2. Getting set up

This codelab assumes you have completed the Document AI Setup steps listed in theIntroductory Codelab.

Please complete the following steps before proceeding:

• Start Cloud Shell
• Enable the Document AI API

3. Create a Processor

You must first create an Invoice Parser processor to use for this lab.

1. In the console, navigate to theDocument AI Overview page.

2. ClickCreate Processor, scroll down toSpecialized (or type"Invoice Parser" in the search bar) and selectInvoice Parser.

3. Give it the namecodelab-invoice-uptraining (Or something else you'll remember) and select the closest region on the list.

4. ClickCreate to create your processor. You should then see the Processor Overview page.

4. Create a Dataset

In order to train our processor, we will have to create a dataset with training and testing data to help the processor identify the entities we want to extract.

You will need to create a new bucket inCloud Storage to store the dataset. Note: This should not be the same bucket where your documents are currently stored.

1. OpenCloud Shell and run the following commands to create a bucket. Alternatively,create a new bucket in the Cloud Console. Save this bucket name, you will need it later.

exportPROJECT_ID=$(gcloudconfigget-valueproject)gsutilmb-p$PROJECT_ID"gs://${PROJECT_ID}-uptraining-codelab"

2. Go to theDataset tab, and click onCreate Dataset

3. Paste the bucket name from the bucket you created in step one into theDestination Path field. (Don't includegs://)

4. Wait for the dataset to be created, then it should direct you to the Dataset management page.

5. Import a Test Document

Now, let's import a sample invoice pdf into our dataset.

1. Click onImport Documents

2. We have a sample PDF for you to use in this lab. Copy and paste the following link into theSource Path box. Leave the "Data split" as "Unassigned" for now. ClickImport.

cloud-samples-data/documentai/codelabs/uptraining/pdfs

3. Wait for the document to import. This took less than 1 minute in my tests.

4. When the import completes, you should see the document in the Dataset management UI. Click on it to enter the labeling console.

6. Label the Test Document

Next, we will identify text elements and labels for the entities we would like to extract. These labels will be used to train our model to parse this specific document structure and identify the correct types.

1. You should now be in the labeling console, which will look something like this.

2. Click on the "Select Text" Tool, then highlight the text "McWilliam Piping International Piping Company" and assign the labelsupplier_name. You can use the text filter to search for label names.

3. Highlight the text "14368 Pipeline Ave Chino, CA 91710" and assign the labelsupplier_address.

4. Highlight the text "10001" and assign the labelinvoice_id.

5. Highlight the text "2020-01-02" and assign the labeldue_date.

6. Switch to the "Bounding Box" tool. Highlight the text "Knuckle Couplers" and assign the labelline_item/description.

7. Highlight the text "9" and assign the labelline_item/quantity.

8. Highlight the text "74.43" and assign the labelline_item/unit_price.

9. Highlight the text "669.87" and assign the labelline_item/amount.

10. Repeat the previous 4 steps for the next two line items. It should look like this when complete.

11. Highlight the text "1,419.57" (next to Subtotal) and assign the labelnet_amount.

12. Highlight the text "113.57" (next to Tax) and assign the labeltotal_tax_amount.

13. Highlight the text "1,533.14" (next to Total) and assign the labeltotal_amount.

14. Highlight one of the "$" characters and assign the labelcurrency.

15. The labeled document should look like this when complete. Note, you can make adjustments to these labels by clicking on the bounding box in the document or the label name/value on the left side menu. ClickSave when you are finished labeling.

16. Here is the full list of labels and values

7. Assign Document to Training Set

You should now be back at the Dataset management console. Notice that the number of Labeled and Unlabeled documents have changed as well as the numbers of active labels.

1. We need to assign this document to either the "Training" or "Test" set. Click on the Document.

2. ClickAssign to Set, then click onTraining.

3. Notice the Data Split numbers have changed.

8. Import Pre-Labeled Data

Document AI Uptraining requires a minimum of 10 documents in both the training and test sets, along with 10 instances of each label in each set.

It's recommended to have at least 50 documents in each set with 50 instances of each label for best performance. More training data generally equates to higher accuracy.

It will take a long time to manually label 100 documents, so we have some pre-labeled documents that you can import for this lab.

You can import pre-labeled document files in theDocument.json format. These can be results from calling a processor and verifying the accuracy usingHuman in the Loop (HITL).

cloud-samples-data/documentai/codelabs/uptraining/training

cloud-samples-data/documentai/codelabs/uptraining/test

11. Uptrain the Model

Now, we are ready to begin training our Invoice Parser.

1. ClickTrain New Version

2. Give your version a name that you'll remember, such ascodelab-uptraining-test-1. The Base version is the model version this new version will be built from. If you're using a new processor, the only option should beGoogle Pretrained Next with Uptraining

3. (Optional) You can also selectView Label Stats to see metrics about the labels in your dataset.

4. Click onStart Training to begin the Uptraining process. You should be redirected to the Dataset management page. You can view the training status on the right side. Training will take a few hours to complete. You can leave this page and return later.

5. If you click on the version name, you will be directed to theManage Versions page, which shows the Version ID and the current status of the Training Job.

12. Test the New Model Version

Once the Training Job is complete (it took about 1 hour in my tests), you can now test out the new model version and start using it for predictions.

1. Go to theManage Versions page. Here you can see the current status and F1 Score.

2. We will need to deploy this model version before it can be used. Click on the vertical dots on the right side and selectDeploy Version.

3. SelectDeploy from the pop-up window, when wait for the version to deploy. This will take a few minutes to complete. After it's deployed, you can also set this version as the Default Version.

4. Once it's finished deploying, go to theEvaluate Tab. Then click on the Version dropdown and select our newly-created version.

5. On this page, you can view evaluation metrics including the F1 score, Precision and Recall for the full document as well as individual labels. You can read more about these metrics in theAutoML Documentation.
6. Download the PDF File linked below. This is a sample document that was not included in the Training or Test set.

file_downloadDownload PDF

7. Click onUpload Test Document and select the PDF file.

8. The extracted entities should look something like this.

13. Conclusion

Congratulations, you've successfully used Document AI to uptrain an Invoice Parser. You can now use this processor to parse invoices just as you would for any Specialized Processor.

You can refer to theSpecialized Processors Codelab to review how to handle the processing response.

Cleanup

To avoid incurring charges to your Google Cloud account for the resources used in this tutorial:

• In the Cloud Console, go to theManage resources page.
• In the project list, select your project then click Delete.
• In the dialog, type the project ID and then click Shut down to delete the project.

Resources

• Document AI Workbench Documentation
• The Future of Documents - YouTube Playlist
• Document AI Documentation
• Document AI Python Client Library
• Document AI Samples

License

This work is licensed under a Creative Commons Attribution 2.0 Generic License.