chem-data-extractor

Chemistry Data Extractor | 化学数据提取器

Extract structured chemical characterization data from chemistry supplementary materials and return in strict JSON format. 从化学论文补充材料中提取结构化表征数据，以严格JSON格式返回。

Supported Data Fields

• compound_name: Full IUPAC or common name (including stereochemistry if given)
• structure_image_description: Brief description of the molecular structure
• physical_state: e.g., "white solid", "colorless oil"
• mass_obtained: in mg
• yield_percent: as number only
• melting_point_range: in °C, as string like "126.6–127.3"
• rf_value: Rf value and solvent system
• optical_rotation: [α]D²⁵ value, concentration, solvent
• hplc_conditions: column, mobile phase, flow rate, wavelength, retention times (major/minor), ee%
• nmr_1H: frequency, solvent, chemical shifts with multiplicity and coupling constants
• nmr_13C: frequency, solvent, chemical shifts with notes (e.g., d, JCF)
• nmr_19F: frequency, solvent, chemical shift
• hrms_data: ion type, calculated m/z, found m/z, formula
• racemic_sample_note: if mentioned

Workflow

Step 1: Ask User for Extraction Mode

When user asks to extract chemistry data, first ask:

Do you want to extract data for:

1. A specific compound (provide compound ID like "3i" or "1a")
2. All compounds in a single document
3. Batch process multiple PDF files (creates folder for each)

---

Mode 1: Batch Process Multiple PDFs

For processing multiple PDF files at once. Creates a separate folder for each PDF with extracted compounds.

Usage

python scripts/batch_extract.py \
    /path/to/pdf_folder \
    -o ./output_folder

Options

• -o, --output: Output base directory (default: ./chem_extract_output)
• --keep-md: Keep intermediate markdown files (default: cleanup after extraction)
• --skip-existing: Skip PDFs that already have output folders

Output Structure

output_folder/
├── batch_summary.json          # Overall summary of all processed PDFs
├── paper1/
│   ├── compounds.json          # All extracted compounds
│   └── summary.json            # Brief summary with compound list
├── paper2/
│   ├── compounds.json
│   └── summary.json
└── ...