CS772: Deep Learning for Natural Language Processing

Announcement

• Join the MS Teams using the code 6umhfai
• Previous iterations of the course: 2022

---

Course Details

CS772: Deep Learning for Natural Language Processing
Department of Computer Science and Engineering
Indian Institute of Technology Bombay

Time Table and Venue

• Monday: 11:35 AM to 12:30 PM
• Tuesday: 8:30 AM to 9:25 AM
• Thursday: 9:30 AM to 10:25 AM
• Venue: F.C Kohli Auditorium, KRESIT

Motivation

Deep Learning (DL) is a framework for solving AI problems based on a network of neurons organized in many layers. DL has found heavy use in Natural Language Processing (NLP) too, including problems like machine translation, sentiment and emotion analysis, question answering, information extraction and so on, improving performance on automatic systems by order of magnitude.

The course CS626 (Speech, NLP and the Web) being taught in the first semester in CSE Dept IIT Bombay for last several years creates a strong foundation of NLP covering the whole NLP stack starting from morphology to part of speech tagging, to parsing and discourse and pragmatics. Students of the course which typically number more than 100, acquire a grip on tasks, techniques and linguistics of a plethora of NLP problems.

CS772( Deep Learning for Natural Language Processing) comes as a natural sequel to CS626. Language tasks are examined through the lens of Deep Learning. Foundations and advancements in Deep Learning are taught, integrated with NLP problems. For example, sequence to sequence transformer is covered with application in machine translation. Similarly, various techniques in word embedding are taught with application to text classification, information extraction etc.

Course Content

• Background: History of Neural Nets; History of NLP; Basic Mathematical Machinery- Linear Algebra, Probability, Information Theory etc.; Basic Linguistic Machinery- Phonology, morphology, syntax, semantics
  
• Introducing Neural Computation: Perceptrons, Feedforward Neural Network and Backpropagation, Recurrent Neural Nets
  
• Difference between Classical Machine Learning and Deep Learning: Representation- Symbolic Representation, Distributed Representation, Compositionality; Parametric and non-parametric learning
  
• Word Embeddings: Word2Vec (CBOW and Skip Gram), Glove, FastText
  
• Application of Word Embedding to Shallow Parsing- Morphological Processing, Part of Speech Tagging and Chunking
  
• Sequence to Sequence (seq2seq) Transformation using Deep Learning: LSTMs and Variants, Attention, Transformers
  
• Deep Neural Net based Language Modeling: XLM, BERT, GPT2-3 etc; Subword Modeling; Transfer Learning and Multilingual Modeling
  
• Application of seq2seq in Machine Translation: supervised, semi supervised and unsupervised MT; encoder-decoder and attention in MT; Memory Networks in MT
  
• Deep Learning and Deep Parsing: Recursive Neural Nets; Neural Constituency Parsing; Neural Dependency Parsing
  
• Deep Learning and Deep Semantics: Word Embeddings and Word Sense Disambiguation; Semantic Role Labeling with Neural Nets
  
• Neural Text Classification; Sentiment and Emotion labelling with Deep Neural Nets (DNN); DNN based Question Answering
  
• The indispensability of DNN in Multimodal NLP; Advanced Problems like Sarcasm, Metaphor, Humour and Fake News Detection using multimodality and DNN
  
• Natural Language Generation; Extractive and Abstractive Summarization with Neural Nets
  
• Explainability
  

References

• Ian Goodfellow, Yoshua Bengio and Aaron Courville, Deep Learning, MIT Press, 2016.
• Dan Jurafsky and James Martin, Speech and Language Processing, 3rd Edition, October 16, 2019.
• Aston Zhang, Zachary C. Lipton, Mu Li, and Alexander J. Smola, Dive into Deep Learning, e-book, 2020.
• Christopher Manning and Heinrich Schutze, Foundations of Statistical Natural Language Processing, MIT Press, 1999.
• Daniel Graupe, Deep Learning Neural Networks: Design and Case Studies, World Scientific Publishing Co., Inc., 2016.
• Pushpak Bhattacharyya, Machine Translation, CRC Press, 2017.

• Journals:Computational Linguistics, Natural Language Engineering, Journal of Machine Learning Research (JMLR), Neural Computation, IEEE Transactions on Neural Networks and Learning Systems
• Conferences: Annual Meeting of the Association of Computational Linguistics (ACL), Neural Information Processing (NeuiPS), Int’l Conf on Machine Learning (ICML), Empirical Methods in NLP (EMNLP).

Pre-requisites

Data Structures and Algorithms, Python (or similar language) Programming skill

---