PDF Extract Experimental Materials

Source: https://github.com/aipoch/medical-research-skills

Validation Shortcut

Run this minimal command first to verify the supported execution path:

python scripts/validate_skill.py --help

When to Use

• You need to build a structured inventory of reagents/antibodies/consumables from a paper's Materials and Methods section.
• A document includes a Key Resources Table and you want to convert it into clean CSV outputs.
• You want to extract instrument/equipment details (brand + model) from methods text or tables.
• You need to capture software and versions mentioned in methods (e.g., analysis pipelines, imaging software).
• You want to standardize reagent preparation instructions (buffers/solutions, concentrations, temperatures, steps) into a table.

Key Features

• Accepts PDF-derived Markdown/text as primary input; falls back to PDF text extraction when needed.
• Table-first parsing: prioritizes structured tables (e.g., Key Resources Table) before scanning prose sections.
• Extracts and outputs three normalized CSV tables with fixed schemas:
  • Main reagents (including antibodies/consumables/software versions)
  • Main instruments
  • Reagent preparation methods
• Uses section heading + keyword targeting to locate relevant regions and fill gaps.
• Enforces non-hallucination rules: unknown brand/model/catalog/version fields remain blank.

Dependencies

• Python 3.10+
• pdfplumber >= 0.10.0 (recommended for text-based PDFs)
• OCR engine (only for scanned PDFs), e.g.:
  • pytesseract >= 0.3.10
  • Tesseract OCR >= 5.0

Example Usage

1) Convert PDF to Markdown (preferred)

Reuse the existing script to generate parseable Markdown; only use OCR for scanned PDFs.

python d:\SKILL\project\pdf-extract\scripts\extract_pdf.py -i input.pdf -o out.md

2) Extract and write CSV outputs

After you have PDF-derived text/Markdown, extract the target fields and write exactly these three files:

• main_reagents.csv
• main_instruments.csv
• reagent_preparation.csv

CSV schemas (must match exactly):

main_reagents.csv

name,brand,catalog_number

main_instruments.csv

instrument,brand,model

reagent_preparation.csv

reagent,preparation_method

Implementation Details

1) Input decision logic

• If Markdown/text is available and structured, parse tables and sections directly.
• If only PDF is available:
  • Prefer pdfplumber for text-based PDFs.
  • Use OCR only when the PDF is scanned (image-only).

2) Target region detection (table-first)

Prioritize extraction in this order:

1. Key Resources Table (or any structured materials table)
2. Materials and Methods tables
3. Prose sections identified by headings/keywords, such as:
   • Key resources table, Materials and Methods, Reagents, Antibodies, Consumables, Software, Equipment, Instruments, Reagent preparation, Buffers, Solutions

When a table exists, parse it first; then scan prose to fill missing items.

3) Field extraction rules

A. Main reagents (name, brand, catalog_number)

• Include: reagents, antibodies, consumables, and software versions.
• Software mapping:
  • name = software name
  • brand = vendor/project/organization (if explicitly stated)
  • catalog_number = version/release (if explicitly stated)
• If brand or catalog_number is not present, leave it blank (do not infer).

B. Main instruments (instrument, brand, model)

• instrument = instrument/equipment name (e.g., "confocal microscope")
• brand = manufacturer/vendor (only if stated)
• model = model identifier string (often alphanumeric; only if stated)

C. Reagent preparation (reagent, preparation_method)

• reagent = buffer/solution/reagent being prepared
• preparation_method = preparation text, including any explicitly stated:
  • concentration, solvent, ratios, steps, incubation/temperature/time, pH adjustments

4) Output constraints and quality checks

• Write three CSV files with the exact column names shown above.
• Deduplicate entries by primary name (e.g., reagent name or instrument name) and keep the most complete row.
• Preserve original casing and punctuation from the source.
• Perform a spot check of at least three items against the PDF source to confirm correctness.

When Not to Use

• Do not use this skill when the required source data, identifiers, files, or credentials are missing.
• Do not use this skill when the user asks for fabricated results, unsupported claims, or out-of-scope conclusions.
• Do not use this skill when a simpler direct answer is more appropriate than the documented workflow.