The repo should help students to prepare for the exam Advanced Speech Recognition. In case you find any errors in the solutions of the exercises feel free to write an issue.
- Exercise 1
- Exercise 2
- Exercise 3 graph decoding uses greedy approach -> should use viterbi decoding
- Exercise 4 graph decoding uses greedy approach -> should use viterbi decoding, use np.log1p() instead of customized getValue function, apply graph rescoring using confidence measures
- Exercise 5
- Exercise 6