Lesson 48 min

Text Classification

Sorting text into categories — the most common NLP task.

What you will learn

Explanation

Text classification assigns a label to a piece of text — spam/not-spam, positive/negative, ticket category.

The classic pipeline is: tokenise → embed → classifier (e.g. logistic regression). You train it on labelled examples.

It's everywhere because it's simple, fast, and useful — and you can often do it with a small model instead of a costly LLM.

python

1	`# sentiment classifier sketch`
2	`tokens = tokenize(text)`
3	`vector = embed(tokens)`
4	`label = classifier.predict(vector) # 'positive' or 'negative'`

Real-world use

Support desks auto-route tickets to the right team using text classification on the message body.

Common mistakes

• Reaching for an LLM when a small, cheap classifier would handle high-volume sorting.

Practice

Pick a text-classification task and list the categories and the labelled data you'd need.

Knowledge check

0/1 answered

1. Spam detection is an example of...

Answer all questions to check.