CS753 - Automatic Speech Recognition
Prof Preethi Jyothi
Time Commitment Required
High, one needs to be in constant touch with the material since the pace is high; it is easy to lose track
Grading Policy and Statistics
Pretty lenient, typical of any CS 6xx or 7xx course
Class Participation has a 5% weightage. Students had to attempt weekly Moodle quizzes for it.
Linear Algebra, Python and a formal course on Machine Learning (in the institute)
Class Participation 5%
3 Assignments 40%
Topics Covered in the Course
HMM models, WFSTs, Neural Network based Acoustic models, RNNs and Language models, Feature analysis, End-to-end architecture, Search and Decoding
The professor is very proactive in the classroom. Takes doubts and clears them very well. Ample material and reading material is provided for beteer understanding
Assignments are moderate to difficult and require a lot of time to solve. The project is supposed to be any topic related to ASR but should be advanced or non-trivial enough. Intoductory ML course projects are discouraged.
Feedback on Exams
The Weekly Quizzes are pretty simple given that one is attentive in class.
Endsem is open-book, majority of the questions target the students’ conceptual understanding.
Motivation for taking this course
I always had the wish to learn how ASR systems like Alexa and Siri work. I had heard the professor taught very well.
How strongly would I recommend this course?
When to take this course?
Semester 6. Any semester third year onwards is good provided you have done the prerequisites
This course open up opportunities in the field of Automatic Speech Recognition, one can uderstand the latest research going on and participate in it too!
Slides and specific papers and textbook chapters which are mentioned at the end of each lecture
CS 753 Review By: Samyak Shah