Research And Implementation Of Intelligent Language Evaluation Model Based On Deep Learning

In recent years,a large number of intelligent electronic devices have poured into our study,life and work.The emergence of smart products has broken the traditional way of human behavior,and various smart products around people emerge in an endless stream.The same is true in the field of language education.Robot dialogue and one-on-one online language guidance have largely changed the way humans behave in language learning and language assessment.As the level of science and technology continues to grow,intelligent products will replace traditional tools.In order to increase the proportion of intelligent products in human language learning,online language evaluation models continue to increase,and people’s demand for online language learning and evaluation is also increasing.Therefore,the development of intelligent language learning and evaluation models is an urgent need of.The traditional online language assessment model only gives the number of language assessment scores,and basically does not have any opinions on language learning.Currently,online language evaluation systems based on gauge tables have been developed,but such systems can only achieve unilateral and objective comments on the evaluation aspects given in the gauge tables.One-to-one manual evaluation is comprehensive and detailed,but it is expensive and only a few people use this method.Therefore,how the language evaluation system realizes the intelligent output of multi-faceted and targeted comments and suggestions is a difficult point in the current language evaluation model research.Aiming at the shortcomings of current language evaluation models,this paper designs an intelligent language evaluation model based on deep learning.The intelligent language evaluation model can provide corresponding constructive comments and subjective suggestions based on the language audio of different testers,providing language learners with a personalized language education and teaching platform.The model adopts the latest sequence-to-sequence model framework.The main architecture is divided into an audio feature encoder based on a Bi-directional Gated Recurrent Unit and a natural language text generation decoder based on a Long Short-Term Memory.In order to improve the accuracy and subjectivity of the model output,the attention mechanism,semantic embedding method and generative adversarial networks are used in the model.The developed model can be called by most language education systems.This article mainly introduces the application of the model to the children’s language education platform.Therefore,this article uses the children’s spoken language evaluation data set.By evaluating the accuracy of the comments generated by the model,the response time of the comments generated by the model,and the scalability of the model,the model can realize normal operation on the language education platform.The development of this intelligent language model will solve the difficulties of single comments in the field of language education,low accuracy,non-objective comments,and inability to automatically generate comments.