YSDA Natural Language Processing course

• Lecture and seminar materials for each week are in ./week* folders
• Create cloud jupyter session from this repo -
• YSDA homework deadlines are listed in Anytask course page.
• Any technical issues, ideas, bugs in course materials, contribution ideas - add an issue
• Installing libraries and troubleshooting: this thread.

Syllabus

• week01 Embeddings

  • Lecture: Word embeddings. Distributional semantics, LSA, Word2Vec, GloVe. Why and when we need them.
  • Seminar: Playing with word and sentence embeddings.

• week02 Text classification

  • Lecture: Text classification. Classical approaches for text representation: BOW, TF-IDF. Neural approaches: embeddings, convolutions, RNNs
  • Seminar: Salary prediction with convolutional neural networks; explaining network predictions.

• week03 Language Models

  • Lecture: Language models: N-gram and neural approaches; visualizing trained models
  • Seminar: Generating ArXiv papers with language models

• week04 Seq2seq/Attention

  • Lecture: Seq2seq: encoder-decoder framework. Attention: Bahdanau model. Self-attention, Transformer. Pointer networks. Attention for analysis.
  • Seminar: Machine translation of hotel and hostel descriptions

• week05 Structured Learning

  • Lecture: Structured Learning: structured perceptron, structured prediction, dynamic oracles, RL basics.
  • Seminar: POS tagging

• week06 Expectation-Maximization

  • Lecture: Expectation-Maximization and Word Alignment Models
  • Seminar: Implementing expectation maximizaiton

• week07 Machine translation

  • Lecture: Machine Translation: a review of the key ideas from PBMT, the application specific ideas that have developed in NMT over the past 3 years and some of the open problems in this area.
  • Seminar: presentations by students

• week08 Transfer learning and Multi-task learning

  • Lecture: What and why does a network learn: "model" is never just "model"! Transfer learning in NLP. Multi-task learning in NLP. How to understand, what kind of information the model representations contain.
  • Seminar: Improving named entity recognition by learning jointly with other tasks

• week09 Domain Adaptation

  • Lecture: General theory. Instance weighting. Proxy-labels methods. Feature matching methods. Distillation-like methods.
  • Seminar: Adapting general machine translation model to a specific domain.

• week10 Dialogue Systems

  • Lecture: Task-oriented vs general conversation systems. Overview of a framework for task-oriented systems. General conversation: retrieval and generative approaches. Generative models for general conversation. Retrieval-based models for general conversation.
  • Seminar: Simple retrieval-based question answering

• week11 Generative models

  • Lecture: generative models recap, generative adversarial networks, variational autoencoders and why should you care about them.
  • Seminar: semi-supervised dictionary learning with adversarial networks

• week12-13 - TBA

Contributors & course staff

Course materials and teaching performed by

• Elena Voita - course admin, lectures, seminars, homeworks
• Boris Kovarsky - lectures, seminars, homeworks
• David Talbot - lectures, seminars, homeworks
• Sergey Gubanov - lectures, seminars, homeworks
• Just Heuristic - lectures, seminars, homeworks