Interactive Poetry Generation using Recurrent Neural Networks

Theoretical (Analytical):

Practical (Implementation):

Literature Work:


  • Finding the best neural network architecture for a given goal can be challenging, especially when it is not possible to assess the output quality of a network automatically.
  • The goal is to design and implement new interactive visual text synthesis methods to generate poetry (e.g. rap lyrics).
  • Use Visual Analytics principles for comparing and explaining outputs of multiple RNNs (recurrent neural networks).

Problem Statement

  • How can we use interactive neural network text synthesis methods to generate poetry?
  • Can we incorporate user feedback to improve text synthesis methods?  


  • Review existing approaches and text synthesis methods using neural networks
  • Work with a large text dataset (approx. 120.000 rap lyrics)
  • Implement different neural networks for text synthesis
  • Implement an interactive tool for text synthesis using neural networks


  • Basic knowledge about information visualization.
  • Good knowledge in machine learning.
  • Good programming skills in Python or C/C++ and web programming skills (HTML/ JavaScript/D3).


  • Bachelor/ Master
  • 3/6 Month Project, 3/6 Month Thesis



  • U. Schlegel, E. Cakmak, J. Buchmüller and D. A. Keim.
    G-Rap: interactive text synthesis using recurrent neural network suggestions.
    European Symposium on Artificial Neural Networks, Computational Intelligence and Machine Learning (ESANN), 2018