Towards a Tool for Interactive Concept Building for Large Scale Analysis in the Humanities

We develop a pipeline consisting of various text processing tools which is designed to assist political scientists in finding specific, complex concepts within large amounts of text. Our main focus is the interaction between the political scientists and the natural language processing groups to ensure a beneficial assistance for the political scientists and new application challenges for NLP. It is of particular importance to find a “common language” between the different disciplines. Therefore, we use an interactive web-interface which is easily usable by non-experts. It interfaces an active learning algorithm which is complemented by the NLP pipeline to provide a rich feature selection. Political scientists are thus enabled to use their own intuitions to find custom concepts.