AWS Textract

Extracts line items, totals, vendor details, and tax fields from a receipt or invoice image using Textract's AnalyzeExpense API.

Runs synchronous OCR on a document page to extract every word, line, and page-level text block, including handwritten content.

Identifies and returns every labelled field and its filled-in value from a structured form, preserving the field-name/value relationship.

Detects and reconstructs tables from a document image, returning rows and cells with their row/column coordinates intact.

Answers specific natural-language questions about a document's content by pointing Textract's Queries feature at target fields.

Locates and flags signature fields within a document, confirming whether a signature is present in each designated block.

Identifies structural layout elements — titles, headers, section text, list items, figures, footers, and page numbers — without performing deep field extraction.

Reads passports, driver's licences, and national ID cards, returning structured fields: name, date of birth, document number, expiry, and issuing country.

Submits a multi-page document stored in S3 for asynchronous OCR text detection, returning a job ID to poll later.

Polls a previously submitted async text detection job by its job ID and returns the extracted text blocks when the job reaches SUCCEEDED status.

Submits a multi-page document to asynchronous analysis for forms, tables, queries, signatures, and/or layout — all in one call — returning a job ID.

Retrieves the structured results of a previously submitted async document analysis job, paginating through all blocks when the job is complete.

Submits a multi-page expense document — receipts, invoices, statements — stored in S3 for asynchronous AnalyzeExpense processing.

Retrieves the structured expense fields — line items, totals, vendor, payment terms — from a previously submitted async expense analysis job.

Submits mortgage, loan, or lending application documents for async analysis using Textract's specialised Lending API, which understands bank statements, pay stubs, W-2s, and tax returns.

Retrieves structured lending document data — income figures, asset balances, employment details — from a completed async lending analysis job.

Creates a named Textract Adapter — a fine-tuned model layer trained on your specific document type — to improve extraction accuracy beyond the general model.

Returns all Textract Adapters in the account with their IDs, names, and status, so agents can select or audit available custom models.

Fires when a new document file lands in a monitored S3 bucket or prefix, initiating the extraction pipeline automatically.

Fires when a previously submitted Textract async job (text detection, document analysis, expense, or lending) transitions to SUCCEEDED or FAILED status.

Fires when an email containing a document attachment arrives in a monitored inbox, handing the attachment to the extraction pipeline.

Fires on a user-defined schedule (daily, weekly, or at a set time) to trigger bulk document processing across all files in a monitored location.

Fires when a user submits a web form or uploads a document through a portal, triggering extraction of the submitted document's fields.

Fires when Textract returns one or more extracted fields with a confidence score below a configured threshold, routing the document to human review.

Fires when a monitored field in an extracted document matches a configured value or condition, enabling downstream conditional workflows.

Fires when a custom Textract Adapter version finishes training and is ready for deployment in analysis jobs.