Question

IQ Bot technology breakdown

  • 5 October 2022
  • 9 replies
  • 2 views

Hello,

 

Is there a detailed breakdown of how IQ bot works and where machine learning is used, if any?

 

The way I see it is, you "train" IQ bot to understand were form fields are located in your documents (maybe some machine learning here), that data is sent to your selected OCR provider and the results are returned, where you can do further validation in Python. Is my understanding correct?

 

Cheers,

Dylan


9 replies

@MICAH SMITH​ Are you able to get more information on this?

Userlevel 3
Badge +7

Hi @Dylan Lesperance​ ,

 

Please find the below,

 

Automate document-centric business processes, end to end, by using IQ Bot, a web-based, Cloud-native intelligent document processing solution that can read and process complex documents and email. This solution combines RPA with AI techniques to extract and classify semi-structured and unstructured data.

 

I'd recommend to you, please check out the below link: 👇

 

https://university.automationanywhere.com/training/rpa-learning-trails/iq-bot-developer//p>

 

Phases in IQ Bot business process:

 

  1. Preprocess documents.
  2. Receive text segmentation and optical character recognition (OCR).
  3. Classify documents in groups.
  4. Extract document data.
  5. Validate and correct failed documents.
  6. Complete validation and save.
  7. Trigger approval.
  8. Obtain final review and approval

 

Where Python used in IQ BOT: 👇

https://developer.automationanywhere.com/blog/how-to-increase-stp-in-iq-bot-using-custom-logic/p>

 

How the documents are classified :

https://developer.automationanywhere.com/blog/understanding-groups-creation-classification-in-iq-bot/p>

 

In the below link, you find the complete documentation about the IQ BOT and Training as well.

 

https://docs.automationanywhere.com/bundle/enterprise-v2019/page/enterprise-cloud/topics/iq-bot/cloud-iqb-process-overview_1.html/p>

 

Please let me know if you have any questions.

 

Thanks!

 

Badge +4

Hi @Dylan Lesperance​ ,

 

There is a new document processing tool out now which is essentially an improved version of IQ Bot. Find some more information here:

 

Document Automation (automationanywhere.com)

Intelligent Document Processing solutions feature comparison matrix (automationanywhere.com)

Userlevel 4
Badge +8

Hi @Dylan Lesperance​ ,

 

OCR helps us to extract the data from document on top of it if you want to do any data transformation we will use Python. Ex: if you want to change the date format from one format to another etc.

 

ML helps the LI to learn from the corrections the users we make via Validator, over a period of time the model gets matured and next time it handles by its own.

Hey Everyone!

 

Thanks for the context. I do understand how to use IQ Bot but wanted to get a better idea of the tech in the backend as I am comparing IQ Bot to other solutions.

Hey @ChanduMohammad S​ 

 

Thanks for the context. It's my understand the only improvements made using ML is the IQ Bot's understanding of the document layout, not any improvements on the data returned in those fields. The data returned in fields are fully depended on the OCR provider selected?

Userlevel 3
Badge +6

The same mentioned above can also be done with classification/classification correction using the classification package. IQ Bot was the predecessor to document automation and doesn't work the same way as described above.

Userlevel 3
Badge +6

Hey @Dylan Lesperance​ - it may be just semantics, but Document Automation (The newest version of document extraction capabilities from Automation Anywhere) does indeed do the learning that you're talking about as far as recognizing how to find additional data based on correction from operators - which will ultimately increase straight through processing. You can see this yourself in CE by processing a few documents, correcting them in AARI by clicking and dragging to draw zones over the missed data, then processing the same/similar forms again. You'll notice that data that previously was not auto-extracted is now extracted based on the corrective feedback provided by the operator

Userlevel 4
Badge +8

Thanks for the very detailed insights @MICAH SMITH​ 

Reply