Instructor

Prof. Pushpak Bhattacharya

Semester

Spring ‘21

Course Difficulty

Slightly harder than average. The course content is straightforward. But exams can be very tricky. Assignments are easy. But project requires substantial efforts. There are numerous evaluations which might seem overwhelming but they are quite easy if you are regular.

Time Commitment Required

As much as that of a regular 6 credit course.

Grading Policy and Statistics

Lenient. Like most CS7xx, 6xx courses
AA 16
AB 30
BB 51
BC 45
CC 3
CD 1
FF 2
Total 148

Attendance Policy

No attendance policy. All lectures and slides were made available.

Pre-requisites

CS626 (Speech, NLP and the Web) offered by Prof. Pushpak himself in the previous semester was a soft prerequisite. But he went over the basics again in the initial few lectures to bring everyone up to speed. Any introductory ML course should suffice.

Evaluation Scheme

Individual effort (Total - 50%):
3 Quizzes - Total 15%
Midsem - 15%
Endsem - 20%

Team effort (Total - 50%):
4 assignments - Total 25%
Project - 25%

Topics Covered in the Course

Language Modeling,
Neural Net Language Models,
Backprop and related concepts,
Seq2seq models,
Attention mechanism,
Transformers,
Sentiment/Emotion/Dialogue analysis

Teaching Style

Online live lectures. With recordings made available offline.
Occasionally had special guest lectures by industry professionals.

Tutorials/Assignments/Projects

Assignments are very basic and easy. Project requires considerable efforts. There are regular project progress evaluations. The TAs and professor are very helpful in guiding us in the right direction in these sessions.

Feedback on Exams

All exams were mostly MCQs or short answer type. There were some instances of ambiguous framing and options. Except endsems, everything was open book.

Motivation for taking this course

I found CS626 quite interesting and this was the natural next step.

Course Highlights

Guest lectures by industry professionals from IBM, Google Research.

Course Importance

DL for NLP makes one acquainted with the concepts underlying SOTA NLP research. There are tremendous research opportunities in NLP and this course can serve as a nice gateway to them.

How strongly would I recommend this course?

8/10

When to take this course?

I took this course in my 6th semester. This course should be taken in or after 3rd year.

References Used

1) Ian Goodfellow, Yoshua Bengio and Aaron Courville, Deep Learning, MIT Press, 2016.
2) Dan Jurafsky and James Martin, Speech and Language Processing, 3rd Edition, 2019
3) Pushpak Bhattacharyya, Machine Translation, CRC Press, 2017

Review By: Harshad Ingole