With a research team based in Paris,  CALFA  is developing text recognition for languages that are not widely covered by other tools.

For research needs, Calfa is providing a model training service able to reach up to 99% of recognition accuracy on the most difficult languages and manuscripts.

As an alternative to Transkribus or eScriptorium, Calfa is offering :

• High accuracy results especially on rare scripts, alphabets, syllabaries
• A data creation/transcription interface designed for non-western and right-to-left writings
• A fast training, with a limited volume of transcription to provide
• A custom training support from experts

Calfa’s custom models have been used to process documents in Ancient Greek, Ethiopian, Hebrew, or Syriac. Calfa is also providing large, generic recognition models, which can be used for Arabic and Armenian with no training.

This presentation will be an opportunity for the public to discover some research projects using Calfa OCR for text recognition, to present the offers dedicated to research professionals, and to give an overview of the model training service, including a guided tour of the interface.