Principle:Evidentlyai Evidently Dataset Score Extraction
| Knowledge Sources | |
|---|---|
| Domains | Data_Engineering, LLM_Evaluation |
| Last Updated | 2026-02-14 12:00 GMT |
Overview
A data extraction mechanism that converts an Evidently Dataset back to a pandas DataFrame including computed descriptor columns.
Description
Dataset Score Extraction retrieves the underlying pandas DataFrame from an Evidently Dataset object, including any descriptor columns that were computed during dataset creation. This allows access to row-level evaluation scores (sentiment, negativity, text length, etc.) for downstream processing such as database storage, filtering, or custom analysis.
Usage
Use after creating a Dataset with descriptors to access the computed evaluation scores as a DataFrame. Essential for the LLM Evaluation Monitoring workflow where individual scores are stored in a database.
Theoretical Basis
This follows the unwrapping pattern, the inverse of the Dataset creation adapter:
# Pseudocode
dataset = Dataset.from_pandas(df, descriptors=[...]) # Wrap + compute
eval_df = dataset.as_dataframe() # Unwrap with computed columns
# eval_df has original columns + descriptor columns