CCARF Practice Question

Paper 1 · Question 6

D5 Context Management & Reliability · Paper 1

A data extraction pipeline extracts financial figures from annual reports. The pipeline cites sources using inline text references like 'According to page 12 of the 2024 Annual Report, revenue was $4.2B.' When the extracted data is consumed by a downstream analytics system, the source attribution is consistently lost. What is the root cause and fix?

• A.The downstream system strips text formatting. Fix by making the citations bold or using a special delimiter
• B.Inline prose citations are fragile. Fix by outputting structured claim-source mappings pairing each value with source, page, and date.
• C.The downstream system needs to be modified to preserve all input text verbatim without any transformation
• D.Add a separate attribution database keyed by value hash that the downstream system queries to look up the source for any extracted figure.

Show answer & explanations

Correct answer: B

A: The problem is not formatting but data structure. Regardless of text formatting, inline prose citations are lost when text is parsed, summarised, or restructured by downstream systems.

B: Structured claim-source mappings survive downstream processing because they are data fields, not prose. An analytics system can consume {'value': '4.2B', 'source': '2024 Annual Report', 'page': 12} without losing attribution, whereas it will discard 'According to page 12...' during text processing.

C: Requiring downstream systems to preserve verbatim text is impractical and fragile. Systems legitimately need to transform, aggregate, and restructure data. Attribution should be in a format that survives transformation.

D: While a database could store attributions, this adds unnecessary infrastructure when the extraction pipeline can simply output structured mappings directly. The attribution should travel with the data, not require a separate lookup.

Related Context Management & Reliability questions

More: Paper 1 · All practice questions