Short Answer

How Human Skill Affects Label Collection

Question: In a task that people can do reliably, how does that affect the process of creating labeled training data?

Sample answer: When people can perform the task accurately, labels are easier to collect because human annotators can produce dependable ground-truth examples with relatively little error. That reduces the difficulty of building the dataset.

Key points:

  • Tasks that humans can do well are easier to label.
  • Human annotators can produce more accurate labels.
  • Reliable labels are easier to obtain when the task is straightforward for people.

Rubric: A correct response should explain that strong human performance makes labeling easier by allowing annotators to generate accurate ground-truth labels.

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