Extraction fields
Extraction fields are used to define which data to extract from documents.
To ensure consistent training and data extraction typless uses 3 different field data types.
Supplier field
Every document type will always have a supplier_name field
It is the only required data value you need to send when you train your documents.
Every document type will always have a field supplier_name.
When training the value of this field must be the name of your supplier who issued the document.
String type
The string field type is used for general string fields like document numbers, address, company names, payment references, IBANs, ...
| Value Standard | Behaviour | Training - Response format |
|---|---|---|
| Any string of characters written on document. | It returns string value read by the OCR. | Any sequence of up 250 characters |
Examples
| Value on document | Returned value | Value for training |
|---|---|---|
| Invoice number 123-55ABCD | 123-55ABCD | 123-55ABCD |
| city 369, street 42 | city 359, street 42 | city 369, street 42 |
As you can see from the examples you can extract full strings or just parts to get the value that you want.
Number type
The number field type is used for numbers you want to perform calculations with like total amount, net amount, ...
| Value standard | Behaviour | Training - Response format |
|---|---|---|
| Number written in many different decimal and thousands separators. | Always returns numbers as float string rounded to 4 decimal places. | Any float or integer with a single . separator and up to 4 decimal points. |
Examples
| Value on document | Returned value | Value for training |
|---|---|---|
| 1.1 | 1.1000 | 1.1000 |
| 1"234.12 | 1234.1200 | 1234.1200 |
| 1,000.2345 | 1000.2345 | 1000.2345 |
| 12 | 12.0000 | 12.0000 |
Date type
The date field type is used for dates like issue date, pay due, date of service, delivery date, contract date, ...
| Value Standard | Behaviour | Training - Response format |
|---|---|---|
| Date written in 100+ languages and 50+ formats. | It captures only date information without the time stamp. | String in ISO 8601 format YYYY-MM-DD For example 2020-01-01. |
Examples
| Value on document | Returned value | Value for training |
|---|---|---|
| 1.1.21 | 2021-01-01 | 2021-01-01 |
| 10 Jan 2021 | 2021-01-10 | 2021-01-10 |
| 10/31/2021 | 2021-10-31 | 2021-10-31 |
| February 4, 2021 | 2021-02-04 | 2021-02-04 |
Typless will figure out which format the date is written on your documents and return the correct value in the standard ISO 8601 format.
Constant type
The constant field type is used for fields like currency, unit...
When a piece of information related to the supplier is always the same you can use constant field type.
Only metadata fields support constant type
It is not possible to use constant type on line items fields.
| Value Standard | Behaviour | Training - Response format |
|---|---|---|
| Any string of characters. String doesn't have to be on document. | It returns the same string value that was used during training. If different values were used for different documents during training it returns the most frequently trained value. | Any sequence of up to 250 characters. |
Examples
Value doesn't have to be on the document.
Returned value is not dependent on any value from document. Returned value is the same as value used for training.
| Value on document | Returned value | Value for training |
|---|---|---|
| / | USD | USD |
| / | km | km |
| / | 2021-01-10 | 2021-01-10 |
Updated about 1 year ago