OCR, or optical character recognition, has been a technology that was only available to high end tech companies for a long period of time. In a world that is rapidly changing and becoming more digital every day, the use cases for OCR have increased significantly. Organizations in banking, insurance, invoice processing, expense management, identity validation and contract management are starting to integrate OCR in their customer facing solutions. The development of your OCR engine, however, is very time consuming and expensive at a poor quality level. At Klippa we started about five years ago. Challenging at first, but with the years and many team members involved, we managed to build high quality OCR engine based on millions of documents and multiple machine and deep learning algorithms. We’re very proud to this piece of art, but the good news is: we’re willing to share! Our OCR is now available as a service via our OCR API. In this blog we will answer some general questions and show you some examples.
What types of documents can the Klippa OCR process?
Receipts, invoices, tickets, contracts, purchase orders, passports, ID’s, driving licenses, you name it and we have processed it for companies all over the world. Our OCR engine is very flexible and can be fine-tuned for specific cases, languages and even file types.
What languages are supported by the OCR engine?
Language is not a dependent factor in our OCR engine. In theory this means that it works for any language. Out of the box it works perfectly for English, Dutch, German, Spanish, Italian and French. In co-creation we can optimize the engine for other languages. Just give us a call and we can discuss our approach on co-creation.
What type of files can be processed?
The input of the engine is flexible. It should be as easy as possible for others to use the OCR. Therefore we support anything from pictures and scans (e.g. JPG, PNG) to documents (e.g. PDF, Docx). Do you have something else you want to have processed? Just schedule a call with us to see how we can make it work. There is a solution for everything!
What does the output look like?
By default, the API response is JSON. This is a structured data format that is lightweight and easy to read. In the image below you can see an example of what the response could look like. JSON is easy to convert to XML, CSV or any other structure. If you have custom requirements regarding the output, please shoot us a message and we can discuss.
How long does it take to implement?
The Klippa OCR as a service works via an easy to use RESTful API. On average, a single developer can implement our API in less than 4 hours. From there you can start testing in your sandbox environment. If everything runs as smooth as it should you can release the production. The average lead time from testing to releasing is less than a week.
Is it possible to test the OCR API?
The short answer is yes you can! On request we can provide an API-key with a certain amount of testing credits (usually 50 credits). This, combined with our API-documentation, is all you need to get started. If you encounter any problems, need more testing credits or just want some advice, you can talk to one of our customer success managers.
Some examples of what the Klippa OCR engine can do
Enough talking, show me the money! Action speak louder than words, so let’s get going. In the first example we will assume that you want to process invoices with OCR for accounting purposes. This reduces human processing time, can reduce the error rate and will increase the user experience. Our engine can extract any field you want, but to keep the example clear we will be focussing on extracting the date, amounts, VAT values and invoice number. On the image below you will see an input document on the left, and (part) of the JSON response on the right.
Invoice input Invoice JSON output
Perhaps you are not into invoice processing, but are looking for a solution to extract data from identity documents like passports, ID’s and similar documents? Klippa is here to help you out! In the example below you will see a passport as input on the left, and the JSON response on the right.
Passport input Passport JSON output
We believe in the power of three and will give you one more common example. This use case is reading receipt line-items for loyalty and data purposes. Interested in learning what consumers buy for what price at what stores? Our OCR as a service is here to help you out. On the left you will see a receipt and on the right you will see the JSON output.
Input receipt Output OCR line-items
Let’s get started!
In this Q&A we gave a short introduction on what our engine can do. We believe that we should always find the best solution for a use case and client with our SaaS OCR service. Therefore we like personal contact, co-developing and brainstorming with our clients. So, let’s get in contact and see where it goes! Shoot us a message at [email protected] or plan a demo with an OCR specialist.