Form Recognizer general document model (preview)

The General document preview model combines powerful Optical Character Recognition (OCR) capabilities with deep learning models to extract key-value pairs and entities from documents. General document is only available with the preview (v3.0) API. For more information on using the preview (v3.0) API, see our migration guide.

The general document API supports most form types and will analyze your documents and associate values to keys and entries to tables that it discovers. It is ideal for extracting common key-value pairs from documents. You can use the general document model as an alternative to training a custom model without labels.

General document features

• There is no need to train a custom model to extract key-value pairs.

• A single API is used to extract key value pairs, entities, text, tables, and structure from documents.

• It is a pre-trained model that will be periodically trained on new data to improve coverage and accuracy.

• The general document model supports structured, semi-structured, and unstructured data.

Sample document processed in the Form Recognizer Studio

Development options

The following resources are supported by Form Recognizer v3.0:

Try Form Recognizer

See how data, including tables, values, and entities, is extracted from forms and documents using the Form Recognizer Studio or our Sample Labeling tool. You'll need the following:

• An Azure subscription—you can create one for free

• A Form Recognizer instance in the Azure portal. You can use the free pricing tier (F0) to try the service. After your resource deploys, select Go to resource to get your API key and endpoint.

Form Recognizer Studio (preview)

1. On the Form Recognizer Studio home page, select General documents

2. You can analyze the sample document or select the + Add button to upload your own sample.

3. Select the Analyze button:

   

   Try Form Recognizer Studio

Key-value pairs

Key value pairs are specific spans within the document that identify a label or key and its associated response or value. In a structured form, this could be the label and the value the user entered for that field or in an unstructured document it could be the date a contract was executed on based on the text in a paragraph. The AI model is trained to extract identifiable keys and values based on a wide variety of document types, formats, and structures.

Keys can also exist in isolation when the model detects that a key exists, with no associated value or when processing optional fields. For example, a middle name field may be left blank on a form in some instances. Key value pairs are always spans of text contained in the document and if you have documents where same value is described in different ways, for example a customer or a user, the associated key will be either customer or user based on what the document contained.

Entities

Natural language processing models can identify parts of speech and classify each token or word. The named entity recognition model is able to identify entities like people, locations, and dates to provide for a richer experience. Identifying entities enables you to distinguish between customer types, for example, an individual or an organization. The key value pair extraction model and entity identification model are run in parallel on the entire document and not just on the values of the extracted key value pairs. This ensures that complex structures where a key cannot be identified is still enriched by identifying the entities referenced. You can still match keys or values to entities based on the offsets of the identified spans.

• The general document is a pre-trained model and can be directly invoked via the REST API.

• The general document model supports named entity recognition (NER) for several entity categories. NER is the ability to identify different entities in text and categorize them into pre-defined classes or types such as: person, location, event, product, and organization. Extracting entities can be useful in scenarios where you want to validate extracted values. The entities are extracted from the entire content and not just the extracted values.

General document model data extraction

Input requirements

• For best results, provide one clear photo or high-quality scan per document.
• Supported file formats: JPEG, PNG, BMP, TIFF, and PDF (text-embedded or scanned). Text-embedded PDFs are best to eliminate the possibility of error in character extraction and location.
• For PDF and TIFF, up to 2000 pages can be processed (with a free tier subscription, only the first two pages are processed).
• The file size must be less than 50 MB.
• Image dimensions must be between 50 x 50 pixels and 10000 x 10000 pixels.
• PDF dimensions are up to 17 x 17 inches, corresponding to Legal or A3 paper size, or smaller.
• The total size of the training data is 500 pages or less.
• If your PDFs are password-locked, you must remove the lock before submission.
• For unsupervised learning (without labeled data):
  • Data must contain keys and values.
  • Keys must appear above or to the left of the values; they can't appear below or to the right.

Supported languages and locales

Named entity recognition (NER) categories

Considerations

• Extracting entities can be useful in scenarios where you want to validate extracted values. The entities are extracted on the entire contents of the documents and not just the extracted values.

• Keys are spans of text extracted from the document, for semi structured documents, keys may need to be mapped to an existing dictionary of keys.

• Expect to see key value pairs with a key, but no value. For example if a user chose to not provide an email address on the form.

Next steps

• Follow our Form Recognizer v3.0 migration guide to learn how to use the preview version in your applications and workflows.

• Explore our REST API (preview) to learn more about the preview version and new capabilities.