Supporting Land Reuse of Former Open Pit Mining Sites using Text Classification and Active Learning

Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th International Joint Conference on Natural Language Processing, pages 4141–4152 August 1–6, 2021. ©2021 Association for Computational Linguistics 4141 Supporting Land Reuse of Former Open Pit Mining Sites using Text Classification and Active Learning Christopher Schr¨oder1,5, Kim B¨urgl1,4,5, Yves Annanias2,5, Andreas Niekler1,5, Lydia M¨uller1,4,5, Daniel Wiegreffe2,5, Christian Bender3,5, Christoph Mengs3,5, Gerik Scheuermann2,4,5, and Gerhard Heyer1,4,5 1Natural Language Processing Group 2Image and Signal Processing Group 3Institute of Public Finance and Public Management 4Institute for Applied Informatics (InfAI), Leipzig, Germany 5Leipzig University, Germany {schroeder,buergl,annanias,aniekler}@informatik.uni-leipzig.de {lydia,daniel,scheuermann,heyer}@informatik.uni-leipzig.de {bender,mengs}@wifa.uni-leipzig.de Abstract Open pit mines left many regions worldwide inhospitable or uninhabitable. Many sites are left behind in a hazardous or contaminated state, show remnants of waste, or have other restrictions imposed upon them, e.g., for the protection of human or nature. Such informa- tion has to be permanently managed in order to reuse those areas in the future. In this work we present and evaluate an automated workflow for supporting the post-mining management of former lignite open pit mines in the eastern part of Germany, where prior to any planned land reuse, aforementioned information has to be acquired to ensure the safety and validity of such an endeavor. Usually, this information is found in expert reports, either in the form of paper documents, or in the best case as dig- itized unstructured text—all of them in Ger- man language. However, due to the size and complexity of these documents, any inquiry is tedious and time-consuming, thereby slow- ing down or even obstructing the reuse of re- lated areas. Since no training data is available, we employ active learning in order to perform multi-label sentence classification for two cat- egories of restrictions and seven categories of topics. The final system integrates optical character recognition (OCR), active-learning- based text classification, and geographic infor- mation system visualization in order to effec- tively extract, query, and visualize this infor- mation for any area of interest. Active learn- ing and text classification results are twofold: Whereas the restriction categories were rea- sonably accurate (>0.85 F1), the seven topic- oriented categories seemed to be complex even for human annotators and achieved mediocre evaluation scores (<0.70 F1). 1 Introduction In many parts of the world, raw materials were mined in open pit mines during the last century, leaving many of these regions inhospitable or unin- habitable. To put these regions back into use, en- tire stretches of land must be renaturalized, which means that land must be ecologically restored with the aim to ultimately increase biodiversity, or recul- tivated, which means its productivity must be re- stored, e.g., reused for agriculture, recreational ar- eas, industrial parks, solar and wind farms, or as building land (Luc et al., 2015). In the following, we subsume both renaturalization and recultivation under land reuse. For land reuse, it is essential that all relevant information about the sites is retained, which used to be recorded in the form of textual reports. Such reports include …

Content truncated to 500 words