Make function and compiled DLL (with source) to reliably capture English text at a given position on the screen using OCR. Should return word at the point, and words on either side. Running time must be less than half a second on a Pentium and compiled size under 500K.
I suggest adapting Tesseract, available from
[url removed, login to view]
(remove TIFF support etc, tweak parameters if neccessary)