Machine translation

Lecture 1. Introduction
Course overview. Machine translation applications. History of Machine Translation in the world, USSR and Russia. Georgetown experiment (1954)
Lecture 2. Introduction to Rule-based Machine Translation
Linguistics and Machine Translation. Morphological Analysis of texts. Morphological categories and morphological features. Morphological ambiguity and methods of its resolution.
Lecture 3. Syntactic Analysis and Generation
Syntactic Analysis (Parsing) and Generation. Methods of Syntactic Stucture Representation of Sentences. Computer dictionaries. Syntactic and Semantic Features. Subcategorization Frames.
Lecture 4. Semantics and Machine Translation
Using deep semantic data in machine translation. Translation with the help of ontologies in individual subject domains.
Lecture 5. Data-driven machine translation: an intro
Why is translation hard? How to build a machine translation system from data? What have been the main breakthroughs? How do we evaluate a system? Statistical modelling: MLE and EM
Lecture 6. Word alignment
Maximum likelihood estimation from data. Expectation-maximization algorithm for handling hidden data. IBM word alignment models.
Lecture 7. Phrase-based MT
Language models (n-gram). Phrase-based translation models. Phrase-based decoding. Hierarchical models. Syntax and Reordering. Syntax and Morphology.
Lecture 8. Neural Machine Translation
Word embeddings. Neural network language models. Sequence models: RNN, LSTM, GRU etc.. Recurrent Continuous Translation models. Encoder-Decoder models. Word embeddings practical.
Lecture 9. Neural Machine Translation, part 2
Encoder-Decoder models (continued). Beam-search. Attention models. Approximating softmax. Open vocabulary NMT: subword units.
Lectures 10-11. Current directions in neural MT
Scaling up NMT. Exploiting monolingual data. Sequence level learning (and RL) in NMT. What does NMT learn? Experimental architectures.