Hi,
We have a Editable PDF which has a static table with fixed width, rows and columns. We only need to capture the material description. The model works well in cases when the material description is short and fits in a single line.
When the description is long it wraps in the same row in two lines. The AI builder recognizes this as two rows instead of one.
I have a ticket open with Microsoft (TrackingID#2109290060003230)and it states that this is a known bug and is being worked upon.
Anyone else face the issue?
Hi Joe,
I have replied back to you via private message about how to recreate the issue and also shared a private preview link which helped fix the issue
Hi again!
I've tested the documents, and the table extraction results seem correct. 🙂 I've sent you a private message to share how I've done the tagging, and keep investigating as we discuss the specifics of your documents.
Thank you for your prompt response. Looking forward to hearing from you
Quick update: I got the details from the support ticket and will run some tests and let you know. Thanks for the patience!
Thanks! Let me get the details on the support ticket and I'll report back.
@JoeF-MSFT i have used 20 PDFs to train the model and the same issue was also reproduced by Microsoft Support
Hi @rahullakshmanan,
Thanks for your question.
How many documents did you use for training? The more samples you can provide for training where the 'material description' is long, the better the model will get to learn how to recognize rows.