Information-optimal Abstaining for Reliable Classification of Building Functions

Abstract

Abstract. In the past decade, major breakthroughs in sensor technology and algorithms have enabled the functional analysis of urban regions based on Earth observation data. It has, for example, become possible to assign functions to areas in cities on a regional scale. With this paper, we develop a novel method for extracting building functions from social media text alone. Therefore, a technique of abstaining is applied in order to overcome the fact that most tweets will not contain information related to a building function albeit they have been sent from a specific building as well as the problem that classification schemes for building functions are overlapping.

Geoinformation Harvesting From Social Media Data: A community remote sensing approach

“…Urban areas are multilingual spaces [51]- [53] and the set of languages discovered in social media posts in cities is diverse [54] (of course, English is dominant on Twitter [55] and Chinese on Weibo). However, multilingual approaches are rare in earth sciences and so offer interesting research opportunities, such as ensemble models covering all languages to classify building functions in urban areas [56] or investigating the information density of Japanese or Chinese social media postings with respect to English [57] within the context of urban land use tasks.…”