AI-assisted categorization and summarization of court rulings

The Swiss Refugee Council advocates for the protection and rights of refugees. If an asylum application is rejected by the Swiss authorities, the asylum seeker can appeal against the decision at the Federal Administrative Court. The court’s decisions are published in a publicly available database — essentially a collection of PDFs with some metadata. The Swiss Refugee Council identifies, categorizes and summarizes the relevant decisions to create an overview that assists legal representatives in their work. Currently, this overview is created manually. This is time-consuming because the court publishes around 100 decisions each week. The aim of this challenge is to develop an AI-based tool to automate this process. This tool would extract useful information such as the number of judges involved in the ruling, the asylum seeker's country of origin, and the outcome of the decision using LLMs or other ML models. In addition, the tool should create a summary of the court ruling that will help lawyers to quickly understand the essentials of a case.