how the text is divided into paragraphs

[wesamalnabki]

I'm trying to OCR the text in the image below. The OCR worked perfectly except that the order of the underlines words was wrong. So, the model detected the first line but missed "sus asesores." and in the second paragraph it detected the whole text except "internacional".
These two words were detected by as a standalone text blocks.

My question is why the model detected them as full sentences. how to change this behavior ?

[felixdittrich92]

Hi @wesamalnabki 👋
You could pass resolve_blocks=False to the ocr_predictor instance :)