Extraction of Spatial Relation in Arabic Text Using Rule-Based Approach

Relation extraction refers to identification of different semantic relations between named entities. Discovering spatial relations between named entities (NEs) is challenging task with many applications. However, the existing researches only focus on general spatial relation extraction in English and some other languages. This paper introduces a rule-based approach to spatial relation extraction in Arabic natural language text. A set of syntactical rules and patterns of spatial relations are induced and then formalized based on spatial relation annotation corpus, morphological information, spatial terms and location named entity recognition technology. Three types of spatial relations are focused on, which are topological relations, directional relations, and distance relations. Then, these rules and patterns are applied to identify and classify these three types of spatial relations in Arabic documents. The system has been successfully tested on 30 geographical documents that include a mixture of types of spatial relations. The result achieved 86.06% of F-measure. This illustrates that our approach is effective in extracting spatial relations in Arabic texts and its performance is satisfactory.