Lesson 1 of 4
Reading a dataset card
The questions to answer before you download anything.
10 min 3-question quiz 3 guides to read next
A dataset card is the documentation that comes with a dataset. Every dataset on this hub has one, and reading it first saves hours later. Open the Adult census income card and look for answers to these questions:
- What is one row? Here, one person from the 1994 US Census.
- Where did it come from, and when? Data from 1994 describes 1994. A model trained on it knows nothing about incomes today.
- What does each column mean, and in what units? Is
fnlwgta weight in kilograms? (No — it is a census sampling weight.) - How are missing values written? In this dataset, as
?. Elsewhere you will see blanks,NA,-999or***. - Are there quirks? The card warns that
adult.teststarts with a comment line and that its labels end in a full stop. - What is the licence? CC BY 4.0: you can reuse it, including commercially, as long as you credit the source.
Look before you load
Use the Data viewer tab to see the first rows of a file. You will spot things no documentation mentions: the wine quality CSVs are separated by semicolons, not commas, and the GISTEMP file has a title line above the header.
If a dataset has no card, no licence or no clear source, treat that as a warning sign rather than an inconvenience.
Check your understanding
3 questions · pass with 2 correct
Enrol for free to save your progress, unlock every lesson and earn a certificate.
Sign in to enrolFurther reading
Guides that go deeper on this lesson.
-
How to Read a Dataset Card
The questions a dataset card should answer — what one row is, where the data came from, its licence and its quirks — before you use it.
1 min read
-
Datasheets for Datasets
A structured way to document how a dataset was created, what it contains and how it should be used.
2 min read
-
Finding Open Datasets
Where to find reliable public datasets for learning and projects, and how to judge whether a dataset is suitable.
1 min read