'''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] [[http://ad-publications.informatik.uni-freiburg.de/CIKM_freebase_qa_BH_2015.materials|Aqqu Materials + Paper + Link zur Demo]]

[2] [[http://ad-research.informatik.uni-freiburg.de/benchmarks/webquestions/|WebQuestions Benchmark]] [[https://www.microsoft.com/en-us/download/details.aspx?id=52763|WQSP (verbesserte Version)]]

[3] [[http://ad-research.informatik.uni-freiburg.de/benchmarks/free917/|Free 917 Benchmark]]

[4] [[http://ad-research.informatik.uni-freiburg.de/benchmarks/|Simple Questions]]

[5] [[https://docs.google.com/spreadsheets/d/1vglGqdrEntJNN4eTrNQLd6A6G1MEZ5dsDDY4asaR8QQ/edit|Problematic questions]] (gives a good feeling, why QA is so hard)