extract-from-pdfs

Extract structured data from scientific PDFs using Claude's vision capabilities combined with configurable schemas, abstract filtering, and JSON repair. The workflow handles bulk PDF collections, applies your extraction schema to each document, validates outputs, and exports results to Python, R, Excel, CSV, or SQLite formats.

Yes. Filter papers by abstract using Claude or local models to focus extraction on relevant literature, reducing processing time and costs. This filtering step runs before full PDF extraction in the pipeline.

Organize metadata from BibTeX, RIS files, DOI lists, or directory structures. The pipeline reads these formats, standardizes the metadata, and uses it to coordinate extraction and enrichment across your PDF collection.

Validate and repair extracted JSON using schema validation and JSON repair functions, then enrich with external databases. The pipeline includes dedicated steps for JSON repair and API validation before final export.

Automate systematic review workflows by organizing metadata, filtering abstracts, extracting structured data from PDFs using Claude vision, repairing and validating outputs, and exporting to analysis-ready formats. This end-to-end pipeline reduces manual effort on large literature collections.

The workflow includes Python scripts for each stage—metadata organization, abstract filtering, PDF extraction, JSON repair, API validation, and database export—so you configure schemas and run steps sequentially rather than writing extraction logic from scratch.