Despite a global rush toward enterprise digital transformation, documents remains at the heart of most businesses, and unfortunately, managing them remains a distinctly manual process.
FREMONT, CA: Despite a worldwide drive toward workplace digital transformation, handling documents still involves a decidedly manual procedure, even though they remain at the core of most enterprises. It essentially forms a component of every knowledge worker's daily workflow.
Being structured, a document's flexibility makes it difficult to automate business processes. Taking data from several line-of-business applications and inserting it in a document requires cutting and pasting from screen to document, and once a document is received, it is frequently necessary to do the same.
Microsoft Syntex, which will be introduced at Ignite in October 2022, will add document processing tools to SharePoint to address some of these tedious manual problems. The system leverages machine learning to aid in document construction and parsing, transforming a manual process into one where humans direct and inspect software while still meeting contractual, legal, and regulatory standards. Learn more about content AI and some of the current use cases for this version in this in-depth analysis of Syntex.
The concept of content AI is exciting since it focuses on developing methods for interacting with the numerous sorts of papers they produce and the frequently unstructured information they include. To offer data in organised formats, systems like Syntex must deduce the context of the data, process it, extract it, and summarise it. It must also be able to work the other way around, employing documents to generate templates that can automate the development of papers.
A group of machine learning models created for content understanding form the basis of this procedure. These pre-built models are similar to those used by Azure's Cognitive Services and were created to interact with SharePoint document libraries. Custom models that might be trained on your content are also available.
Custom Models in Microsoft Syntex
The custom models are the most intriguing since they define how Syntex interacts with unstructured content to support document production and processing. When training a custom model, significant content pieces are identified by labelling existing documents as part of the training procedure. The secret to working with unstructured material is to highlight the main phrases and recurring themes that characterise a document's crucial components.
While formal letters and contracts are frequently the unstructured documents they work with, despite their structural differences, they frequently share certain terms and structures with other papers that have particular commercial implications.
Document Understanding As a Service
Discovered Microsoft's document interpretation service when researching the company's Cognitive Services AI platform. This was the unstructured document processing model used by Syntex before it. A variety of documents in a SharePoint content centre are required to train a model for document comprehension to function, resulting in the creation of two different sorts of tools: classifiers and extractors.
Documents that have been loaded into a library are identified using classifiers. One could, for instance, develop one that would locate all the Request for Proposal documents stored in a library. Then, extractors locate important information in the file, giving it to outside programmes. The name of the client who submitted the RFP can be used by extractors to add the client's contact information as a sales prospect in the CRM system.
Microsoft Syntex Use Cases
Microsoft Syntex in several ways to arrange, categorise, and comprehend documents. Here are a few of the most typical Syntex commercial use cases.
A new SharePoint content type that will be connected to all detected documents is created when a classifier is created. Starting with a selection of pertinent documents that can be utilised to develop the model, training is pretty simple. They must have at least five sample documents that are positive and one document that is negative before they can complete the training procedure.
After noting the essential phrases and terms utilised, one must then explain why those documents should be classified as a particular kind. The system will then include this into the model and verify that it accurately matches all of the example papers. If it doesn't work, just add more information and try again.
The learning curve is not too steep. Once finished, it can use more documents to assess the model's performance as a document classifier. Expect it to take a few iterations of this loop, like most machine learning systems, to develop a model that is effective for your documents.
Creating an Extractor
Similar steps are involved in creating an extractor, including labelling the information wish to take out of files. Here, they must use a different model for each distinct piece of information they wish to highlight on pages.
If the model requires more complicated rules or wants to eliminate duplicates, can improve the extractors. Extractors that are now in use additional data to SharePoint document library columns so that it can be provided to other programmes via SharePoint APIs.
Applying Content Templates
Microsoft offers tools that make it easier to create explanations for both classifiers and extractors by using templates that already cover a wide range of frequently used content formats for documents. For instance, can choose the suitable date template and include it in your explanation rather than including all the various data variations that appear in documents.
A lengthy variety of templates is available, including techniques to extract international phone numbers and email addresses as well as significant financial data types. Users can configure it to determine if email addresses belong to a sender or a receiver.
Generating documents using Syntex and Power Platform
Syntex's document models can be used to create documents as well, populating what Microsoft refers to as a contemporary template with information from well-known data sources. By uploading existing Word documents to SharePoint and adding fields to the default document structure using Syntex's Template Studio, it is possible to create templates from those documents. You may then automate document generation by linking these fields to a SharePoint list or library.
This is more of a contemporary low-code mail merge replacement than a flexible document generation tool. A Power Automate flow that pulls data from several corporate systems and uses it to create common documents like contracts, invoices, and invitations can produce a modern template managed by Syntex.
Following the creation and publication of a template, Syntex offers a framework for adding new values from a SharePoint list and, when possible, automatically filling in some fields.
As an alternative can automate the procedure using a Power Automate action. For instance, you might have Dynamics 365 automatically produce an employee handbook whenever a new sale is logged or Azure Active Directory adds a new name.
The future of Syntex and SharePoint
Although the current prototype of Syntex is still somewhat constrained in comparison to other document processing services powered by machine learning, it is evident that Microsoft has huge intentions for automating documents.
But it makes a lot of sense to build on top of SharePoint. Making it the location of Syntex's document generation and processing hub fits well with its current purpose and establishes an intriguing route for SharePoint's future role as an important enterprise content management system.
Read Also
