Sypht fields demystified

You’ll notice that Sypht offers lots of different data fields. And you’ll also notice that our AI OCR competitors have different names for their fields. Honestly, it can get confusing. Sorry for that 😔. There is no standards body that defines these names, so each company makes up its own.

Below we demystify our fields so you can better understand them and Sypht’s versatility.

Think of our fields hierarchically like this:


Key Points

  • Two classes of fields: derived fields & captured fields

  • Three types of derived fields: matched, conditional, and signal

  • Three types of captured fields: text, category, and table

What are Captured Fields?

Think of captured fields as any data in that exists in the document that the AI had extracted. Like, tables, invoice numbers, ABN number, numbers, text, etc, are all captured fields because they are data from the document. Text, category, and tables are all types of captured fields.

What are Derived Fields?

Derived fields are very different from captured fields. Derived fields are data that’s been taken from the document and then processed for insights. A good example of a derived field is our ABN matched field lookup. The lookup will take the ABN number from the document then derive the business name, location, and so on from an external ABN service and put that into the document.

Types of derived fields:

  • matched fields – derived by looking up public and/or private data. For example, (If you’re familiar with Excel, think of matched fields as VLOOKUP.

  • conditional fields - derived by logical operations e.g. (if x = 1, then y = 1.00) or something like that. Conditional fields are used for data cleansing and data transformation.

  • signal fields – derived by leveraging previously extracted data and then running calculations and operations on it. Examples include: duplicate signals, fraud detection, and so on.

Was this article helpful?
0 out of 0 found this helpful