Recipe catalog / review-recommendation-kind

Classify a peer review's recommendation

What recommendation does the wording of review express: accept, minor revision, major revision, or reject?

An editor or review-management tool needs to read the recommendation implied by a free-text review, for example when the reviewer skipped the form field or when the text and the ticked box disagree.

Explore this recipe interactively · Source and implementation guide

Use review-recommendation-kind in TypeScript

Install with npm install jev-recipes. Requires Node.js 22.9 or newer and ES modules. Set TYPESAFE_API_KEY in your server environment for live calls, which send input to TypeSafe and use API quota. See the installation guide.

import { reviewRecommendationKind } from 'jev-recipes/review-recommendation-kind';

const result = await reviewRecommendationKind({
  "review": "The paper addresses an important question and the experimental setup is generally careful. However, the central claim in Section 5 rests on the independence assumption introduced in Section 3, which the authors never test, and the baseline comparison omits the two strongest recent methods. I would need to see the assumption validated and the comparison extended before I could support publication. The writing issues I noted in the margins are secondary and easy to fix.",
  "minConfidence": 0.8
});
console.log(result);

Input contract

{
  "input": {
    "$schema": "https://json-schema.org/draft/2020-12/schema",
    "type": "object",
    "properties": {
      "review": {
        "type": "string"
      },
      "minConfidence": {
        "type": "number",
        "minimum": 0,
        "maximum": 1
      }
    },
    "required": [
      "review"
    ]
  },
  "result": {
    "$schema": "https://json-schema.org/draft/2020-12/schema",
    "type": "object",
    "properties": {
      "model": {
        "type": "string"
      },
      "usage": {
        "type": "object",
        "properties": {
          "input_tokens": {
            "type": "integer",
            "minimum": 0,
            "maximum": 9007199254740991
          },
          "output_tokens": {
            "type": "integer",
            "minimum": 0,
            "maximum": 9007199254740991
          }
        },
        "required": [
          "input_tokens",
          "output_tokens"
        ],
        "additionalProperties": false
      },
      "status": {
        "type": "string",
        "enum": [
          "ready",
          "review"
        ]
      },
      "verdict": {
        "type": "string",
        "enum": [
          "accept",
          "minor_revision",
          "major_revision",
          "reject",
          "unclear"
        ]
      },
      "confidence": {
        "type": "number",
        "minimum": 0,
        "maximum": 1
      },
      "probabilities": {
        "type": "object",
        "propertyNames": {
          "type": "string",
          "enum": [
            "accept",
            "minor_revision",
            "major_revision",
            "reject",
            "unclear"
          ]
        },
        "additionalProperties": {
          "type": "number",
          "minimum": 0,
          "maximum": 1
        },
        "required": [
          "accept",
          "minor_revision",
          "major_revision",
          "reject",
          "unclear"
        ]
      }
    },
    "required": [
      "model",
      "usage",
      "status",
      "verdict",
      "confidence",
      "probabilities"
    ],
    "additionalProperties": false
  }
}

Saved example result

This hand-authored response demonstrates the contract. It is not a model accuracy measurement. Run it without an API key: npx jev-recipes demo review-recommendation-kind.

{
  "model": "demo-fixture",
  "usage": {
    "input_tokens": 0,
    "output_tokens": 0
  },
  "status": "ready",
  "verdict": "major_revision",
  "confidence": 0.86,
  "probabilities": {
    "accept": 0.01,
    "minor_revision": 0.04,
    "major_revision": 0.86,
    "reject": 0.07,
    "unclear": 0.02
  }
}

Evaluation evidence

Use the evaluation guide to measure this decision on your own labeled cases.

Limitations

• Reads the recommendation the wording expresses, not whether the reviewer's assessment is fair or the paper is any good.
• A review that mixes praise and severe objections is classified by the disposition its wording commits to; when it commits to none, the result is unclear.
• Journal-specific decision categories, such as reject and resubmit, must be mapped from these four in application code.

Related recipes

• draft-compare: Use draft-compare to judge which of two revisions is better, rather than what a reviewer recommended.
• feedback-kind: Use feedback-kind to classify general user feedback, rather than the recommendation in a formal peer review.