SharePoint Syntex

You can use SharePoint Syntex to create and train models. These models can be used to automate the classification of files and the extraction of metadata from the files. The metadata can then help your organization drive business processes, compliance, and information search and retrieval.

A content center is a new site template for creating, training, and managing SharePoint Syntex models. The first content center (the default content center) is created as part of the SharePoint Syntex setup. Additional content centers can be created by an admin in the admin center. Models can be organized by groups (such as departments). All models are rolled up to the default content center, where you can view all content centers and models as well as usage analytics for them.

A model is an algorithm “trained” using content and human input to replicate a decision an expert would make with that same information. In the case of SharePoint Syntex, this lets you classify a file of a particular business type and extract specific entity information from it. SharePoint Syntex models are trained to automate these decisions and actions to improve the accuracy and efficiency of your business processes.

Anyone who uses or consumes SharePoint Syntex capabilities requires a license. This includes the following scenarios:

• Accessing a content center.
• Training and managing models in a content center.
• Uploading content to a library where a document understanding model is associated (whether in a content center or elsewhere).
• Viewing a library where a SharePoint Syntex model has been applied.
• Manually executing a document understanding model.
• Creating a form processing model via the entry point in a SharePoint library.
• Uploading content to a library where a form processing model has been applied.

You can create a limited preview content center without a SharePoint Syntex license to evaluate the product features. However, to use the models you create, you will need to purchase SharePoint Syntex licensing.

SharePoint Syntex models use content types as the classification object and for defining the schema of entities. When creating a new model, you can create a new content type or map the model to an existing one.

Both model types are generally used for the same purpose, which is to automate the classification of content and extraction of information from them. However, there are differences – and reasons based on file structure and model reuse – that should be considered. For more information, see Difference between document understanding and form processing models. 

Document understanding models are trained in SharePoint and stored there. For more information, see Where your Microsoft 365 customer data is stored.

Form processing models trained in AI Builder are deployed in the region that hosts your environment and the datacenter where that environment is hosted. For more information, see Administer AI Builder.

Both document understanding and form processing models currently support documents in English.

Models can be applied to SharePoint document libraries.

There are two different types of models – one for structured content (form processing) and one for unstructured content (document understanding). You apply form processing models for structured content to a library because you start from a library to go into AI Builder to build your model for that library. You apply document understanding models for unstructured content, which you build in the content center, to multiple libraries.

Yes, both types of models can be applied to the same document library.

Not currently. However, this will change in the future and will be generally in line with consumption costs for other parts of SharePoint Syntex.

You can apply SharePoint Syntex models to Office documents, PDFs, and images. For a full list of file limitations and supported file types, see Document understanding overview.

Explanations are used to help to define the information you want to label and extract in your document understanding models. SharePoint Syntex provides different explanation types to support the different kinds of information you need to extract.

Classifier models are used to automate the identification and classification of a document type. Extractor models are used to identify and extract specific information from them. Extractor models are associated with a parent classifier model.

Either can be trained first, but a classifier is necessary to publish or apply the model. If you plan to train extractors, they can be trained first to be used as explanations for the classifier.

A 1 million credit add-on is included if you have 300 or more SharePoint Syntex licenses. This capacity is renewed each month if you maintain the 300-license minimum. This capacity will be allocated to the default Power Apps environment, but you can change the environment where the AI Builder models are saved and run.

No, but the 1 million capacity add-on is renewed each month if you maintain the 300-license minimum.

Form processing models trained in AI Builder are deployed in the region that hosts your environment and the datacenter where that environment is hosted. For more information, see Administer AI Builder.

JPG, PNG, or PDF format (text or scanned) are supported. Text-embedded PDFs are better because there will not be any errors in character extraction and location.

Trained form processing models are applied to document libraries. Unlike document understanding models, the form processing model is applied directly to the library from where the model was created.

Not currently.

With the SharePoint Syntex add-on, the following managed metadata capabilities are available:

• SKOS-based term set import.
• Push enterprise content types to a hub site, which also adds them to the associated sites and any newly created lists or libraries.
• Term store reports, providing insights into published term sets and their use across your tenant.

Yes, when you create an extractor in your document understanding model, you can take advantage of global term sets in the term store to display preferred terms for data that you extract. For more information, see Leverage term store taxonomy when creating an extractor.