Implementation:Evidentlyai Evidently Legacy Text Evals Preset

Knowledge Sources	Evidentlyai_Evidently
Domains	ML Monitoring, NLP, Text Analysis
Last Updated	2026-02-14 12:00 GMT

Overview

TextEvals is a metric preset that generates column summary metrics for a set of text descriptors applied to a specified text column.

Description

The TextEvals class extends MetricPreset and provides a convenient way to evaluate text quality characteristics on a given column. It applies a list of text feature descriptors to the target column and produces a ColumnSummaryMetric for each descriptor.

If no custom descriptors are provided, the preset uses a default set of five descriptors:

• TextLength -- measures the length of text
• SentenceCount -- counts the number of sentences
• Sentiment -- computes sentiment scores
• OOV -- measures out-of-vocabulary word ratio
• NonLetterCharacterPercentage -- computes the percentage of non-letter characters

Each descriptor is applied to the specified column via desc.on(self.column_name), and the result is wrapped in a ColumnSummaryMetric.

The class is registered with the type alias "evidently:metric_preset:TextEvals".

Usage

Use this preset when you need to quickly evaluate text quality characteristics of a text column in your dataset. It is suitable for monitoring text data quality in NLP pipelines and detecting changes in text properties between reference and current data.

Code Reference

Source Location

• Repository: Evidentlyai_Evidently
• File: src/evidently/legacy/metric_preset/text_evals.py

Signature

class TextEvals(MetricPreset):
    class Config:
        type_alias = "evidently:metric_preset:TextEvals"

    column_name: str
    descriptors: Optional[List[FeatureDescriptor]] = None

    def __init__(self, column_name: str, descriptors: Optional[List[FeatureDescriptor]] = None):
        ...

    def generate_metrics(
        self, data_definition: DataDefinition, additional_data: Optional[Dict[str, Any]]
    ) -> List[AnyMetric]:
        ...

Import

from evidently.legacy.metric_preset.text_evals import TextEvals

I/O Contract

Inputs

Name	Type	Required	Description
column_name	str	Yes	The name of the text column to evaluate.
descriptors	Optional[List[FeatureDescriptor]]	No	A list of feature descriptors to apply. Defaults to [TextLength, SentenceCount, Sentiment, OOV, NonLetterCharacterPercentage].

Outputs

Name	Type	Description
return	List[AnyMetric]	A list of ColumnSummaryMetric instances, one per descriptor applied to the specified text column.

Usage Examples

from evidently.legacy.metric_preset.text_evals import TextEvals
from evidently.legacy.report import Report

# Evaluate text column with default descriptors
report = Report(metrics=[TextEvals(column_name="review_text")])
report.run(reference_data=ref_df, current_data=curr_df)

# Evaluate text column with custom descriptors
from evidently.legacy.descriptors import TextLength, Sentiment
report = Report(metrics=[
    TextEvals(column_name="review_text", descriptors=[TextLength(), Sentiment()])
])
report.run(reference_data=ref_df, current_data=curr_df)

Related Pages

• Environment:Evidentlyai_Evidently_Python_Core_Environment