Implementation:Explodinggradients Ragas RougeScore Metric

Field	Value
source	Repo
domains	Metrics, NLP
last_updated	2026-02-10

Overview

RougeScore computes ROUGE (Recall-Oriented Understudy for Gisting Evaluation) n-gram overlap scores between a reference and a generated response using the rouge-score library.

Description

The RougeScore class evaluates text generation quality by computing ROUGE scores, which measure n-gram overlap between a reference and response text. It supports two ROUGE variants (rouge1 for unigrams and rougeL for longest common subsequence) and three scoring modes (fmeasure, precision, recall). The metric uses stemming for improved matching. It does not require an LLM and inherits only from SingleTurnMetric.

Key attributes:

• rouge_type -- The type of ROUGE score: "rouge1" or "rougeL" (default "rougeL").
• mode -- The scoring mode: "fmeasure" (default), "precision", or "recall".

Dependency: Requires the rouge-score package (pip install rouge-score).

Usage

The metric requires reference and response columns.

Code Reference

Property	Value
Source Location	src/ragas/metrics/_rouge_score.py L11-43
Class Signature	class RougeScore(SingleTurnMetric)
Import	from ragas.metrics import RougeScore

I/O Contract

Inputs

Parameter	Type	Required	Description
reference	str	Yes	The ground truth reference text
response	str	Yes	The generated response to evaluate

Outputs

Output	Type	Description
score	float	ROUGE score (0.0 to 1.0) based on the selected type and mode

Usage Examples

from ragas.metrics import RougeScore
from ragas.dataset_schema import SingleTurnSample

metric = RougeScore(rouge_type="rougeL", mode="fmeasure")

sample = SingleTurnSample(
    reference="The cat sat on the mat.",
    response="The cat is sitting on the mat."
)
# score = await metric.single_turn_ascore(sample)

Related Pages

• Explodinggradients_Ragas_BleuScore_Metric -- BLEU-based n-gram precision metric
• Explodinggradients_Ragas_ChrfScore_Metric -- Character n-gram F-score metric
• Explodinggradients_Ragas_StringMetrics_Module -- Non-LLM string comparison metrics