Type: Project or Thesis, preferably both in succession. Nice topic if you like Deep Learning and Natural Language Processing. The topic is also relatively "stand-alone" in that it does not rely on other complex systems but you can design and build everything yourself from the ground up.

Goal: Design and build a system that accepts a (relatively simple) question in natural language and automatically corrects typos etc. For example, given the question "wo inveted pyton", output "who invented python". This should be realized with a character-based language model learned using deep learning (e.g., with an RNN). One question is whether solving this problem for questions is easier or harder than solving it for arbitrary sentences.

[1] Aqqu Materials + Paper + Link zur Demo

[2] WebQuestions Benchmark WQSP (verbesserte Version)

[3] Free 917 Benchmark

[4] Simple Questions

[5] Problematic questions (gives a good feeling, why QA is so hard)