Technology Research On Generation Of Character Animation In Text To Scene Field

Character animation generation system which is responsible for visualization of animation information obtained from the analysis of the text in the form of animation is one of most important modules of Text-to-Scene system. The Natural Science Foundation Project"the three-dimensional visualization of spatial relationships in text based on ontology"is the study about the extraction and visualization of spatial and event relationships. Our research in this paper, which is an important part of the project, focus on the interpretation and visualization of the action relations of dynamic role in the event relationship and automatic realization in three-dimensional scene.In this paper, we first analyze the mapping relation from compound verb to original verb. We find that we can not interpret the compound verb to the sequence of original verbs with"HowNet". Therefore, this paper presents a model of script interpretation based on example library. The model main interpret the role action script whose core is the compound verb to the sequence of original script whose core is the original verb. The basic principle of the model is that we use the corpus which contain examples of the interpretation from role action script to sequence of original script to train model. That is means that the corpus will be classified according to a verb to create the script example library. For the test of role action script, we find the examples in the library which has highest similarity with the role action script that will be interpreted by the model. Then we extract the prototype of the example and apply it to the script which will be interpreted and get the interpretation.The character animation system based on the interpretation model consists of two parts: the model of script interpretation based on example library and the animation generation system. The animation generation system which based on VirtoolsSDK provide interface for each kind of original script. According to the kind of original script, system can simply call the appropriate interface to display the character animation.The animation generation system consists of three parts: scenes classification module, scene initialization module, role action processing module. The scenes classification module decided which scene to be loaded into the system. The scene initialization module gets the information of entity such as coordinates, size and direction and puts the entity into the scene. The role action processing module which based on the first two modules acquires the sequence of original script and calls the appropriate handling function, generate the character animation.