Code and data for our VarDial 2018 paper on spoken versus written image descriptions - View it on GitHub