Does it handle complex tables and charts in PDFs?

Yes, it utilizes Claude’s advanced vision capabilities to interpret complex layouts and extract data into structured JSON format.

What file formats can I export the results to?

Extracted and validated data can be exported to various formats including Python objects, R dataframes, CSV, Excel, and SQLite databases.

Can I use local models for paper filtering?

Yes, this skill supports Ollama for local, privacy-focused filtering of abstracts before proceeding with full PDF processing.

Can it cross-reference with external scientific databases?

Yes, it supports data enrichment through built-in API configurations for GBIF, NCBI, PubChem, GeoNames, and more.

How do I ensure the extracted data is scientifically accurate?

The skill includes a dedicated validation workflow that allows you to calculate precision, recall, and F1 metrics against a sample set of papers.

Scientific Data Extractor

Name: Scientific Data Extractor
Author: brunoasm

bybrunoasm

•

データサイエンスとML

Extracts and validates structured data from scientific PDF collections for systematic reviews and meta-analyses.

This skill automates the labor-intensive process of transforming collections of scientific research papers into structured, analyzable datasets. By leveraging Claude's vision capabilities and customizable JSON schemas, it enables researchers to extract complex metadata, repair data formats, and validate results against external databases like NCBI or GBIF. It is particularly valuable for systematic literature reviews where data integrity is paramount, providing built-in validation metrics like precision and recall to ensure high-quality outputs ready for statistical analysis in Python or R.

主な機能

01Schema-driven extraction using Claude's vision capabilities for complex PDF layouts

02Automatic JSON repair and validation against domain-specific external APIs

03Integrated filtering pipeline using local models (Ollama) or Claude (Haiku/Sonnet)

04Comprehensive quality assurance with precision, recall, and F1 score calculation

05Multi-format export supporting Python, R, CSV, Excel, and SQLite

063 GitHub stars

ユースケース

01Conducting systematic literature reviews that require large-scale data extraction from research papers

02Validating AI-extracted data against ground-truth datasets to ensure scientific accuracy

03Building curated databases from scientific publications for meta-analyses and database creation

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add brunoasm/my_claude_skills extract_from_pdfs

For use in Claude.ai and ChatGPT

Download Skill