement learning, and its specific implementation includes the following steps:

[0028] Step R1, perform preprocessing such as word segmentation, part-of-speech tagging, and named entity recognition on the text and sentences in the question, and map words into corresponding word vectors in the vocabulary (usually using GloVe word vectors or combining them with CoVe word vectors). At the same time, for each word, according to its part-of-speech features and named entity category features, each feature is also mapped to a low-dimensional feature vector, which is spliced ​​together with the word vector. In addition, for each word in the text, according to its matching degree with the word in the question, two more features are added:

[0029] 1) Exact matching features, expressed as: β(p i )=II(p i ∈q), that is, when a word p in the text i When it appears i...

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 a deep neural network and reinforcement learning-based generative machine reading comprehension method. According to the method, texts and questions are encoded through an attention mechanism-combined deep neural network so as to form question information-fused text vector expressions, and decoding is carried out through a unidirectional LSTM decoder so as to gradually generate corresponding answer texts. According to the reading comprehension method, the advantages of extractive models and generative models are fused, training is carried out by adoption of a multi-taskcombined optimization manner, and a reinforcement learning method is used in the training process, so that benefit is brought to generate more correct and fluent answer texts.

Description

technical field [0001] The invention relates to the technical field of natural language processing, in particular to a generative machine reading comprehension method based on deep neural network and reinforcement learning. Background technique [0002] As a subfield of natural language processing, machine reading comprehension has received great attention in recent years and has also made great progress. The emergence of richer data and more diverse and advanced models enables machines to better process natural language text input and, to a certain extent, be able to answer relevant questions raised on the input text. This is important for building more advanced natural language processing applications, such as automatic question answering (QA), dialogue system (DialogueSystem), providing more intelligent, efficient, and personalized search engine services, and even building real strong artificial intelligence. basic meaning. [0003] At present, most of the mainstream ma...

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): G06F17/30G06F17/27G06N3/04
CPCG06F16/3329G06F40/289G06F40/30G06N3/045
Inventor 朱国轩王家兵
Owner SOUTH CHINA UNIV OF TECH
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