AI glossary · Business
Text classification
In one sentence
Text classification is the AI task of automatically assigning one or more categories to a piece of text, such as an email or a message: type of request, topic, priority or language. It's the basis for sorting, prioritising and routing work.
Also known as: text categorisation · document classification · automatic text classification · text tagging
What is text classification?
It’s one of the oldest and most useful tasks in natural language processing. Your email spam filter is text classification: for each message it picks one of two categories. The same principle applies to any question that can be answered with a label: what’s this email about? Which department should it go to? Is it urgent?
Put it to work
An agent that sorts and answers your email
From €39/mo + VAT
It sorts the inbox by priority and drafts replies with context.
See the email agentThere are several types, depending on the categories:
| Type | What it decides | Example |
|---|---|---|
| Binary | One of two | Spam / not spam |
| Multi-class | One of several | Order, invoice, issue, enquiry |
| Multi-label | Several at once | Issue + urgent + key account |
| Hierarchical | Category and sub-category | Issue > Delivery > Delay |
Two well-known uses are text classification under another name: intent detection in a chatbot, and sentiment analysis.
Related to this term
UTIO Email
- Add-on for any voice or WhatsApp plan
- 2 mailboxes and 2,000 emails a month
- Standalone email: from €89/mo
From €39/mo + VAT
See the email agentHow it works
- Define the categories with clear criteria. If two people can’t agree on how to label an email, the model won’t get it right either.
- Turn the text into numbers that capture its meaning.
- A model assigns the category. There are three approaches:
- Rules: keywords (“invoice” → Accounts). Fast but brittle.
- Trained model: learns from labelled examples. Accurate and cheap per text, but needs maintaining.
- Language model with instructions: understands a description of each category. Flexible, no training needed.
- Act on the label: move it to a folder, assign it to someone, raise its priority or reply.
Example
A van hire company in Norwich gets around 150 emails a day in a shared mailbox. Every morning someone spent an hour reading and forwarding them. With text classification, each email arrives tagged by type (booking, change, invoice, damage, other) and by urgency. Damage reports for vans due back today rise to the top; marketing emails drop out of sight.
Text classification at UTIO
UTIO’s email agent applies classification to your mail: it sorts it into threads and prioritises them, urgent at the top and automated mail out of the way, and quarantines suspicious and irrelevant messages. Every email lands on the contact’s CRM record with an owner, a status and a next step in the pipeline, alongside their calls and WhatsApp messages.
On top of that sorting, the agent drafts replies in your tone, which you review and send, or let go out on their own depending on the autonomy level you choose.
Frequently asked questions
What's the difference between text classification and sentiment analysis?
Sentiment analysis is one specific kind of text classification, where the categories are positive, negative or neutral. Text classification is broader: the categories can be anything, such as topics, departments or urgency levels.
Do you need to train a model to classify text?
Not always any more. You used to need hundreds of labelled examples per category. With language models, describing the categories and giving a few examples is often enough. At very high volumes, a small trained model is still cheaper.
How do I know whether the classification is working?
Review a sample of classified texts and count how many it gets right per category. Pay most attention to the costly mistakes, such as an urgent email being marked as irrelevant.
Related terms
- Natural language processing (NLP)
Natural language processing (NLP) is the branch of artificial intelligence that lets machines work with human language: read, analyse, classify, translate and produce text or speech the way people use it.
- Sentiment analysis
Sentiment analysis is the AI technique that detects the emotional tone of a text (positive, negative or neutral, and sometimes emotions such as anger or urgency) so you know how the writer feels without reading everything by hand.
- Intent
An intent is the goal behind what a person writes or says to a chatbot or AI agent, such as booking, cancelling or asking for a price. Recognising it lets the system decide what to do with the message.
- Machine learning
Machine learning is the branch of AI in which a system learns to do a task from examples, rather than following hand-written rules, and gets better as it sees more data.
- Embeddings
An embedding is a list of numbers that represents the meaning of a piece of text, an image or audio. Items with similar meaning have embeddings that sit close together, which lets you search by meaning rather than by exact words.
Updated on