rouge-chinese

1.0.3 verified Fri May 01 auth: no python

A Python implementation of ROUGE score specifically designed for Chinese text evaluation. It supports both character-level and word-level ROUGE-1, ROUGE-2, and ROUGE-L metrics. Version 1.0.3 is the latest, with no recent updates (last release likely 2021).

pip install rouge-chinese

Common errors

error AttributeError: module 'rouge' has no attribute 'Rouge' ↓

cause Wrong import: using 'import rouge' instead of 'from rouge_chinese import Rouge'.

fix

Install rouge-chinese and import as 'from rouge_chinese import Rouge'.

error TypeError: get_scores() takes 3 positional arguments but 4 were given ↓

cause Passing arguments incorrectly; the API expects (hypotheses, references) as lists of strings.

fix

Call as 'rouge.get_scores(hypotheses, references)' with two lists.

error ImportError: No module named 'jieba' ↓

cause Missing dependency jieba required for Chinese word segmentation.

fix

Install jieba: 'pip install jieba'.

Warnings

breaking The library 'rouge-chinese' is commonly confused with 'rouge' (the general ROUGE implementation). Importing from 'rouge' instead of 'rouge_chinese' will import a different library with different API. ↓

fix Use 'from rouge_chinese import Rouge' instead of 'import rouge'.

gotcha The library returns scores as a dictionary with keys like 'rouge-1', 'rouge-2', 'rouge-l', but the output format differs from the original 'rouge' library. The 'f' key is present but some users expect 'f1-score'. ↓

fix Access scores via scores[0]['rouge-1']['f'] and similar.

deprecated The library has not been updated since 2021 and may have compatibility issues with newer Python versions (e.g., Python 3.10+). It relies on 'jieba' which may change its API. ↓

fix Consider pinning dependencies (jieba==0.42.1) or switching to 'rouge' library with proper Chinese tokenization.

Imports

rouge
wrong
```
import rouge
```
correct
```
from rouge_chinese import Rouge
```
The correct import is from rouge_chinese, not the generic 'rouge' package.

Quickstart

Basic usage computing ROUGE scores for Chinese text.

from rouge_chinese import Rouge

# Example hypotheses and references (Chinese text)
hypotheses = ["这是一个测试句子"]
references = ["这是一个参考句子"]

rouge = Rouge()
scores = rouge.get_scores(hypotheses, references)
print(scores)