The recognition result of fields in a form comes up as separate blocks of text.
This can happen when there are large gaps between characters in a frame.
The best way to extract the text in the boxes as one word is to manually specify the recognition areas using the AnalyzeRegion method. In order to do that, create a Region object and then use the AddRegion method to add all the required text blocks. For the text in frames, create Region objects in such a way that they are inside the text frame. Visually, it should look something like this: