Skip to lesson

Path: Working well on data projects

Do not trade accuracy for throughput

Decide when to flag, when to skip, and when a guess costs the project more than an unfinished item.

Starting level
Comfortable with the basics
Time
18 minutes
Public lesson guide
EN + हिंदी

You will learn to

Decide when to flag, when to skip, and when slowing down is the correct choice.

Why this matters

A flagged item is honest, finished work. A guessed one looks finished and is trusted later by people who have no reason to check it.

Three things to remember

  • Flag the unclear item; never guess to keep the count moving.
  • Speed comes from fewer re-reads, not from fewer checks.
  • Tell the project what you flagged when you hand the batch in.

See one example

Task

An hour is left in a transcription batch and a handful of lines are genuinely inaudible.

Try the judgement yourself

Both answers respond to the task above. One of them does the job better. Which, and why?

Make something yourself

Write your own working rule for a batch under time pressure: what you flag, what you skip, when you slow down, and what you tell the project at hand-in.

A useful starting pointThe flag condition, the skip condition, the slow-down signal, and the hand-in note.

Read further

The source is optional. The complete teaching and practice are available inside AI2Bharat.

Read the primary or official source: Gebru et al., Datasheets for Datasets (arXiv)