The Entity Attribute Structure In Text-to-Scene Conversion

Human beings regard vision as a window to understand the objective world.They can accumulate large amounts of knowledge and form experience though vision,which can help human beings understand the objective world.With the development of visual information technology,people pay more attention to information visualization,and the visualization of Chinese information has gradually been developed.The text-to-scene conversion is a process that is dependent on computer to realize the creation of 3D scenes from natural language description automatically.It has a broad application prospect in the fields of design,animation,education,military,artificial intelligence and so on.Meanwhile,it is of great scientific and practical significance to improve the intelligently understanding of the computer and free our hands and brain.But the existing research is not mature in the basic theory and technical scheme.The entity attribute structure is one of the main contents of entity visualization in text-to-scene conversion.The main research contains entities' classification,visualization attribute structure,the reasoning and extraction of information,etc.Currently,it is still lack of systematic basic theory and effective solutions about these problems.Facing all this,the factors that influence entities' visualization effect are studied,such as entity shape,surface texture,constraint conditions and constraint relationships of all kinds of entities' visualization factors,and visual entities are classified detail,general model of entity visualization attribute structure is proposed.Then,considering the imperfections of visual factors,it canbe deduced by the improved similarity reasoning method that is proposed in the thesis based on the existing visualization factor datas.In addition,in consideration of visual information extraction from the text description,attribute words extraction method based on semantic classification is proposed.The experimental results show that the visual attribute structure model proposed in this thesis can play a key role in the transformation process of entity noun to entity parameterization.In addition,using the improved similarity reasoning method,there is a reduction in order of magnitude on the Mean Absolute Error(MAE)and the Mean Square Error(MSE)of height reasoning experiments and weight reasoning experiments.The attribute words extraction experiments based on semantic classification makes obvious effect in words extraction.The results prove the validity and feasibility of the basic theory and technology route proposed in this thesis.