Under the first year of the meta universe,VR has become one of the most important technologies.The construction of virtual world is inseparable from virtual scene construction technology.In this context,the digitization of traditional opera is the only way to realize the protection and inheritance of traditional opera.However,the traditional opera virtual scene has some problems,such as complex construction process,single modeling method,low reuse rate.This paper combined with NLP and scene construction technology to carry out the research on the technological innovation of drama virtual scene construction based on semantics.The purpose is to improve the automation of drama virtual scene construction,reduce repetitive labor.The specific research work is as follows:Firstly,for the analysis and extraction of semantic information.This paper proposes a hybrid mode word segmentation method based on drama to analyze the morphology of the text.Syntactic analysis of the text from the perspective of sentence components.Finally,the information related to scene construction is extracted from the text through semantic analysis,and the spatial relationship embodied in different texts is expressed by the triad of location words,projectiles and boundary markers.Secondly,for opera scene construction.This paper divides the spatial relationship into relative position spatial relationship constraints and common sense spatial relationship constraints.For the spatial relationship of relative position,the spatial relationship triplet can be directly extracted from semantics to determine the placement position of the model.For the common sense spatial relationship,a placement method based on a priori probability of opera stage is proposed.Take the common spatial relationship of opera scenes as a priori knowledge.A priori probability is used to determine the common sense spatial relationship constraints in the scene.Complete the placement of all opera scene models.Finally,design an opera virtual scene generation system.Input Chinese text,extract the scene model information and the spatial relationship,and then use the a priori probability to complete the placement of the model,generate the scene description file and visualize it. |