The Extraction of Trajectories from Real Texts Based on Linear Classification

Text-to-scene conversion systems need to share the spatial descriptions between natural language and the 3D scene. Such applications are the ideal candidates for the extraction of spatial relations from free texts, in which the extraction to trajectories that are focus objects in spatial descriptions is an essential problem. We present an analysis of how the space relations are described in Chinese. Based on this study, we propose a method where the extraction of trajectories is modeled as a binary classification problem and resolved based on a linear classifier with syntactic features. Moreover, experimental results are analyzed in detail to demonstrate the effectiveness of the linear classifier to the extraction problem of the trajectory concept.