Natural Language Generation (NLG): Which choices describe core planning decisions an NLG system must make before producing output?

Introduction / Context:
Natural Language Generation systems plan and realize text from structured data. This question evaluates understanding of content determination and discourse planning, two fundamental NLG planning stages.

Given Data / Assumptions:

• NLG requires selecting relevant information (what to say).
• NLG must organize and time information (when and how to say it).
• High-level business purpose is typically specified externally by the application.

Concept / Approach:
Core NLG pipeline: content selection, document planning (ordering/timing), microplanning (lexicalization/aggregation), and surface realization. The system must decide what content to include and how to structure its presentation over a discourse.

Step-by-Step Solution:

Verification / Alternative check:
Standard NLG architectures consistently list content selection and discourse planning as essential pre-realization steps.

Why Other Options Are Wrong:

• (a) or (b) alone is incomplete; both are necessary.
• (c) is external to the generator; it does not describe a generation decision.
• None of the above contradicts NLG fundamentals.

Common Pitfalls:
Confusing strategic goals of the application with the tactical planning decisions the NLG system performs.

Final Answer:
Both (a) and (b)