Inscribe's parsing feature extracts key details from English language bank statements, pay slips, utility bills, tax documents, and identity documents. This article will help you understand how Inscribe's parsing feature can help protect you against fraudulent applicants, and how you can use its outputs to automate your processes.
What we parse
Key details
For all the documents above, Inscribe will look for three key details: the name of the document owner, the address of the document owner, and the document date. For bank statements, the document date is the last day of the period to which the statement pertains; for utility bills, the document date is the date the bill was prepared; and for pay slips, the document date is the check date or advice date.
Details specific to bank statements
In addition to the key details above, we also parse the account number, account name beginning balance, ending balance and transaction data from bank statements.
Details specific to pay slips
In addition to the key details above, we also parse the name and address of the business issuing the payslip, the pay period, gross and net amounts of pay during that period, and the year-to-date gross and net amounts of pay.
Details specific to identity documents
In addition to the key details above, we also parse the expiry date and date of birth.
Details specific to tax documents (CP 575, 147C, W2)
In addition to the key details above, we also parse all EINs and SSNs present on the document.
Working with parsed details
Getting the results
Parsed details are returned in the Get Document Results API response, under the 'parsed_details' key. For each parsed detail, a confidence score between 0 and 1 indicates Inscribe's certainty about the result.
Result formats
Depending on the detail parsed, how the detail is formatted in the API response may vary, and additional enriched data is returned where appropriate:
Date information will always include data in the standardised YYYY-MM-DD format.
Address information will always include structured output that identifies country, state and lower level address details.
Pay amounts and transaction amounts will always include a standardised output that expresses the amount as an integer number of cents.
Confidence scores
Parsing performance can vary depending on the quality and format of the document. Inscribe's confidence scores allow you to identify when you wish to accept the details parsed by Inscribe.
Each detail parsed by Inscribe comes with an individual confidence score, with the exception of transaction data, where a single score returned corresponds to all transactions.
Selecting a confidence score cut-off will depend on your use case. Choosing a low confidence score threshold will allow you to automate more documents, whereas choosing a high confidence score cut-off will reduce the risk of accepting incorrect parsed details.
Inscribe recommends starting with a high confidence score cut-off, such as 0.98, and confirming that performance is satisfactory on your document set before reducing to meet your automation goals.
Getting the best performance
Inscribe's parsing feature works best on high quality documents. As a rule of thumb, if it's difficult for a human to make out the text in the document, it will also be difficult for Inscribe.