Confidence Filtering

Overview:

The system filters tags based on a configurable confidence threshold to remove low-confidence results before returning or persisting data.

Prerequisites:

• Parsed tag output from the shared parser
• Lambda integration
• Environment variable support for threshold configuration

Implementation/Steps:

1. The system applies filtering after tag normalization.
2. The system removes tags below the active confidence threshold.
3. The system reads the threshold from an environment variable or defaults to 80.0.
4. The system returns and persists only filtered tags.
5. The system allows empty tag results when all values fall below the threshold.

Key Takeaways:

• Filtering occurs in a single consistent stage.
• The threshold is inclusive at 80.0.
• Default behavior applies when no configuration is provided.
• Empty results are valid outcomes.