How to determine the correct block coordinates for text recognition in a picture?

Question

Text recognising

Good day to all.
There is such a task. An image of the scanned text is given. For the image, the coordinates of some areas that need to be recognized are indicated. These areas may not be accurate.
Is there any algorithm that allows you to find a more correct location for each block? Those. use not the exact coordinates given earlier, but apply some "tolerances"?
Tesseract OCR is used for text recognition.

Answer 1

" ABBYY FlexiLayout Studio
ABBYY FlexiLayout Studio – мощный инструмент по созданию гибких описаний для документов с нежесткой структурой, то есть тех документов, на которых расположение однотипных данных может варьироваться от одного экземпляра к другому. Для гибких описаний в интерфейсе программы принято название FlexiLayout. Гибкое описание FlexiLayout позволяет системе извлекать из документов с нежесткой структурой необходимые данные, как то: текстовые строки, штрихкоды, даты, валютные символы, числа, разделители, таблицы.
For example, using only one flexible description, the program can find all the necessary fields on invoices coming from different suppliers and having different formats, even if they are multi-page documents. ABBYY FlexiLayout Studio offers efficient and convenient tools for testing and customizing FlexiLayouts on a set of document images: hypothesis tree, reference descriptions, FlexiLayout language.
The system self-learning technologies implemented in ABBYY FlexiCapture allow not only to automate the process of creating document descriptions, but also enable specialists to edit such descriptions immediately after they are created to improve the result of the system in the future. This significantly reduces the complexity of this stage of work and its duration.
Maybe you should change the recognition tool?