AI Builder To Extract Data From Document In Power Platform

Introduction

AI Builder is a component of Microsoft Power Platform which adds artificial intelligence in applications, workflow, chatbots and data analytics. This article will describe the simple steps to build train the model using AI builder which can be used to automate processes intelligently. AI builder helps to add intelligence in workflows and automate business processes using power apps and power automate. In this article we will learn how to build a model and train it for further use in power platform.

Here, I will show how to extract the data from a form and train the model accordingly. We will use form processing to extract data from the document. Form Processing identifies format or structure of the provided sample documents and extracts text from the matching form. Those extracted data can be save into our database, SharePoint list, library property and loaded in power apps form for further use. We will use invoices as sample documents to train and test the model for demonstration purposes in this article.

Steps

Power Apps Studio or Microsoft Flow website provides navigation pane to access AI Builder. Go to AI builder from Power Apps Studio or Power Automate/Flow Website.

AI Builder to Extract Data from Document in Power Platform

We can get different types of AI builders in power apps studio or Microsoft flow site, such as Category Classification, Entity Extraction, Form Processing, Object Detection, Prediction and many more. In our case, we will use form processing as we are going to extract the data from invoice document. Form Processing works like Optical Character Recognition, so we need this for our scenario. It is recommended to use standard documents( same format of documents) for training and testing the model to get proper accuracy.

Secondly, select and click on Form Processing. Give the name of your AI model and click on create. For my case, I am giving name as Demo AI Model.

Then, click on create as depicted above which automatically brings you to the below screen.

Now, it’s time to choose information or fields which we are going to extract using this model. I am going to show an example of extracting data from an invoice document. Thus, I am adding fields: Invoice Number, Invoice Date, Invoice Amount, Invoice From and Invoice For as extracting fields.

Add necessary fields and click on Next.

Subsequently, we need to add sample training documents which should have the same format of those that we are going to extract data from.

For learning and testing purposes we can get sample documents from the Get sample documents option beside Add documents as illustrated in the above picture. For training purposes, at least we have to upload a minimum of five documents.

Click on Add documents and upload documents from the appropriate location.

Upload the documents and click on Close after upload completion. Now, click on analyze, it analyzes documents and reads information from documents.

Analyzing may take a few minutes.

After analyzing we have to Tag documents, select appropriate area and tag with the correct field that we have added earlier. For instance, tag Invoice From to appropriate text area in document and do the same for all five fields: Invoice For, Invoice Date, Invoice Number and Invoice Amount.

Repeat this for all uploaded documents and click on Next after tagging all uploaded documents.

Then, click on Training. It trains the model. This may take a few seconds to train the model.

Now, you can go to details of the model by clicking on Model. You can publish and do a quick test of the document.

To Test the Model, Click on Quick Test and upload test documents. It will show the results of your test, you can hover and see the test output.

After testing your model Click on publish to publish it. Then you can use it in your flow or application.

Conclusion

This article has described simple steps to train and build models using AI Builder in Power Platform which adds intelligence to automate business processes, applications and flow. These days intelligence has become the most crucial thing in every application, processes and flow. Hence, after creating this model we can use this AI model in our power automate, flow, chatbot, business intelligence and power apps to adds intelligence in business process, flow and applications.