Animation Scene Generation Based On Situational Semantics

People generally tend to prefer using animation to show the content,such as the popular AR(augmented reality)technology,for the same reason,the automatically generate animation from the natural language has also appeal to a lot of researchers.Based on the project "The research and demonstration of experiencing Shaoshan tourism culture and digital performance and display",which is a sub-project of"Research and development of cultural tourism resources mining and experiential platform" as the research background,this article has studied and designed the animation scene generation system based on situation semantics,and achieved that automatically generating a 3D scene which conforms to that text by analyzing and processing a period of natural language.The main work in this paper includes the use of dependencies to extract the spatial relationship in the text,the scene generation method based on prior template,etc.,more details are as follows:First of all,according to the extraction of scenario semantic information,this paper puts forward the method of using Stanford dependency analysis to complete the extraction of spatial relations in the sentence,and writes a set of relatively complete extraction rules,and compares the recalling rate of spatial relations extracting of the methods in this paper and other through the experiment.Experiments show that the rules of extracting spatial relations in this paper are significant for general sentence extraction.Secondly,for the layout of scene,this paper presents a algorithm which is based on prior template.First,about the corpus,the template is generated based on the simple conditional probability statistics.Then,the scene type is determined by the probability of occurrence of the object in the text,and the template is matched and applied according to the scene type to generate a relatively perfect scene.This method is to solve the oral story often omit the scene description information and can not be placed on the issue of objects.Finally,a scene generation system was designed according to the methods proposed above.With voice as input,converted it into text,extracted the scene semantic information and the scene display information,then using the connected graph to traversal the target objects,calculating the coordinates of each object,and generating the scene description file,finally through the Unity3D parsing scene description file to generate the final animation scene.Through the experiment with corpus,we proved the validity and practicability of the animation scene generation method proposed in this paper.