Your regular expression catches very few penises. A very simple case:

    8------>

But you can easily solve that one. A much harder case would be catching stuff like this:

     _
    / \_________
                |\
                | |
       _________|/
    \_/

And even if you could do some image recognition, I think you'd have a hard time differentiating between stuff like this:

    O
   -|-
    |
   / \

And this:

    O
   -|-
    |_
   / \

...because what makes it a penis is not the shape, but the context, which is very hard to recognize.

Don't underestimate the difficulty of filtering human-generated data.

Also, I can't believe I just spent 5 minutes making ASCII penises.