Extraction of Visual Material and Spatial Information from Text Description for Scene Visualization

摘要

The generation of 3D virtual scene could be greatly simplified by integration of natural language interface. This involves complex task of extracting of visual object, material and their spatial information from language description. The knowledge of visual perception, spatial language, natural language processing and graphic representation can be combined with one another to accomplish the task of generating a virtual scene. This paper first mentions a variety of research in developing language based scene application. This is followed by a brief introduction of natural language processing technology. Some related theories of human visual perception and spatial language are investigated in section 3. A representation formalism based on word-concept-visual information has been proposed by linking the levels of meaning and visualization in section 4. Finally, we draw conclusion and mentioned future work.