1.3.4 · Lesson 11 of 28
Enhance input data quality
Improve the information given to the FM before inference. Examples:
WHAT
Improve the information given to the FM before inference.
Examples:
- normalize names
- extract entities
- standardize terminology
- reformat text
- remove irrelevant content
- enrich metadata
AWS services
- Amazon Bedrock
- Amazon Comprehend
- AWS Lambda
Example
Raw customer text
↓
Entity extraction
↓
Normalization
↓
Structured representation
↓
FMWHY
Cleaner, more structured input can improve consistency and reduce ambiguity.
EXAM TRAP
Input enhancement is not the same as hallucination-producing enrichment.
Only add trusted information.
---
Key points
- Cleaner, more structured input can improve consistency and reduce ambiguity.
Exam focus
Focus on requirement-driven choices and common architecture traps.