BLEU Score Calculator
Calculate BLEU (Bilingual Evaluation Understudy) scores for translation and text generation
Input Texts
6 tokens
7 tokens
BLEU Score
0.1
Very Poor
1-gram
57.1%
2-gram
33.3%
3-gram
20.0%
4-gram
0.0%
Calculation Details
P1: 4/7 = 57.1%
P2: 2/6 = 33.3%
P3: 1/5 = 20.0%
P4: 0/4 = 0.0%
BP: 1.0000 (hyp=7, ref=6)
BLEU = 1.0000 × 0.0014 = 0.0014
Related Tools
Cosine Similarity Calculator
Calculate similarity between two vectors or text embeddings
Embedding 3D Visualizer
Visualize high-dimensional embeddings in 2D/3D using PCA/t-SNE simulation
Perplexity Explainer
Calculate and understand perplexity from probability distributions
ROUGE Score Calculator
Calculate ROUGE-N and ROUGE-L scores for summarization tasks
LLM Temperature Simulator
Visualise how temperature and top-p sampling affect next-token probabilities
Tokenization Visualizer
See how text is broken down into tokens by different tokenizers (BPE, WordPiece)
What is BLEU Score?
BLEU (Bilingual Evaluation Understudy) is a metric for evaluating machine translation quality. It measures how similar a generated text is to a reference text by comparing n-gram overlaps.
BLEU scores range from 0 to 1 (or 0 to 100 when expressed as percentages), where higher scores indicate better matches with the reference.
How BLEU Works
N-gram Precision
Count matching n-grams (1-4 words) between hypothesis and reference.
Clipped Counts
Clip n-gram counts to prevent gaming by repeating words.
Brevity Penalty
Penalize translations that are too short compared to reference.
Geometric Mean
Combine precisions using geometric mean for final score.
Score Interpretation
| Score | Quality | Interpretation |
|---|---|---|
| >60 | Excellent | High quality, near-human translation |
| 40-60 | Good | Understandable, mostly accurate |
| 20-40 | Fair | Gist is preserved, rough translation |
| <20 | Poor | Low similarity to reference |
Limitations of BLEU
BLEU only measures surface-level n-gram overlap. It doesn't capture meaning, fluency, or semantic equivalence. A paraphrase may have low BLEU despite being a good translation.
Related Tools
ROUGE Score
Calculate ROUGE for summarization evaluation.
Perplexity Calculator
Calculate perplexity for language models.