I trained a document extraction - how to continue?

(1) Share

Report

Posted on by CU04081300-0

Hey guys

A little description to what I actually want to achieve:

So at another location ( bsides the office I am sitting in ), we got a weighbridge with our recycling yard. We recycle materials from demolished houses and streets and stuff like that to create a recycling gravel out of it.

Now each weighing process creates a pdf, which includes the basic stuff like construction site, adresses, timestamps, carrier, weights and so on.

What we office people would like to have, is a self refreshing diagram, every time a weighing process was completed; or if you look at p.automate it would be:

Everytime a file is created

The first problem obviously is, to scrape/extract the data I want. In my case the following:
Delivery type
Material number

Material description/text

Net weight

With those definitions I want the diagram to show:

Currently concrete taken in: xxx tons
Currently asphalt taken in: xxx tons

Currently recycling gravel taken out: xxx tons

and so on.

I tried to extract those data fields with power query, but I dont get really too far, and even the nice people who are helping me there, dont seem to have an answer to the errors I create ^^
I also tried severeal scraping apps like parseur and parsio and so on, and I really love the simplicity how you determine a specific field and it gets read file for file. However, the prices there are way over our limit.

Right now I got the trial running and just trained the ai builder, which came up with a possible detection rate of 99% ^^ ( since our weighing pdf´s are very very similiar to each other obviously )

But what now? Do I go back to automate and try to arrange a flow?

How to tell the ai modul, use those extracted and either

a) create a diagram with it ( if there is an option/app for that in automate )

b) send the extracted values to an excel sheet/table somewhere

If possible, id avoid going into excel if theres already something that could read/use the extracted data and creats me a nice and shiny diagram.

Maybe you got a better way for me to go instead of my thoughts here?

Anways, any help is appreciated

Kind regards

Tobias

I have the same question (0)

All responses (3)

Answers (0)

CU04081300-0 4 on at

Like (0)

Report

I trained a document extraction - how to continue?

Update: So I used this ready to use template

and found the right dynamics to get my values.

So right now I think this is a good approach, since the weighing process triggers this event and I get a eMail with just the facts I need.

When I now use a new flow based on that;

whenever new email

- Filter by sender

- Filter by subject

then use the content to add values into a table.

What would I have to do now, now extract the emails content ( the already needed values ) and have them placed on the right spot ( based on the values itself, like the first value determines the material type and for this the location in the table ) ?

Was this reply helpful? Yes No
CU04081300-0 4 on at

Like (0)

Report

I trained a document extraction - how to continue?

Hey @Michael E. Gernaey

Thanks for the support.

Ill try to define it a bit more.

Again my planned steps:

1- Everytime a file is created, it should create a copy to my folder

2- Again, everytime a files is created, extrat data with the trained model

3- Send this data to whatever app that can use it to form a diagram

4- Let this whole procedure run permanently so it refreshes itself after every weighing pdf

How the diagram is supposed to look like is something I am also unsure about to be honest.

It can be whatever is useful on first sight.

Like mentioned, id like to make a monthly overview, maybe as a bar chart, or graphs?

I just took an example from Excel, which looks not too bad, youd have to determine everything I know, but just as an example

With this Id have every current running month and the months before with a little graph that shows trend/compared to earlier months.

The problem I have is, how to go on with my processed data. To you it might not seem hard, but to me it is. I am really struggling to understand the logics in automate, escpecially why its so hard to link to onedrive folders..... ^^

How I do call the extracted data, how to get a "grip" on it? What app can be used to create something visual out of the extracted data?

Could you please tell me that?

Was this reply helpful? Yes No
Suggested answer

Michael E. Gernaey 47,125 Super User 2025 Season 2 on at

Like (0)

Report

I trained a document extraction - how to continue?

hi @CU04081300-0

I am unfortunately lost here 1) as this needs to be moved to Power Automate not power apps.

You say diagram but that doesn't mean a whole lot as you haven't shared what you want it to look like.

But you would definitely train the model, then leverage the model in the AI Builder actions to gather what you want. Not super hard but after that I have no idea how to build a Diagram for you as mentioned, no idea what its supposed to look like.

But yeah

1) when a file is created

2) use AI builder action and specific trained model to extract the data

3) the data will become Dynamic Properties and or Accessible via the primary Dynamic Property of your AI Extraction Action

4) then you can write it to wherever, but it more sounds like you want something "updated" not simply creating a diagram per run, but updating an existing document

Is that correct?

P.S. Also moving this to the right forum.

Was this reply helpful? Yes No