Towards Intelligent Legal Advisors for Document Retrieval and Question-Answering in German Legal Documents
Christoph Hoppe, David Pelkmann, Nico Migenda, Daniel Hötte, Wolfram Schenck
Abstract
The legal system is one of the most important pillars of human society. While digitization is integrated into many areas of everyday life, the legal system is still very traditionally positioned. Recent advances in storing and processing large number of documents initiated the work of intelligent legal advisors. While first approaches in the English language suggest an enormous potential to generate knowledge from legal documents, there are no approaches in the German legal system. We present an intelligent legal advisor based on semantic document retrieval, to improve knowledge extraction from German legal documents. In addition, we set up a question-answering system. We implemented a BERT and a BM25-model for German document retrieval in legal documents. The approach is validated on a data set consisting of German question-answer pairs.