Affinda’s invoice extractor is an AI based tool that allows you to extract structured data from invoices, no matter what format they come in. We power AP automation, enabling you to skip the manual data entry and populate your accounting software with invoice data automatically.
Here are questions to some of the most frequently asked questions about Affinda’s invoice extraction software.
Who Uses Invoice Extraction?
Any business with an accounts payable team can benefit from using an invoice extractor. However, for accounts payable and ERP platform providers, an invoice parser is essential. By simply adding an invoice parsing integration to your software, you’re providing tremendous benefits to your customers.
How Does an Invoice Parser Recognise Different Fields?
We’ve put our invoice parser through rigorous training on millions of documents in 56 languages. Invoices tend to have the same widespread key attributes.
Fields that are commonplace on invoices include:
- Due date
- PO Number
- Bank details
- Invoice number
- Payment instructions
Our software can detect which is which based on the text in the document. For example, if the invoice says ‘Due by:’ followed by a date, the software can be pretty sure that date is the due date. But our data extraction isn’t based on a simple keyword search model. Instead, we use natural language processing and visual processing to determine exactly what information your invoice is meant to convey.
If you have specific fields you need to have parsed, we have the potential to train our model based on your invoice data. Just let us know what you’d like!
Is Affinda’s Invoice Parser Customisable?
While Affinda’s invoice extractor has been trained on invoices in dozens of different fields, we realize that every industry has its own needs and we’d like nothing better than to customize our product to meet your needs. We can optimise the parsing software by training it with your specific invoices, helping it learn how to read and categorise the information more accurately. We can also create custom fields that match your invoices.
How Can I Validate the Parsing Results?
Invoice data needs to be reliable and accurate. That’s why we provide a validation tool along with our invoice parser. Once you integrate this into your platform you’ll be able to check parsed data before pushing it through to your system.
If you like, we can set a custom auto validation threshold. Our data comes with a probability of confidence rating; for example, the technology might be 98.7% sure that a given field is a PO number. You can set manual checks for data that comes under a particular rating. Many users automatically accept data that have a confidence rating of above 96% and manually check anything under that level.
What is OCR?
Optical character recognition, more commonly known as OCR, is a technology that’s been around for years. It is used on documents that don’t have a text layer; usually images or scanned files. Since these are just an image, computer software cannot read the words on the page unless it has been processed through OCR.
OCR scans the document and locates characters on it, including letters, numbers, spaces, special characters, and full stops. OCR can turn an image into a file with a text layer that is readable by computer software, or into a simple text document.
Affinda’s invoice extractor works by parsing documents using our AI technology. This technology requires a text layer so that it can categorise and structure data, so any files you upload without a text layer are first put through our integrated OCR so that they can then be parsed.
What’s Adaptive Recognition?
Our software takes a machine learning-led approach to intelligent document processing. That includes adaptive recognition.
When we say that our software is trained on millions of documents, it means we feed real invoice files through the AI to see how accurate the parsed results are. After that, we label and annotate the data to train the engine to better recognise those types of invoices and information. Then we parse a new batch of documents, find out where the gaps are and go through the process again. Each time, we annotate the data so that the algorithm becomes more accurate with every iteration. As we correct information, the software learns over time. That’s adaptive recognition.
The result? More accurate parsing and a technology that can adapt to your specific use case.
How Does Integration Work?
To integrate our invoice parser with your accounting platform, you will need a developer. Our data extraction software works on REST API, which we have chosen because it can be used with almost any platform. Developers write our documentation, making it super easy to follow and implement. We also offer a range of client libraries to suit your needs.
Why are REST APIS So Great?
REST offers integration options with a wide variety of platforms. Our API provides a JSON output, which many programmes require.
REST APIs support messaging in various formats, including HTML, plain text, XML, and JSON. Alternatives such as SOAP only support messaging in XML, making them much more limited in application. This ability to use numerous formats for data is a top reason that REST is the API style of choice for many developers.
Where Does Our Invoice Parser Fit in the Market?
As you may have noticed, our invoice extraction tool is a lot cheaper than many of our competitors (up to 60%!). Many people put software into two camps: high quality and high cost or low quality and low cost. Affinda has decided that we fit into neither category, instead offering high-quality software at a reasonable price.
As a newer company, we are also more flexible with the ability to meet customization requests in a reasonable timeframe. We aim to make it easier for companies to automate their processes and we work hard to make this as accessible as possible.
How Can I Try Out the Affinda Invoice Parser?
Not ready to commit yet? Fair enough – that’s why we offer the free online version of our software. Head to https://affinda.com/invoice-extractor/ to check it out. We recommend uploading 100+ documents to perform a comprehensive test and see if this could be the extractor for you.
There’s no point in us trying to convince you that we are more accurate when you can just see the results for yourself.