Extract free text from unstructured documents using large language model (LLM)-based SenseML methods. For example, extract information from legal paragraphs in contracts and leases, or results from research papers.

The following LLM-based methods are alternatives to layout-based methods for structured documents, for example, tax documents or insurance forms.

Method	Example use case	Chained-prompt example¹	Notes
List method	"For each vehicle in an auto insurance declaration, extract the VIN, model, and year."	"For the extracted list of vehicles, sort them by year of manufacture"	Extracts a list of data from a document, where you don't know how the data are represented.
Query Group method	"When does the policy period end?" "What's the account number?"	"Redact the account number by returning only the last 4 digits"	Extracts a single fact or data point.
NLP Table method	"For each transaction in a bank statement table, extract the date and amount."	N/A	Extracts a list of data from a document, where you know they're in a table.

Notes

¹ For information about chained prompts, see Computed field methods.
For layout-based extraction, see Layout-based methods.