Developing Machine Learning Models for the Analysis ofCitizens’ Contributions from E-Participation

Citizen participation is increasingly relevant in acquiring local knowledge for urban projects. However, the increasing number of participants has created a problem of data overload, as manual analysis is hardly feasible due to its expense, time intensity, and slowness. AI and ML can help reduce or solve this issue. However, due to the lack of proper ML-based approaches, we investigate how textual contributions from citizens can be analyzed using machine learning techniques. To achieve this, we followed the knowledge discovery in the database framework, collected data, and trained several machine learning models, which we analyzed and compared. Our findings demonstrate that urban development contributions often cover multiple topics, making classification challenging, which also corresponds to the length of citizens’ contributions. Transformer models, however, show remarkable precision when compared to SVM models. With our findings, we contribute to the analysis of citizen contributions to support democratic processes and scalable citizen participation.