Many applications in natural language processing require information about objects and their topological and geometric properties known as spatial relations. Arabic spatial role labelling addresses the challenge of extracting spatial relations between objects from Arabic texts. This paper extends spatial role labelling achieved on English texts to Arabic as a demanding computational linguistic track. Spatial roles including trajectors, landmarks, spatial indicators, motion indicators, and paths are associated with a solid annotation scheme in Arabic. Spatial role labelling method is implemented using conditional random fields for feature factorization and by training uni-class and multi-class probabilistic classifiers. Spatial relations are further classified into direction, distance or region using SVM classification approach. Experimental works are divided into two parts; static and dynamic relation roles. Our classifiers are trained using two supervised settings and evaluated using precision, recall, and F1-measure against gold annotated dataset. Overall results are encouraging in comparison with a straightforward baseline. Results from spatial roles identification and classification show the need to incorporate more linguistic features and lexicons to classify the relations correctly by SVM. The overall performance in dynamic spatial role labelling indicates the need for improved dataset and classifier training to achieve better precision and recall.

自然语言处理中的许多应用都需要有关对象及其拓扑和几何特性（称为空间关系）的信息。阿拉伯语空间角色标签解决了从阿拉伯语文本中提取对象之间空间关系的挑战。本文将在英语文本上实现的空间角色标记扩展为阿拉伯语，以此作为苛刻的计算语言轨道。包括轨迹，地标，空间指示符，运动指示符和路径的空间角色与阿拉伯语的固体注释方案相关联。使用条件随机字段进行特征分解并训练单类和多类概率分类器来实现空间角色标记方法。使用SVM分类方法将空间关系进一步分类为方向，距离或区域。实验工作分为两个部分：静态和动态关系角色。我们的分类器使用两个监督设置进行训练，并针对金标数据集使用精度，召回率和F1度量进行评估。与简单的基准相比，总体结果令人鼓舞。空间角色识别和分类的结果表明，需要合并更多的语言功能和词典，以通过SVM正确地对关系进行分类。动态空间角色标记的总体性能表明需要改进数据集和分类器训练，以实现更高的精度和召回率。与简单的基准相比，总体结果令人鼓舞。空间角色识别和分类的结果表明，需要合并更多的语言功能和词典，以通过SVM正确地对关系进行分类。动态空间角色标记的总体性能表明需要改进数据集和分类器训练，以实现更高的精度和召回率。与简单的基准相比，总体结果令人鼓舞。空间角色识别和分类的结果表明，需要合并更多的语言功能和词典，以通过SVM正确地对关系进行分类。动态空间角色标记的总体性能表明需要改进数据集和分类器训练，以实现更高的精度和召回率。