What is the primary output of this skill?

It generates a structured data directory containing a source summary, a JSON file of all feedback units, and a JSONL file containing one training pair per line for easy machine learning ingestion.

Can this skill help categorize the type of feedback extracted?

Yes, it automatically categorizes feedback by scope (specific, section, or document), type (structural, methodological, etc.), and suggested action (revise, add, clarify, etc.).

How does it handle feedback that doesn't use a specific quote?

It uses a pattern recognition approach to synthesize the relevant source context, capturing 2-3 representative examples of the issue to ensure the LLM understands the underlying pattern.

What file formats does this skill support for source documents?

The skill supports PDF, DOCX, and plain text files. It uses specialized extraction libraries to preserve document structure, including page numbers and headings.

Academic Review Training Extractor

Name: Academic Review Training Extractor
Author: nicsuzor

bynicsuzor

0•

데이터 과학 및 ML

Extracts structured training pairs from academic peer reviews and source documents to build high-quality datasets for LLM fine-tuning.

This skill automates the creation of specialized training datasets designed to teach LLMs how to perform academic peer reviews in a specific style. By processing matched pairs of reviews and source documents (PDF, DOCX, or TXT), it identifies distinct feedback units and maps them back to the exact evidence or patterns in the original text. It categorizes critiques by scope, type, and action, providing the necessary context for models to learn complex pattern recognition, such as identifying methodological gaps or structural inconsistencies in academic writing.

주요 기능

01Standardized JSONL output optimized for direct ingestion into LLM training pipelines

02Synthesis of high-level patterns for training models on non-specific critiques

03Intelligent categorization of feedback into scope, type, and action labels

040 GitHub stars

05Support for PDF, DOCX, and TXT source file formats with structure preservation

06Automated extraction of feedback-evidence pairs from academic documents

사용 사례

01Creating pedagogical datasets for teaching graduate students how to formulate effective critiques

02Analyzing historical review data to identify common patterns in academic manuscript issues

03Building datasets to fine-tune AI models in a specific researcher's peer review style

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add nicsuzor/academicops review-training

For use in Claude.ai and ChatGPT

Download Skill