Deep Learning for Speech and Natural Language Processing

This hands on course aims to build foundations of deep learning for NLP and speech, while also introducing the students to current state of the art of NLP.

June 24, 2019
5 Days
PES Participants:
Rs. 5,000
Non-PES Participants:
Rs. 10,000



Crucible of Continuing Education (CCE)
PES University Campus
100 Feet Ring Road, BSK III Stage
Bengaluru – 560 085   View map

Course Objectives(What will you learn)

This hands on course aims to build foundations of deep learning for NLP and speech, while also introducing the students to current state of the art of NLP. After completing this course the students will be able to:

  • Find it easier to delve much deeper into research in the field of speech and NLP
  • Apply the concepts learned as per the industry requirements in their jobs.


  • Knowledge of Python
  • Knowledge of basic ML concepts – Probability, basic understanding of supervised and unsupervised learning, linear regression, classification, naive bayes classification, perceptron, single layer and multi layer neural networks, backpropogation.


Based on hands-on exercises. Students will be expected to work in teams of 3. The hands-on exercises will be broken down into a number of tasks. After the hands-on exercises, the team will be asked to write down a description of the approach used, and results. Additionally, each team member will be asked to write down their specific contribution to the exercise. ​Saying “we all worked together” will mean that the evaluator will use their own judgement as to how much work each person has done, and grades will be given accordingly.

Out station students / candidates have to make their arrangements for accommodation and boarding

Course Outline and schedule


  • Introduction
  • Word vector representations: basic: word2vec, glove,
  • advanced: ELMO, BERT and current state of the art
  • Applications and evaluations of word embeddings.
  • Hands on exercises:
    • Training and testing word embeddings, using embeddings in downstream NLP tasks.


  • Introduction to Speech recognition and processing with Kaldi toolkit
  • Feature extraction
  • Acoustic modeling
  • Language modeling
  • Decoding
  • Hands on exercises:
    • Training a GMM-HMM model for speech
    • Training a neural net model for speech.


  • Recurrent neural networks and language modelling (word level and character level)
  • GRU, LSTM and attention mechanisms
  • Seq2seq models
  • Hands on exercises:
    • Building language models using RNNs for next word prediction.
    • Text generation using RNNs and LSTMs.


  • Deep delve into seq2seq models and their applications
  • Transformer models – basics and deeper look into BERT, openAI GPT and GPT2 architectures.
  • Machine translation – current state of the art
  • Hands on exercises:
    • Simple machine translation system using seq2seq architecture.
    • Incorporating embeddings into the model
    • Testing out transformers


  • Applications of NLP – sentiment analysis and question answering
  • NLP models for sentence and document classification
  • CNNs and GANs for text processing
  • Hands on exercises:
    • Applying the concepts learned so far to
      • Sentiment analysis
      • QA systems
      • Text classification models (on raw real time data)

Share this page