You can't. No software that scans pages and converts them to text is certified 100 per cent accurate. At best, such software, called OCR (Optical Character Reader) highlight possible mistakes and prompt the user for corrections. You can reduce errors by using more advanced, and therefore more expensive, OCRs like OmniScan. An alternative approach is to use Adobe Acrobat, with its Capture utility to save the scanned page as a PDF and to automatically hide the translated text 'behind' the scanned image so the 'find' or 'search' works.