Font Size: a A A

Research On Text Generation Of Comprehensive Quality Assessment System Based On Sequence Generative Adversarial Nets

Posted on:2021-01-12Degree:MasterType:Thesis
Country:ChinaCandidate:L Y HuangFull Text:PDF
GTID:2428330602476860Subject:Software engineering
Abstract/Summary:PDF Full Text Request
Comprehensive quality assessment is a form of quality education in school,to facilitate the management of students' learning process,it is necessary to develop an online management system for quality assessment,which can also serve as an important reference for further improving the ability of implementing quality education.Since currently the quality assessment is subjective sometimes,which mainly reflects in that the control power of the head teacher is too much,and assessment process is often not transparent in some school,that leads to the kind of way is still questioned.This thesis uses the comprehensive quality assessment data of middle school students in Jiangxi province from previous years,through the text generation method based on big data and complete online management,to reduce the subjective influence of the participants to a certain extent,and improve the impartiality of the assessment.At the same time,it helps teachers save the time spend on writing comments and improve work efficiency.Aiming at the generation of quality education comment,this thesis makes an in-depth research and analysis on the automatic text generation.Firstly,the research achievements in the field of text generation are classified and summarized,and the next,the basic concepts,specific structures and training methods of several related neural networks are introduced.On this basis,according to the downstream task proposed in the thesis,the generation method using SeqGAN with abstract semantic control is realized,which add impacts of associated semantics in feature extracting and hidden vectors of discriminator and generator respectively,so that the system can generate the comment which conforms to the expected effect according to the input assessment level.In addition,automatic and manual evaluations were used to establish the evaluation system.Experimental result shows that the method could achieve the design target.Finally,this thesis implements the practical application for comment generation in the quality assessment management system,and verifies its effectiveness.
Keywords/Search Tags:Text Generation, Generative Adversarial Nets, Semantic Control, Policy Gradient
PDF Full Text Request
Related items