Language modelling

​IA161 Advanced NLP Course, Course Guarantee: Aleš Horák

Prepared by: Vít Baisa

State of the Art

The goal of a language model is a) to predict a following word or phrase based on a given history and b) to assign a probability (= score) to any possible input sentence. In the past, this was achieved mainly by n-gram models known since WWII. But recently, the buzzword deep learning penetrated also into language modelling and it turned out to be substantially better than Markov's n-gram models.

References

1. Bengio, Yoshua, et al. "A neural probabilistic language model." The Journal of Machine Learning Research 3 (2003): 1137-1155.
2. Mikolov, Tomas, et al. "Efficient estimation of word representations in vector space." arXiv preprint arXiv:1301.3781 (2013).
3. Mikolov, Tomas, et al. "Distributed representations of words and phrases and their compositionality." Advances in Neural Information Processing Systems. 2013.
4. Chelba, Ciprian, et al. "One billion word benchmark for measuring progress in statistical language modeling." arXiv preprint arXiv:1312.3005 (2013).

Practical Session

We will build a simple language model (skip-gram) which has very interesting properties. When trained properly, the vectors of words obey simple space arithmetics, e.g. vector "king" − vector "man" + vector "woman" ~= vector "queen". We will train this model on a large Czech and English corpora and evaluate the results.