Event banner
Event details
97 Comments
- EmilyPerina
Community Manager
That’s a wrap! Thank you for joining our Microsoft Syntex AMA today! We appreciate all the great questions and hope you learned something new! I'll be locking this event to new questions, but our experts will still be able to respond to any unanswered questions. Stay tuned for more Microsoft Syntex news!
For any other questions about Syntex, head to the Microsoft Syntex Community.
- IanStory
Microsoft
Thank you all for the great questions and engagement! As Emily noted, please share other questions about Syntex at the community, but I'll note that if you have specific product feedback you'd like to share, like a feature request, we welcome those here - https://feedbackportal.microsoft.com/feedback/forum/59b83b65-b78a-ec11-93b0-0022482db33b - Thank you all for your time and answers. Appreciate it!
- Muchas GRACIAS a todos, Saludos Cordiales
- As well as the improvement of ANNOTATIONS, in the future could an audio or video be attached to the document? Do you see it as a topic that would contribute to our clients?
- IanStoryHi Jaime - yes, we plan to add audio and video annotation features in the future, we are aware of some competitors that have this, and think there are great use cases for it (like attaching a video from a meeting when you discussed a document as an annotation, or attaching a voice recording from a doctor reading an xray to the xray itself). This isn't near term, but we're aware and thinking about it 🙂
- Perfect !!!
- Where are the annotations stored? Is there a limit how much annotations can be created per document?
Kristen_KamathAnnotations are stored with comments. There are practical limitations to performance, so we don't recommend creating tens of thousands of annotations.- Great, thanks for your answer.
- What do you think about generating synthetic data? Would it be helpful for the models?
- Shreya_GangulyHi Jaime - Taking a stab at this question from one lens. You can leverage content assembly to create an Invoice or Contract modern template and generate lot of documents with synthetic data.
- IanStoryHi Jaime, that said, for best model performance, you'd probably want more variation in the actual formats/layouts of documents than synthetic data would imply/create. You could use content assembly, as Shreya noted, to create synthetic data, and we've definitely seen customers using content assembly to make documents not just for training, but for demos and of course production use cases with Syntex. Just note that you'll want more variation for the best model perf (unless all you end up processing are documents created with content assembly and the templates you've built).
- Are there plans to use versioning for annotations like for documents?
- Kristen_KamathAnnotations will be respective to the version of the document and you will be able to see them with those versions. However, we dont version the annotations specifically.
- That's a pity, thanks for your answer.
- Have you considered using another solution similar to AI BUILDER? for example Form Recognizer or another external to Microsoft ?
- IanStoryHi Jaime - Syntex today uses AI Builder - it is what powers our freeform and structured document processing capabilities. Underneath AI Builder, it uses Form Recognizer. We do not plan to use something external to Microsoft directly in the product, that creates all kinds of licensing and OEM and compliance complications for us - however, this is something our wonderful business partners can of course do using the various Syntex and SharePoint and Microsoft 365 APIs all up.
- Could you guide me how to migrate a syntex model from a non prod environment to the prod environment.
- JamesEcclesCurrently we have options to do this for unstructured document processing models only, either through PnP PowerShell or the PnP SDK. For example, the PowerShell approach is documented here- https://learn.microsoft.com/en-gb/microsoft-365/syntex/powershell-syntex-import-export?view=o365-worldwide
- As I am sure you are aware, both in Latam and in other regions, there are a variety of types and formats of documents, even with a high percentage in manuscript, what is the strategy to deal with or progress on the subject?
- IanStoryHi Jaime - can you elaborate on what you mean by manuscript? Are you talking handwriting? Printed documents that need to be scanned? For handwriting, you can use Syntex today to process handwritten documents, in particular I'd look at freeform and structured document processing. If printed documents that need to be scanned, we don't plan to build scanner software as there are many, many software packages that work with scanners today to take paper and turn it into a PDF or TIF, which can then be processed by Syntex (in fact, every scanner made for the last few years includes various software packages to operate the scanner itself).
- exact, handwritten With respect to digitization solutions, it is correct, there are several, and we have already covered several functionalities...in my opinion, it would only be necessary to add certain functionalities in the pre-processing of the image to ensure that the best possible image reaches Syntex and facilitate the removal
- JamesEcclesWe continue to expand the portfolio of models in Syntex to handle different types of file. For example unstructured and freeform models can be used for more varied datasets. In addition, we will be releasing the Taxonomy Tagger feature which will do keyword extraction from documents based on term sets, with no training required by users.
- We'll Syntex be able to auto tag (extract metadata) from documents that don't follow a prebuild document model?
- Shreya_Ganguly
Yes Zuly, that will be possible. You can leverage the soon-to-be-release Taxonomy Tagger feature to auto tag your documents with your organization-specific taxonomy without having to build/train or use any existing model.
You can also explore freeform and unstructured document processing to extract metadata from unstructured content.
- Currently the annotation feature is included in the Syntex per user license. Are there changes planned to switch to a different licensing/pricing model, e.g. like a pay-as-you-go model?
- Kristen_KamathAnnotation will continue to be part of the Microsoft Syntex seat license.
- Perfect, thanks for your answer.
