Analysing the Usage of Spatial Prepositions in Short Messages

Abstract

Spatial prepositions such as in, on and near are important to describe where things are located in relation to other geographic features. Location-based services (LBS) usually disregard such spatial prepositions. Their automatic detection and interpretation is challenging, because prepositions are quite often used in non-spatial context (e.g., “in the afternoon”). This paper analyses spatial relations in short messages. Short messages typically have special characteristics (e.g., slang, abbreviations, etc.) and thus represent a special type of natural language. A sample corpus of short messages was used to extract descriptions based on spatial prepositions and to analyse their commonness of use. A frequency-based probability for each term to be spatial was calculated, which can serve as an indicator of a verbal spatial description and support the development of intelligent spatial language interpretation in automatic systems.