UiPath Document Understanding - Forms AI - ML For Structured Documents - RPA
Vložit
- čas přidán 3. 08. 2024
- There are many approaches for extracting data from structured documents. Among those is template creation. However, template creation can be tricky when some parts change from time to time, such as rows of a structured table. This is where machine learning can help us easily predict and extract data from structured documents.
#UiPath Forms AI lets us easily train a machine learning model without writing any code to extract data from documents. This video is focused on using Forms AI for document processing.
▬ Contents of this video ▬▬▬▬▬▬▬▬▬▬
0:00 - Introduction
02:40 - Exploring UiPath Document Understanding App
03:50 - Creating New Project in Document Understanding App
06:35 - Configuring and Training Model in Forms AI
24:00 - Configuring UiPath Document Understanding Workflow
32:00 - Examine the Output and Discussing Additional Points
#UIPathCommunity #RPA - Věda a technologie
thanks for easy-to-follow video.
Nice one
Thank you so much!! :)
Hi, thank you for a great content. I need a link to download all the form 4 samples you used in the demo. thanks a lot😊
Hi Lahiru. Sorry for the dumb question . When you are saying the slash key for selecting the row you mean forward slash and backward slash key. Also how can we access these docs for practicing purpose
Hey.. Im not sure how I missed your comment. Sorry for the late reply..
I'm referring to this one: "/"
You can actually download Form 4 documents online for any company. It's publicly available. Do a quick Google search for SEC Form 4 for any company.
Nice one Lahiru. Can you also have something added to extract the check box data, is that straightforward?
Hey.. I missed your commen bro... Yes, we can do checkbox data too.. We can try predicting checkboxes by indicating the checkbox (if you don't have a unique label next to it) or, the label next to the checkbox during the labeling session. This way, the model learns to extract the values from check boxes that are ticked..
Keep in mind that we only need to label the value of the checkbox that is checked.. Labels of Unchecked checkboxes are not labeled.
great video bro! in form ai we should add only one type of document(form4/invoice/po) per project right bro? one more qn bro- only ml extractor should be used in form au right?
Hey bro.. Good questions as always :)
So yes, we need to add only one type of documents per Forms AI model (Document Manager session). If you need to have multiple document types in one project, you can always create multiple Document Manager sessions in that project that creates separate extractors for each.
For your next question, Yes.. we need to use only Machine Learning Extractor for this one.
@@LahiruFernando thankss bro! so in one project we can only use multiple extractors for multiple type of documents?
@@yashobantadash6670 Yes bro.. in one project, we can have multiple extractors used for that project. Just like you see in AI Center.. One project can have multiple Skills.. It's the same concept, done in a more simplified way :)
@@LahiruFernando thanks bro 😍🙌
Hi Lahiru, thanks for this awesome tutorial! I want to know that can we use custom validations inside the "Validation Station". Ex: I need to validate an invoice line item code with external system data. if it does not match, give an option to the user to select a correct item code and pass that value.
Hello Gayashan,
At the moment we cannot do custom validations within the validation station or in Taxonomy manager. But in future releases these features will slowly come into the picture for sure.
For the time being, we will have to write our logic to perform validations, and update the extraction results variable before pushing to validation station in case you want to show that data to your users..
Thank you so much for the explaination !
I have installed taxanomy manager , when i created a group and tried inputting category ,suddenly it stopped saying "communication between uipath studio and taxonomy manager ended unexpectedly".i am unable to open it again. Then I created new process but when i put category same thing is happening. Can you please help?
Hi Sindhu,
Hmm.. Seems like there is some issue with the Studio. Can you try uninstalling and installing again?
You can also try the following:
- Use stable version of the latest dependencies to see if it works
- Use a bit older version than the latest to see if that works
Let me know if this helps..
Thank you so much! I will try
@@LahiruFernando Thanks for your help !! It worked.
Hey Lahiru,
what is the difference between
UiPath - AiCenter - create project - Data Labeling - Document Understanding
and
UiPath - Document understanding - create project
can we use any one of them ?
Hi Padmashree,
AI Center Daya labeling is used for training document understanding models we have in AI Center. It's all about loading documents, doing the labeling on the fields we need to extract and training our models.
The other create project I believe is the one we have on Automation Cloud as an app. That's the new DU platform that's being developed. For now we have Forms AI available there. That is something you can create to process structures documents using the power of AI. But I believe they will improve it to make the whole AI Center thing simple for DU.
Hope this helps..
@@LahiruFernando Thanks for responding. For me, both of these are doing similar job. Loading and training documents. But what is the difference. When we can use aicenter for training documents and when to use document understanding for training documents ? Can we use any of these? Which one is better ?
@@padmashreesandeep8264 Yes... It does the same thing. For the time being, on the latest version, we can use AI Center for training semi-strucured documents like invoices purchase orders etc. The Data labeling you have in the Document Understanding App is used for Structured templates when you want to use the power of AI instead of creating templates for each. That is the biggest difference between the two.
@@LahiruFernando got it ! Thanks again for explaining 💐
Hi Lahiru, I need some help in the below use case. Below is the requirement.
Step 1. Extract invoice details using Document Understanding - AI Fabric
Step 2. Fill the extracted data from step1 in Data Services and generate a unique code.
Step 3. Place these processed invoices in a separate folder in google drive. In case of exception place the failed invoices in another folder.
I am able execute step 1 and step 2.
I have non invoice PDF's in input folder. Data is not extracted from non invoice PDF's using AI Fabric, I need to identify these non invoice PDF's and place them in failed invoices folder.
In which part of the code can i check non invoice PDF's? How to check this condition?
Hi Sindhu,
Identifying non invoice documents can be done through classification. You can define the unique keywords that you see in invoices and not in other documents. This way, the classifier will identify documents that needs to be excluded. You can write a simple logic based on classification results to handle it
@@LahiruFernando Thank you so much!! I will try it out
Hi , what’s the difference between form extractor and forms AI?
Hello Rajeena,
Below is the difference:
Form Extractor: This one is purely based on templates. Form Extractor creates an exact template of the document so we can extract values from those defined structure. This does not have any AI behind it.
Forms AI: This is also for processing structured documents. However, this one does not need any templates to extract data. Forms AI is powered by an AI model that runs behind it to identify the values we need to extract and from where. This way, slight changes in the page (due to scan orientations etc) can be overcome easily. Further, Forms AI will also support dynamic row values when you have multiple rows on a fixed form. This is not possible with Form Extractor. Form AI can do it because of AI model. However, this part is not perfect yet, but it is getting there.
Hope this helps...
Thank you so much , so if the project has different types of documents coming in, do we use that many ML extractor with the end points created ?
@@rajeenasuresh Depends on the types of documents you get. If it is all structured forms, you can use Form Extractor or Forms AI to train it. If you get semi-structured documents like invoices, you can have another ML Extractor for that. But in case you are going with Forms AI, yes, you will have ML Extractor activity with those endpoints given to you by the model you create.
Thank you 😊
Document understanding with data labelling , training or Forms AI which is the best way? Is there any extra license needed for Forms AI? Currently I am able to use it on structured documents with community license . Thanks