​
Input

  • output (str): The generated text
  • expectedOutput (str): The reference text

​
Output

  • Result (float): A score between 0 and 1.

​
Interpretation

  • Higher scores (closer to 1): Better bigram overlap (captures local word order and fluency)
  • Lower scores (closer to 0): Poor bigram overlap

​
Formula

ROUGE-2=Count of overlapping bigramsTotal bigrams in reference\mathrm{ROUGE\text{-}2} = \frac{\text{Count of overlapping bigrams}}{\text{Total bigrams in reference}}

​
Example Calculation:

  • Reference bigrams: [the cat, cat sat, sat on, on the, the mat]
  • Candidate bigrams: [the cat, cat sits, sits on, on the, the mat]
  • Overlapping bigrams: [the cat, on the, the mat] = 3
  • Total reference bigrams = 5
ROUGE-2=35=0.6\mathrm{ROUGE\text{-}2} = \frac{3}{5} = 0.6
This is a Similarity Metric

​
Use Cases

  • Evaluating text summarization where sentence structure matters
  • Assessing machine translation quality