Advanced Document Processing using Machine Learning

Hi Team,

If Advanced Document Processing is implemented using ML concept. It would be great.

For example, I am receiving an invoice in pdf format and in an another email the pdf field positions are changed and what if we make the robot to learn this using ML concept to indicate the changed field so the bot would automatically fetch the data next time it come across that document even if the format is changed.

Thanks,
Ula

UiPath can integrate with such applications via APIs so that ML can easily form part of your workflow

Thank you. I will try to know about that but i suggested is something different.

Hi @Boopathi

I believe these activities would suit best for it for now:
https://activities.uipath.com/docs/about-the-intelligentocr-activities-pack
It uses templates for each type of document that is being processed.

If I understood you correctly, you would wish for those templates to auto-correct/auto-adapt themselves with Machine Learning algorithms? :slight_smile:

2 Likes

Thank you for this link and I am going through it. Yes, if we capture an element and the bot learns the element always takes that position in a pdf or any other document using Machine learning algorithms (If integrated with the RPA Platform)

Hi,

Thank you so much for the details and it works as expected.
But when i am trying to fetch details from invoices with different languages like Finnish, Swedish etc. its failing and not extracting all the details as expected.

Do we have any language barrier for the activities ?
I am using Machine learning extractor and invoice ML Model to fetch the details.

Thanks in advance

1 Like

Hi @ronak_94

I believe support for more languages is coming, but @Ioana_Gligan can give more details here.

1 Like

Hello @ronak_94,

The Machine Learning Extractor with the Invoices endpoint works with a pre-trained model that only targets a few languages for now.

To improve this, we are working on two directions:

  • training the machine learning model on more languages (takes time as we need to collect sufficient samples for each language, and tag them appropriately). @alexcabuz can offer more info on the supported languages at this moment.
  • offering training/retraining capabilities, so that you will be able to tag and train your own model with your own specific data.

All this is work in progress, rest assured we will let you know when it’s available. :slight_smile:

3 Likes

thank you so much for the update :slight_smile:

2 Likes

is this document intelligent processing has separate license or part of studio?

Hi @vboddu

Correct, see here: