• Irina Kavchik

Ocerra essentials: invoice annotation tips and tricks



Ocerra has its own data extraction engine that reads and extracts data from various invoice layouts. This engine is powered by Artificial intelligence (AI) and Machine Learning (ML) technologies. It is a fascinating tool as you can also train it, so it can learn and get smarter!


While we do train the system internally on a regular basis and add new invoice layouts and variations, we also offer a user-friendly training mechanism that allows anyone to train the system with a few clicks when required.


What is invoice annotation?


Invoice annotation is the way to train the AI of the system to extract data from the invoice layout where the system did not extract it the first time.


Support: Overview of the annotation process.


Learning to annotate is an essential skill when using an AI-powered system like Ocerra. It will help you to add information to the invoice quickly and make the system remember it next time. This will help save more time compared to manually adjusting these invoice documents.


There are a few types of annotation techniques, as well as tips and tricks, that we would like to cover in this article. As a bonus, you will find new video tutorials with invoice training examples.


Value-based annotation to pick up key data

Value-based annotation is used to pick up values such as invoice number, dates, subtotal, tax, and total amounts.


Learn how to quickly add these key data with a few clicks:



Check this support article for more information on Value-based annotation.


Tip: This technique can be used to quickly add any figures from the invoice layout. For example, invoice line-item amounts or quantities.


Table-based annotation to pick invoice table

The invoice table includes line-item description, quantities and amounts on each line.


Learn how to quickly train the system to identify the table on the invoice layout:



Tip: By default, Ocerra will aim to extract all the lines from the invoice. If you don’t need all the lines, you can train it to extract only one line from a specific invoice layout.


Check this support article for more information on Table-based annotation.


Column-based annotation to pick invoice table

Column-based annotation is applied where invoice columns were not identified in the first place.


Learn how to quickly train the system to pick up the correct column from the invoice layout:



Tip: If more than one column needs training, you can do it all in one go before annotating. Same with other information on the invoice. Add it first, and then click “annotate” once.


Check this support article for more information on Column-base annotation.


Other tips on training and annotation


Sending original PDF documents to Ocerra

  • Original PDF documents have higher data extraction rates and training. Scanned invoices on the other hand may not provide you with the same results. It is recommended to send original PDF files to Ocerra to achieve consistent invoice training results.

Training to pick up Purchase Order number

  • The purchase order number can be trained using the value-based annotation. For the annotation to work, the PO needs to have an exact match with the PO number from your system. In all other cases, you can quickly add using the reverse search.

Training to pick up Other cost such as freight

  • Freight, discount or other amount that is outside of the invoice table is trained to go to the Other field in Ocerra.

  • You can also untrain it if you would like to manually create a new line for Other cost. Learn more about other cost here.

Annotation help & support

  • As part of the setup and onboarding services, our team will pre-train invoices from your key suppliers. If you have any questions on invoice training, please send the URL link of the invoice to our support, and they’ll gladly help you with it.


Read next: 10 tips on getting started with Ocerra.

47 views0 comments