Automated Topic Categorisation of Citizens’ Contributions: Reducing Manual Labelling Efforts Through Active Learning

Abstract

Political authorities in democratic countries regularly consult the public on specific issues but subsequently evaluating the contributions requires substantial human resources, often leading to inefficiencies and delays in the decision-making process. Among the solutions proposed is to support human analysts by thematically grouping the contributions through automated means. While supervised machine learning would naturally lend itself to the task of classifying citizens’ proposal according to certain predefined topics, the amount of training data required is often prohibitive given the idiosyncratic nature of most public participation processes. One potential solution to minimise the amount of training data is the use of active learning. While this semi-supervised procedure has proliferated in recent years, these promising approaches have never been applied to the evaluation of participation contributions. Therefore we utilise data from online participation processes in three German cities, provide classification baselines and subsequently assess how different active learning strategies can reduce manual labelling efforts while maintaining a good model performance. Our results show not only that supervised machine learning models can reliably classify topic categories for public participation contributions, but that active learning significantly reduces the amount of training data required. This has important implications for the practice of public participation because it dramatically cuts the time required for evaluation from which in particular processes with a larger number of contributions benefit.