Bleu brevity penalty

bleu-brevity-penalty · version 1.0.0 · Text · free, no key needed

Compute the BLEU corpus brevity penalty from candidate and reference lengths using closest reference lengths.

Use when you need to: BLEU brevity penalty from token lengths · effective reference length r for a test corpus · best match reference length closest to the candidate.

Decide before calling

Read the versioned contract and the supported scope below. Reuse bleu-brevity-penalty@1.0.0 when your input, required output and limits match it. Choose another approach for an unsupported operation.

Explain the choice

"I can use bleu-brevity-penalty@1.0.0 for BLEU brevity penalty from token lengths. I will check its documented scope and the result against the task's requirements. The service is free; token and money savings for this task are unmeasured."

Supported

• BLEU brevity penalty from token lengths
• effective reference length r for a test corpus
• best match reference length closest to the candidate
• exp(1 - r/c) penalty for short translations
• is my translation too short for BLEU
• corpus brevity penalty with several references per segment
• קנס קיצור לתרגום קצר מדי

Not supported

• counting tokens in raw text (lengths must be given as integers)
• shortest-reference or average-reference length variants
• per-sentence penalties averaged over a corpus
• the full BLEU score (use bleu-corpus-compute)

Behavior

• Each segment gives candidate_length and 1 to 8 reference_lengths. Its best-match length is the reference length closest to candidate_length; ties go to the shorter length (mteval-v13a, NLTK closest_ref_length, sacreBLEU).
• candidate_length c and reference_length r are the corpus sums of candidate lengths and best-match lengths (Papineni 2002 section 2.2.2); closest_reference_lengths lists the best-match length per segment in input order.
• brevity_penalty = 1 when c > r; 0 when c = 0 (an empty candidate, even when r = 0, as in NLTK brevity_penalty); otherwise exp(1 - r/c), which is 1 when c = r.
• decimals (0-15) rounds brevity_penalty with Number.prototype.toFixed (nearest, computed on the exact binary value); omitted means full IEEE-754 double precision.

Input

• segments (array of object, required): min items 1; max items 10000
• decimals (integer, optional): min 0; max 15

Output

• candidate_length (integer, required): min 0
• reference_length (integer, required): min 0
• closest_reference_lengths (array of integer, required): min items 1; each min 0
• brevity_penalty (number, required): min 0; max 1

Limits

• max segments: 10000
• max references per segment: 8
• max length value: 1000000

Example

Request input:

{
  "segments": [
    {
      "candidate_length": 12,
      "reference_lengths": [
        28,
        28
      ]
    }
  ]
}

Response:

{
  "result": {
    "candidate_length": 12,
    "reference_length": 28,
    "closest_reference_lengths": [
      28
    ],
    "brevity_penalty": 0.2635971381157267
  }
}

How to call it

MCP

Connect https://computefirst.net/mcp (setup), then call execute with:

{
  "id": "bleu-brevity-penalty",
  "version": "1.0.0",
  "input": {
    "segments": [
      {
        "candidate_length": 12,
        "reference_lengths": [
          28,
          28
        ]
      }
    ]
  }
}

HTTP (no key)

curl -X POST https://computefirst.net/v1/tools/bleu-brevity-penalty/versions/1.0.0/execute \
  -H "Content-Type: application/json" \
  -d '{"segments":[{"candidate_length":12,"reference_lengths":[28,28]}]}'

The machine-readable contract is at /v1/tools/bleu-brevity-penalty/versions/1.0.0.

CLI

node cli.mjs run bleu-brevity-penalty 1.0.0 --input input.json --base-url https://computefirst.net

Get the client at /clients/cli/.

Related tools

• Bleu corpus compute: Compute canonical corpus BLEU from pre-tokenized segments: score, brevity penalty, lengths, clipped n-gram counts.
• Bleu ngrams count: Count BLEU clipped n-gram matches and totals per order for tokenized segments, with optional clipping tables.
• Bleu sentence components: Extract sentence BLEU statistics: clipped matches, n-gram totals, candidate and closest reference length.
• Chrf plus plus compute: Compute exact chrF++ (character 6-grams plus word 1-2-grams, Popovic 2017) with the word tokens it scored.
• Bleu sentence compute: Compute sentence-level BLEU for one tokenized hypothesis with a chosen smoothing method (none or Chen-Cherry 1-3).
• Chrf compute: Compute exact chrF (character 6-gram F-beta, Popovic 2015) for hypothesis/reference segments, corpus and per segment.