Design And Application Of The Innovative Oral English Training System Based On The Metaverse Environment

English is a global language around the world,and now China attaches great importance to the quality of English teaching.As China is not a native English speaking country and relatively lacks oral training methods,Chinese students’ oral English proficiency is relatively poor.Under special circumstances such as COVID-19,the metaverse environment can guarantee the continuity and stability of education and promote the overall upgrading of language education.In the metaverse environment,this thesis combines virtual reality and artificial intelligence,and creates an immersive learning environment through the deep integration of virtual and reality,which can avoid the problems of single learning mode,boring pages,lack of interaction and difficult evaluation of online education and traditional education.Based on oral comprehension technology,an oral English training system is developed,which can be used for learners’ oral English practice and evaluation.The research contents of this thesis are as follows:First of all,according to the Y demand analysis model,from the perspectives of user needs,user objectives and product functions,combined with Maslow Theory,this thesis investigates user needs and precisely locates target users to complete the system framework and function design.The system deploys virtual reality technology to solve the problem of lack of immersive English learning environment,and artificial intelligence technology is used to support the training tasks including oral semantic understanding,word and sentence pronunciation,and training evaluation.Secondly,in task-based multi-round dialogues,the semantics of user sentences are closely related to historical information,coding historical information is a very critical step.Aiming at the weight distribution of historical sentences and the question of dialogue jumps,this thesis proposes an oral comprehension method based on role information.The high-dimensional semantic embedding word vectors are extracted through BERT preprocessing method,and multi-layer perceptron are used to calculate the weights of historical sentences involved in encoding.Since the role and dialogue habits of speakers have a greater impact on the sentence parsing,this thesis calculates the historical influence vector according to the role information,fully extracts the historical sentence information through the Bi-directional Long Short-Term Memory network,and finally uses softmax function to obtain the sentence intention and the probability distribution of slot value filling.Through verification,the intention recognition and slot filling accuracy of this method on DSTC4 data set is significantly improved.Finally,this thesis builds a dialogue system for oral English training.In the metaverse environment,the function of reading and writing words,reading sentences,multiple rounds of dialogue and dialogue quality evaluation can be realized.The semantic understanding algorithm is applied to the English dialogue system,and users can practice oral English by talking with virtual characters.The methods of speech recognition,virtual role call,speech evaluation and rule matching are adopted to realize multiple rounds of dialogue designed for the hotel environment.Virtual characters include real character actions and animation effects,which provide learners with a good learning experience.By analyzing feedback of users through user investigation,the results show that students can practice and evaluate oral English through dialogue and interaction with virtual characters in an immersive metaverse Environment,which is helpful to students’ oral English learning.