Often patterns need to match "any token". A clean way to denote this
is with the empty dict {}: this sets no constraints on the token,
so should always match.
The problem was that having attributes length==0 was used as an
end-of-array signal, so the matcher didn't handle this case correctly.
This patch compiles empty token spec dicts into a constraint
NULL_ATTR==0. The NULL_ATTR attribute, 0, is always set to 0 on the
lexeme -- so this always matches.
The TextCategorizer class is supposed to support multi-label
text classification, and allow training data to contain missing
values.
For this to work, the gradient of the loss should be 0 when labels
are missing. Instead, there was no way to actually denote "missing"
in the GoldParse class, and so the TextCategorizer class treated
the label set within gold.cats as complete.
To fix this, we change GoldParse.cats to be a dict instead of a list.
The GoldParse.cats dict should map to floats, with 1. denoting
'present' and 0. denoting 'absent'. Gradients are zeroed for categories
absent from the gold.cats dict. A nice bonus is that you can also set
values between 0 and 1 for partial membership. You can also set numeric
values, if you're using a text classification model that uses an
appropriate loss function.
Unfortunately this is a breaking change; although the functionality
was only recently introduced and hasn't been properly documented
yet. I've updated the example script accordingly.
