Case Study

Decide which recording issues apply to a distant, noisy audio sample.

Case context: An annotator reviews a weak voice recording from a dataset for speech systems. The clip has strong cafeteria noise, and the speaker is several feet from the microphone. The speaker's pace is normal.

Question: Which labels should be selected, which label should be left off, and how may notes be used?

Sample answer: Select “Heavy ambient noise” and “Speaker far from mic.” Leave “Speech is too fast” off because the speaker was not rushed. Notes can capture any extra details that help explain the clip.

Key points:

  • Select heavy ambient noise.
  • Select speaker far from mic.
  • Leave speech is too fast off.
  • Use notes for extra details.

Rubric: Full credit requires selecting the two observed issues, omitting the unobserved fast-speech label, and explaining that notes can store additional observations.

0

1

Updated 2026-08-12

Contributors are:

Who are from:

Tags

Machine Learning

Deep Learning

Supervised Learning

Dive into Deep Learning @ D2L

Data Science

Machine Learning Strategy

Machine Learning Yearning @ DeepLearning.AI