What does this implement/fix? Explain your changes.

In the test_kernel_theta function, s.rstrip('_bounds') is used to strip "_bounds" from the end of a string. This is almost certainly a misuse of the rstrip function, which strips off any number of the characters in the argument from the end, so, for example:

>>> 'some_string_bounds'.rstrip('_bounds')   # This one happens to do what you'd expect
'some_string'
>>> 'noun_unbound_bounds'.rstrip('_bounds')   # All characters from '_bounds' - oh no!
''
>>> 'noun_unbound_bounds'[0:-len('_bounds')] # This is what we expect
'noun_unbound'

Normally I would just do [s if not s.endswith('_bounds') else s[:-len('_bounds')] for s in args], but there's already a filter on args that only takes values that already end with _bounds.