Snowflake SnowPro Specialty: Gen AI (GES-C02) · 15% practice allocation

Snowflake GES-C02: Document Processing Practice Questions

5 free sample questions from the Document Processing domain, with answers and explanations. The full GES-C02 bank has 241 questions across 4 domains. Questions are scenario-based and written from the official Snowflake exam guide.

1.

A team wants to build full-text search over PDF contracts on a stage — no specific fields need to be pulled out, just the searchable text. Which function is the right starting point?

  • A. SPLIT_TEXT_RECURSIVE_CHARACTER, run directly on the raw PDF bytes
  • B. AI_EXTRACT, since it is the general-purpose entry point for any document task
  • C. A directory table, which extracts text as a side effect of listing files
  • D. PARSE_DOCUMENT (or AI_PARSE_DOCUMENT), to extract the text itself✓ Correct
Explanation

When the goal is just the text itself, PARSE_DOCUMENT is the correct and sufficient starting point. AI_EXTRACT is for structured fields (a different, narrower need), chunking needs already-extracted text to operate on, and a directory table only holds file metadata.

2.

A finance team wants queryable invoice_number, total_amount, and due_date columns sourced from scanned invoices. Which function should they use after the invoices are parsed?

  • A. SPLIT_TEXT_RECURSIVE_CHARACTER, since chunking produces structured fields
  • B. AI_EXTRACT, to pull the named fields into structured columns✓ Correct
  • C. PARSE_DOCUMENT again, since running it a second time yields structured fields
  • D. EMBED_TEXT_768, since embeddings are structured data
Explanation

AI_EXTRACT is purpose-built for pulling named structured fields like invoice number or total into queryable columns. Re-running PARSE_DOCUMENT still only yields text, and chunking or embedding do not produce named fields.

3.

A team needs the raw text of scanned personal letters for keyword search only; no tables or headings in the source carry meaning. Which is the more appropriate, cost-effective choice?

  • A. OCR mode, since structure preservation isn't needed here✓ Correct
  • B. Neither mode — call SPLIT_TEXT_RECURSIVE_CHARACTER directly on the raw file
  • C. LAYOUT mode, since it should always be preferred regardless of content
  • D. Neither mode — call AI_EXTRACT directly on the raw file
Explanation

When structure carries no meaning, OCR mode's plain text extraction is sufficient and simpler. LAYOUT mode isn't wrong here, just unnecessary; AI_EXTRACT and chunking both require parsed text as input, not a raw file.

4.

Select two categories of Snowflake stage that can serve as the source location for documents processed by PARSE_DOCUMENT.

  • A. Resource monitors
  • B. Directory tables
  • C. External stages✓ Correct
  • D. Virtual warehouses
  • E. Internal stages✓ Correct
Explanation

Internal (Snowflake-managed) and external (customer-owned cloud storage granted access) stages are where source documents live. A directory table is metadata over a stage, not a stage type itself, and virtual warehouses and resource monitors don't hold source files.

5.

A pipeline parses contracts from an external stage, must preserve table structure for exhibit schedules, and needs to produce both a searchable text store and a table of extracted effective dates. Select two functions or features this pipeline must use.

  • A. SPLIT_TEXT_RECURSIVE_CHARACTER as the only step needed for the whole pipeline
  • B. A masking policy to pull out the effective dates
  • C. AI_EXTRACT for the effective dates✓ Correct
  • D. A resource monitor to pull out the effective dates
  • E. PARSE_DOCUMENT in LAYOUT mode✓ Correct
Explanation

LAYOUT mode is required because table structure in the exhibit schedules must survive parsing, and AI_EXTRACT is the function built to pull the named effective-date field into a queryable form. Chunking alone can't extract structured fields, and resource monitors and masking policies have nothing to do with field extraction.

Test yourself across every GES-C02 domain

The free GES-C02 practice test mixes about 20 questions weighted like the exam. The full 241-question bank adds instant grading, flashcards and a timed mock exam.

Take the free GES-C02 practice testPractice the full bank

Study this domain

Document Processing study guide (GES-C02) →

Other GES-C02 domains

← All Snowflake SnowPro Specialty: Gen AI study resources