TMs &Term databases
• Pre-clean and normalize for prompt injection/RAG.
• “Flatten” and enrich with metadata.
• Remove noise.
• Split into relevant partitions.
Prompt creation
• Goes hand-in-hand with data preparation.
• Prompts are adjusted for each TM partition.
• Data and context are injected at runtime.
DATA PREPARATION IS KEY
Preparing input data is
half of the overall effort.
• Pre-clean and normalize for prompt injection/RAG.
• “Flatten” and enrich with metadata.
• Remove noise.
• Split into relevant partitions.
Prompt creation
• Goes hand-in-hand with data preparation.
• Prompts are adjusted for each TM partition.
• Data and context are injected at runtime.
DATA PREPARATION IS KEY
Preparing input data is
half of the overall effort.