This browser is no longer supported.
Upgrade to Microsoft Edge to take advantage of the latest features, security updates, and technical support.
![](data:image/svg+xml, %3Csvg xmlns='http://www.w3.org/2000/svg' height='64' class='font-weight-bold' style='font: 600 30.11764705882353px "SegoeUI", Arial' width='64'%3E%3Ccircle fill='hsl(64, 48%, 16%)' cx='32' cy='32' r='32' /%3E%3Ctext x='50%25' y='55%25' dominant-baseline='middle' text-anchor='middle' fill='%23FFF' %3EKH%3C/text%3E%3C/svg%3E)
Hi,
In our organisation, we receive thousands of documents (structured and not structured).
We only need to classify documents and no need to extract data.
We thinking to use Form Recogniser to only identify the document using custom method with model training.
Is Form recogniser a good tool for our need ? (only recognise a document in order to classify it)
Thank you
![](data:image/svg+xml, %3Csvg xmlns='http://www.w3.org/2000/svg' height='64' class='font-weight-bold' style='font: 600 30.11764705882353px "SegoeUI", Arial' width='64'%3E%3Ccircle fill='hsl(214.4, 45%, 30%)' cx='32' cy='32' r='32' /%3E%3Ctext x='50%25' y='55%25' dominant-baseline='middle' text-anchor='middle' fill='%23FFF' %3EGM%3C/text%3E%3C/svg%3E)
Hi, do you mean classify the type of document or specific contents in the document?
i mean classify the type of document. when i receive a document , i pass it to form recogniser and if the DocTypeConfidence is high then the type of the document is the same of the documents in the model.
Hi, currently, Form Recognizer does not support this feature. However, you can explore other services such as Computer Vision, Custom Vision, Text Analytics API to implement a solution, or build your own classifier using Azure ML. Here's a technical document on classifying documents based on topics. Here are other examples using Custom Vision and Azure ML. Hope this helps!
I do some tests, but i dont know if it is a good way, here is what i done :
1) I made a model with 5 instances of the same form. Let say the form Type name = ''Medical_diagnostic_form_1936''
2) I labelled my forms with fixed elements like Form title, Logo, Section titles etc... I don't used value key pair like for example : (Employee Name : John Mcenroe) since i'm not interested in data extraction.
3) My goal is to identify document type vs the model, so i test a document and if the DocTypeConfidence is between 90 ans 100 i consider that the document type match the Model and identified as ''Medical_diagnostic_form_1936''
That's all i need from form Recogniser, the next step, is to put the document in the right place (Directory or system), and for that, anothr system will do that.
Is what i describe is a good way (Receipee) to use form recogniser ?
Note: in my tests i have a score of 20% for handwritten documents, and between 90 and 97 for forms filled in electronic way.
here is a screenshot that show the tags i choosed in my model :
Hi, thanks for following up. I just confirmed that using DocTypeConfidence to identify incoming document is a valid way of using Form Recognizer. For labeling, we recommend labeling parts of the form that are changing values as it might affect DocTypeConfidence.