Matthew Honnibal
e79fc41ff8
Merge pull request #1391 from explosion/feature/multilabel-textcat
...
💫 Fix multi-label support for text classification
2017-10-09 04:22:31 +02:00
Matthew Honnibal
6c79841c0d
Fix tests for history features
2017-10-09 04:12:24 +02:00
Matthew Honnibal
be4f0b6460
Update defaults
2017-10-08 02:08:12 -05:00
Matthew Honnibal
42b401d08b
Change default hidden depth to 1
2017-10-07 21:05:21 -05:00
Matthew Honnibal
9d66a915da
Update training defaults
2017-10-07 21:02:38 -05:00
Matthew Honnibal
d163115e91
Add non-linearity after history features
2017-10-07 21:00:43 -05:00
Matthew Honnibal
92c5d78b42
Unhack NER.add_action
2017-10-07 19:02:40 +02:00
Matthew Honnibal
f2b590f672
Increment version
2017-10-07 19:01:01 +02:00
Matthew Honnibal
eb0595bea9
Merge pull request #1392 from explosion/feature/parser-history-model
...
💫 Parser history features
2017-10-07 15:07:02 +02:00
ines
d70cf19158
Fix formatting
2017-10-07 15:06:38 +02:00
Ines Montani
36c68015f3
Merge pull request #1397 from explosion/feature/matcher-wildcard-token
...
💫 Allow empty dictionaries to match any token in Matcher
2017-10-07 15:05:24 +02:00
ines
c970b4f226
Add missing token attribute
2017-10-07 15:04:16 +02:00
ines
37f755897f
Update rule-based matching docs
2017-10-07 15:04:09 +02:00
Matthew Honnibal
3d22ccf495
Update default hyper-parameters
2017-10-07 07:16:41 -05:00
Matthew Honnibal
e22067e3b5
Document new hyper-parameters
2017-10-07 07:10:10 -05:00
Matthew Honnibal
09442d25ec
Merge remote-tracking branch 'origin/develop' into feature/parser-history-model
2017-10-07 07:05:04 -05:00
Matthew Honnibal
3b67eabfea
Allow empty dictionaries to match any token in Matcher
...
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.
2017-10-07 03:36:15 +02:00
Matthew Honnibal
8be46d766e
Remove print statement
2017-10-06 16:19:02 -05:00
ines
3468d535ad
Update model benchmarks
2017-10-06 21:39:06 +02:00
Matthew Honnibal
8e731009fe
Fix parser config serialization
2017-10-06 13:50:52 -05:00
Matthew Honnibal
f4c9a98166
Fix spacy evaluate command on non-GPU
2017-10-06 13:17:47 -05:00
Matthew Honnibal
16ba6aa8a6
Fix parser config serialization
2017-10-06 13:17:31 -05:00
ines
96a4e79d13
Fix PhraseMatcher example
2017-10-06 18:22:10 +02:00
Matthew Honnibal
c66399d8ae
Fix depth definition with history features
2017-10-06 06:20:05 -05:00
Matthew Honnibal
5c750a9c2f
Reserve 0 for 'missing' in history features
2017-10-06 06:10:13 -05:00
Matthew Honnibal
fbba7c517e
Pass dropout through to embed tables
2017-10-06 06:09:18 -05:00
Matthew Honnibal
21d11936fe
Fix significant train/test skew error in history feats
2017-10-06 06:08:50 -05:00
Matthew Honnibal
555d8c8bff
Fix beam history features
2017-10-05 22:21:50 -05:00
Matthew Honnibal
3db0a32fd6
Fix dropout for history features
2017-10-05 22:21:30 -05:00
Matthew Honnibal
b0618def8d
Add support for 2-token state option
2017-10-05 21:54:12 -05:00
Matthew Honnibal
363aa47b40
Clean up dead parsing code
2017-10-05 21:53:49 -05:00
Matthew Honnibal
ca12764772
Enable history features for beam parser
2017-10-05 21:53:29 -05:00
Matthew Honnibal
fc06b0a333
Fix training when hist_size==0
2017-10-05 21:52:28 -05:00
Matthew Honnibal
e25ffcb11f
Move history size under feature flags
2017-10-05 19:38:13 -05:00
Matthew Honnibal
563f46f026
Fix multi-label support for text classification
...
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.
2017-10-05 18:43:02 -05:00
Matthew Honnibal
c36d4596bf
Merge branch 'develop' of https://github.com/explosion/spaCy into develop
2017-10-05 18:27:56 +02:00
Matthew Honnibal
056b08c0df
Delete obsolete nn_text_class example
2017-10-05 18:27:10 +02:00
Matthew Honnibal
c6cd81f192
Wrap try/except around model saving
2017-10-05 08:14:24 -05:00
Matthew Honnibal
5743b06e36
Wrap model saving in try/except
2017-10-05 08:12:50 -05:00
Matthew Honnibal
fd4baff475
Update tests
2017-10-05 08:12:27 -05:00
Matthew Honnibal
dcdfa071aa
Disable LayerNorm hack
2017-10-04 20:06:52 -05:00
Matthew Honnibal
943af4423a
Make depth setting in parser work again
2017-10-04 20:06:05 -05:00
Matthew Honnibal
bfabc333be
Merge remote-tracking branch 'origin/develop' into feature/parser-history-model
2017-10-04 20:00:36 -05:00
Matthew Honnibal
92066b04d6
Fix Embed and HistoryFeatures
2017-10-04 19:55:34 -05:00
ines
b621a2e964
Fix build emoji
2017-10-04 18:37:27 +02:00
Matthew Honnibal
5560c46a59
Update buildkite
2017-10-04 18:29:41 +02:00
Matthew Honnibal
e3c93f87a4
Update sdist
2017-10-04 18:18:07 +02:00
Matthew Honnibal
c4c7def9ce
Fix yml
2017-10-04 18:14:33 +02:00
Matthew Honnibal
71825f9737
Fix yml
2017-10-04 18:12:16 +02:00
Matthew Honnibal
6304c5e146
Fix yml
2017-10-04 18:08:34 +02:00
