You are tasked with using the 'PARSE DOCUMENT' function in Snowflake to extract key information (name, address, phone number) from a large collection of scanned invoices stored as PDF files in an AWS

Question

Options

• AUse `PARSE DOCUMENT with default settings and load the raw JSON output into a VARIANT
• BUse `PARSE DOCUMENT with a pre-defined JSON schema to enforce a rigid structure on the
• CCreate a custom UDF (User-Defined Function) that calls 'PARSE_DOCUMENT and then uses
• DEmploy a combination of 'PARSE DOCUMENT and Snowflake's external functions to integrate
• EDirectly load PDF files into a relational table's TEXT column and write SQL queries utilizing LIKE

• A
  3% (1)
• B
  8% (3)
• C
  84% (31)
• E
  5% (2)

Explanation

Creating a custom UDF that calls PARSE_DOCUMENT and applies additional post-processing logic is the most effective approach for invoices with varying formats and quality. The UDF can normalize the raw JSON output from PARSE_DOCUMENT, apply validation rules, handle missing fields gracefully, and map extracted values to a consistent structure (name, address, phone number) regardless of invoice layout differences. This encapsulates the transformation logic in a reusable, maintainable unit. Option A (loading raw JSON into VARIANT) defers all structuring to query time, making analysis harder. Option B (pre-defined rigid schema) fails for varying invoice formats. Option D (external functions) adds unnecessary complexity and latency for this use case. Option E (LIKE queries on raw text) is fragile and impractical for structured extraction from PDFs.

No community discussion yet for this question.