Unanswered

AI Builder Scan showing invalid characters

(0) Share

Report

Posted on by Ashish - CS

When the Model is being trained, the AI Builder scan picks up additional characters which are not visible on the pdf. This happens across multiple documents. See examples below:

- Characters Y added in between extracted text

- Amber Valley Borough Council being extracted as 'Amber Valley 3or0ugh Council'

- Table extract shows 0's being added.

Categories:

AI Builder

All responses (4)

Answers (0)

Sort by

CedrickB Microsoft Employee on at

Like
a
(0)

Report
Copy link

Link copied!

This problem usually happen when watermarks are overlaying (Those Y)

We have just improved the tagging in case the document contains watermarks.

The change should be live en January.

If your document doesn't contain watermark can you share a screenshot of the entire file please so we can investigate?
Ashish - CS 10 on at

Like
a
(0)

Report
Copy link

Link copied!

Hi,
There are no watermarks on the document. Please find the file screenshot for investigation. Is there a way I can attach the file for you?
Let me know if you need any more details.
CedrickB Microsoft Employee on at

Like
a
(0)

Report
Copy link

Link copied!

I have just tested your image and it works fine on my end.

This seems something related to your PDF then.

Do you mind to send a sample to aihelpen@microsoft.com so we can try to reproduce with the original document please?
Ashish - CS 10 on at

Like
a
(0)

Report
Copy link

Link copied!

this file is from the second screenshot - the table which is getting the additional characters. I will email you over. We have quite a few such scenarios coming up on different files. I will send a few examples and issues on the files on that email.