Language models are needed to describe the probability of words following each other (select previous word to see).

You will use tools (SRILM) to build N-gram models in the standard format and evaluate the effects of training size, N-gram order on model performance. You may also investigate model adaptation and possibly class-based language models. If you want, you can try out RNN-based LM techniques such as the ones implemented in faster-rnnlm.

-- TimoBaumann - 06 Apr 2016