I don't know if the Tess4J wrapper supports multiple languages, or which language you're using as a base language, but you might consider training it and any other symbols you need into an entirely separate language and then OCRing using the Tess4J equivalent of -l eng+mylang (or whatever your base language is).
There is the equation "language" with code equ, but it apparently doesn't include that style division sign, which I was a little surprised at. You might try -l eng+equ to start though and see perhaps the "wavy division sign" or some other symbol is close enough that it gets detected reliable in place of the plus sign, before going to the trouble of doing your own training.
I've attached the characters which are included in the equ training.
Tom