Tesseract OCR in PHP

I am wondering one thing. Here is a question- [login to view URL] You will see an image in the StackOverflow question. If I put the image in Tesseract running on my local machine and the [login to view URL] site, both gives the text "Add 536 to 592", that is wrong and this should be "Add 686 to 692 . The [login to view URL] site gives the right answer. For the convenience I have also attached the image here. All the referenced sites use Tesseract. So I guess somehow we need to set focus o training. So we want someone expert in Tesseract. Thanks.