Construction Of Three-Dimensional Scene Using Natural Language Understanding

If the course that generating 3D scene from language is finished automatically, a lot of manpower and material resources will be saved. In order to decrease the cost, many scholars have done a lot of research in that domain and many achievements have been worked out. The project that generate 3D scene from language is composed three steps: extraction spatial relation from user input, compute every model's coordinate in the scene and generate the 3D scene. In this thesis, the work that we did on researching in the domain will be shown.The thesis contains the content about building 3D scene under the help of NLP. Thesis builds wordhouse and modelhouse. The wordhouse id designed to our system, it is composed by many tables, and this framework suits our need. There are a lot of 3D models in modelhouse, each model file is added some information that is useful during computation.While extracting spatial relation from user's input, the course is departed into three steps, First, we participle user's input, the result is called jeresult; Secondly, we participle the jeresult under the help of wordhouse; Third, we extract spatial relation from jeresult. During the course of the thirdly step, we put forward some regulations and tagged every word in jeresult.At last, generate 3D scene,On the basis of the building of wordhouse and modelhouse,spatial relation, we generate 3D scene. In order to work out the problem that how to denote spatial relation in numeralization, each kind of location word has one compute model.After all of the disposement as front tells,the system in thesis let user build 3D scene that is the same as user's input without any difficult work.