Xberg for document extraction across languages

How it works

The project provides a shared extraction core with bindings for application runtimes. Optional OCR and format-specific extractors handle cases where text is absent or embedded differently.

What you need

Choose a supported language binding and enable the extraction features your formats require. Test the deployment package with the optional assets it will need at runtime.

What to try first

Convert a normal PDF and an image-only PDF in the binding your application uses. Confirm what metadata and error details survive the binding boundary.

Include an encrypted or damaged input in a fixture set. Applications should distinguish a handled extraction error from a successful conversion that happened to return empty text.

Where to look for failures

Different formats need different extraction paths. Review optional runtime assets and compare results for malformed documents before opening an upload endpoint.

Related Franklin pages

Compare converted output on a few documents with different structures. Choose according to the information downstream consumers need to recover, such as tables, source locations or heading boundaries.

• Extract document structure, reading order and tables for search or analysis. Docling for structured document conversion
• Turn documents into typed elements that a retrieval pipeline can filter and chunk. Unstructured for partitioning source documents

For how recognizing characters fits into the larger document processing workflow, see Franklin’s glossary. Optical character recognition

Project documentation

• Current project README
• README at the reviewed revision