# Long-form Extraction

- Published on May 6, 2026
- 2 minute(s) read

Hyperscience’s solution for documents that consist of longer paragraphs of text that span across multiple pages is called _Long-form Extraction_. It builds upon the Long-form Extraction model. To learn more, see [Long-form Extraction Field ID model](https://help.hyperscience.ai/v43/docs/training-a-new-field-identification-model#long-form-extraction-field-id-model).

Long-form Extraction is available for:

- Flexible Extraction
- Field ID & Field ID QA

To make the extraction of longer text possible, we implemented a new data type called “Clause.”

The Clause data type can only be added to fields, not tables, in the Layout Editor.

## Prerequisites

To use Long-form Extraction, you first need to ensure that your infrastructure, data type, and engine type are set up correctly.

### GPU trainer

A GPU trainer is required in order to use Long-form Extraction. To learn more about the technical requirements for GPUs, see [Infrastructure Requirements](https://help.hyperscience.ai/deployment/docs/infrastructure-requirements#gpu) (for Docker and Podman) or [Kubernetes Installation Overview](https://help.hyperscience.ai/deployment/docs/kubernetes-installation-requirements#gpus-for-trainers).

If you have an on-premise deployment of Hyperscience, you need to follow the steps in the "Enabling Trainers with GPUs" article for [Docker](https://help.hyperscience.ai/deployment/docs/enabling-trainers-with-gpus-in-on-premise-docker-deployments), [Podman](https://help.hyperscience.ai/deployment/docs/enabling-trainers-with-gpus-in-on-premise-podman-deployments), or [Kubernetes](https://help.hyperscience.ai/deployment/docs/enabling-trainers-with-gpus-in-on-premise-kubernetes-deployments) before using your trainer.

### Clause data type

Before using Long-form Extraction, make sure to set the data type of the long-form field to **Clause**. Learn more in [Creating Semi-Structured Layouts](https://help.hyperscience.ai/v43/docs/creating-semi-structured-layouts).

Custom Data types also work with Clause, as long as the ML Configuration is set to **Entry - Clause**.

Learn more about ML Configurations in [Creating Data Types with ML Configurations](https://help.hyperscience.ai/v43/docs/creating-data-types-with-ml-configurations).

### Long-form Extraction Engine type

The default Field Identification models use a generic engine type that cannot extract data points from documents with unstructured text. To learn more, see [Training a New Field Identification Model](https://help.hyperscience.ai/v43/docs/training-a-new-field-identification-model).

To extract data points from Unstructured documents using a specific layout, you need to select the **Long Form Extraction** engine type for that layout before training the model.

Follow the steps below to change the engine type of your layout:

1. Go to **Library** \> **Layouts** and open your layout.
2. On the **Configuration** card, click the **Edit** button under the **Engine Type** setting.
3. Select **Long Form Extraction** from the drop-down menu.
4. Click **Change Type**.
5. Retrain your model after you’ve changed the engine type of the layout.

## Using Long-form Extraction

1. Annotate your documents by using the **Add another text segment** option for fields that span across multiple pages. Learn more in [Multiple bounding boxes for fields](https://help.hyperscience.ai/v43/docs/field-identification#multiple-occurrences-model).
2. Preview the transcribed values on the right-hand side of the Document Viewer.

## Long-form Extraction limits

Below is a list of the limits that apply for training a Long-form Extraction model.

- 2,000 text segments per page
- 200 pages per document
- 200,000 text segments per document
- 100,000,000 text segments total

To learn more about segments, see [Text Segmentation](https://help.hyperscience.ai/v43/docs/segmentation).