LightOn adapts its OCR model to the Arabic language

LightOn unveils the adaptation of its OCR model, LightOnOCR-2, to the Arabic language. This advancement is based on targeted training using 12,000 synthetic pages produced with a modified version of their document generator. This model takes into account specific features such as right-to-left cursive writing, which often poses a problem in document automation.

The adaptation addresses the needs of the Middle East, strengthening LightOn's presence in the region. The model, offered as open source under the Apache 2.0 license, is central to LightOn Console, the company's self-service solution. LightOn provides guides for replicating this fine-tuning, thus extending its usefulness to other document contexts.