Are you a
Builder?
Do you think your approach to natural language processing problems can
withstand attacks?
Is it
robust to unexpected inputs, or domain mismatch?
Will a
rule-based approach hold up better than a
learning-based approach?
Participate in our shared tasks by solving an NLP problem—by rules or by learning—and learn how your approach stacks up against others on
adversarially, linguistically constructed inputs.
Or maybe you're a
Breaker? Click to flip...
Are you a
Breaker?
Are you sick of those natural language processing folks saying that they've
solved language problems, when you know their models are too naive?
Do you think you can
construct examples that completely fool NLP technology, or poke at their
biases?
Participate in our shared tasks by constructing
linguistically-motivated examples that will defeat the best technology out there.
Help expose holes in existing techniques and
drive future research.
Or maybe you're a
Builder? Click to flip...
Build It, Break It
