Natural language problem generation method in civil construction information field based on Transformer

A natural language problem, civil engineering technology, applied in computer artificial neural network natural language processing, civil engineering information field, can solve the problem of reducing the workload of manual labeling, natural language problem generation is difficult to automate, etc., to achieve the effect of improving generation ability

Active Publication Date: 2021-03-26
XIAN UNIV OF TECH
View PDF9 Cites 8 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0003] The purpose of the present invention is to provide a Transformer-based method for generating natural language questions in the field of civil engineering and construction information, which reduces the workload of manual labeling and solves the problem that the generation of natural language questions in the field of civil engineering and construction information in the prior art is difficult to automate

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
  • Natural language problem generation method in civil construction information field based on Transformer
  • Natural language problem generation method in civil construction information field based on Transformer
  • Natural language problem generation method in civil construction information field based on Transformer

Examples

Experimental program
Comparison scheme
Effect test

Embodiment

[0051] The present invention is based on Transformer's method for generating natural language questions in the field of civil engineering and construction information, comprising the following steps:

[0052] Step 1: Pre-training on Wikipedia open-domain text. A Transformer-based 12-layer stacking module is constructed to manually preprocess the public Chinese Wikipedia text corpus to form a unified structure in the form of upper and lower sentences, and then input the processed Wikipedia corpus into the model for model pre-training. The pre-training stage mainly uses the same two-way cover pre-training mechanism and the second half sentence prediction training mechanism as Bert. The pre-training of a large amount of open-domain text makes the model have a certain common sense. The model in this paper conducts word frequency statistics on the Chinese Wiki Encyclopedia data, and the scale of the constructed dictionary is 32162. For tokens outside the dictionary during trainin...

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 Transformer-based natural language problem generation method in the civil construction information field, which comprises the following steps of: analyzing information distribution learned by each layer in a Bert model, training different corpora of different modules of a Transformer in combination with finite characteristics of training data, and proposing a low-layer network for training syntax grammatical characteristics, the high-level network is used for training a mechanism for obtaining semantic characteristics. Then, a UniLM thought is adopted to finely adjusta downstream task of the Bert, so that the natural language text generation capability of the model in the field of civil construction information is improved; the model provided by the invention hasrelatively high feasibility and effectiveness in problem generation in the field of civil construction information, and reaches a relatively high natural language problem generation level.

Description

technical field [0001] The invention belongs to the technical field of computer artificial neural network natural language processing, and in particular relates to a method for generating natural language questions in the field of civil engineering and construction information based on a Transformer encoding and decoding structure. Background technique [0002] With the continuous development of artificial intelligence and big data technology, intelligent question answering systems are rapidly changing people's lifestyles, and greatly improving work efficiency and user experience. Question Generation QG (Question Generation) is the basic task of a question answering system. Its goal is to automatically generate natural language questions given a sentence or paragraph. This task has been widely used in many fields, such as education, medical treatment, Internet of Things, etc. In the field of education, learning and teaching tasks are assisted by reading comprehension-style ...

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): G06F16/332G06F16/35G06N3/08
CPCG06F16/3329G06F16/353G06N3/08
Inventor 朱磊焦瑞黑新宏赵钦杨明松姚燕妮彭伟董林靖
Owner XIAN 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