Descriptive Sentence Extraction for Text to 3D Scene Generation

Three-dimensional objects (3D) allow extensive and heterogeneous information to be stored in single models which can be exploited by users to satisfy various research and study needs. Moreover, the 3D visualization would be even more interesting if it were the result of the "materialization" of descriptive sentences extrapolated from texts related to the subject matter. In other words, a direct connection between 3D models and the associated texts or drawings could provide a useful and stimulating explication of the case of study. The extrapolation of specific information from texts, however, is time-consuming and it requires the user to have in-depth knowledge of the referring domain. An innovative solution to the problem, then, would be to develop a system that can analyse and "comprehend" the documents in order to automatically provide, as output, portions of text containing geometric and spatial information useful for the 3D scenes generation. In this paper, the analysis of the framework of the above-mentioned system is presented and its implementation on a specific corpus concerning the "World City" project, is evaluated.