DEVELOPMENT OF AN ANALYTICAL SYSTEM FOR PROCESSING CITIZEN OBJECTIONS

In the context of infrastructure planning, citizen objections play a central role in the participation process and in political decision-making. As these objections are often unstructured and not machine-readable, processing them is time-consuming and causes delays and inefficiencies in the management of public concerns. The aim of this study is to develop an analytical system that digitally records analog citizen objections, determines their scope and classifies them thematically. Based on this, it will be investigated to what extent automated pre-processing and categorization of objections is possible in order to make the participation process more efficient. In addition, the extent to which artificial intelligence (AI), in particular natural language processing (NLP), can support the pre-processing of objections in order to further improve the efficiency of the participation process through an automated analysis will be investigated. The analytical system developed categorizes objections using predefined keywords derived from a manual analysis of several infrastructure procedures. In the process, 13 different subject areas and numerous relevant keywords were identified based on a document analysis which were subsequently integrated into the system. The automated categorization within the analytical system was compared with a manual analysis and showed a high level of agreement. In addition, in some cases the system recognized keywords that were overlooked in the manual analysis, indicating a high level of accuracy. The application to other data sets also confirms that the system can be reliably transferred to different data sets and enables consistent thematic pre-sorting. Investigations into the possible use of AI in this context showed that AI can contribute to an improvement in the categorization of objections. However, it became clear that AI-generated terms were often specific to the respective procedure and could not be easily transferred to other data sets. Consequently, targeted control of the AI parameters appears necessary to ensure a more precise and generally valid integration into the analytical process. In future studies, the transferability of the analytical system, which only looked at public objections in the context of road construction projects, to other infrastructure sectors should be investigated. Furthermore, the investigation and implementation of other use cases, such as a duplication check of objections or the generation of response suggestions based on past responses, appears promising.