# Information Extraction

> Pulling facts like names and dates out of text. They go into a neat table.

**Field:** Artificial Intelligence  
**Category:** NLP & Language  
**Also known as:** Relation Extraction, Text Information Extraction

## Think of it like

It is like using a pen to mark the dates and names in a letter. Then you copy them into a form.

## Definition

Information extraction pulls set facts out of free text and puts them in a neat table. It finds things like names, dates and amounts, and how they link, such as who works for which company. Named entity recognition is often the first step. The output has fixed fields, so a database or a sheet can use it. Today, a language model can do much of this from a short prompt and an output format.

## Example

Emma's firm gets 300 contracts a month. A model pulls each one's parties, start date and price into a spreadsheet.

## Example code

```json
{
  "party": "Acme Ltd",
  "start_date": "2026-03-01",
  "price_eur": 12000
}
```

One contract, and the three fields pulled out of its pages of text.

## Why it matters

Most facts live in emails, reports and forms written for people. Extraction turns them into data you can search and add up.

## Common mistake

Trusting every field it pulls. A model can fill in a date that is not in the text, so check a sample by hand.

## Three principles

- **Fixed fields.** Decide what to pull first.
- **Text to table.** Free text in, rows out.
- **Check a sample.** Models can make fields up.

## Related terms

- [Named Entity Recognition (NER)](https://learn.benibix.com/term/ai/named-entity-recognition)
- [Structured Output (LLM)](https://learn.benibix.com/term/ai/structured-output)
- [Intelligent Document Processing (IDP)](https://learn.benibix.com/term/ai/intelligent-document-processing)
- [Unstructured Data](https://learn.benibix.com/term/data/unstructured-data)
- [Structured Data](https://learn.benibix.com/term/data/structured-data)
- [Natural Language Understanding (NLU)](https://learn.benibix.com/term/ai/natural-language-understanding)

## Told apart from

- [Named Entity Recognition vs Information Extraction](https://learn.benibix.com/compare/named-entity-recognition-vs-information-extraction): People think finding names and pulling out facts are the same job. Named entity recognition is only the first step. It finds the names in a text and says what kind each one is. Information extraction goes on to fill set fields, such as which firm paid which sum on which date.

![One customer message, 'My card was charged twice, please fix it today', points to three outputs: a group, billing; an intent, refund request; and a set of fields for the problem, the count and the deadline.](https://learn.benibix.com/diagrams/ai5-text-analysis.svg)

---

Source: https://learn.benibix.com/term/ai/information-extraction
Author: Behnia Zarei (a developer in Vienna, Austria, who builds and teaches data, AI and business intelligence at BENIBIX)
Publisher: BENIBIX Learn (https://learn.benibix.com/)
Last updated: 2026-11-18
Licence: CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). Quote freely with attribution.
