Commit Graph

362 Commits

Author SHA1 Message Date
Ines Montani
cae4457c38 💫 Add .similarity warnings for no vectors and option to exclude warnings (#2197)
* Add logic to filter out warning IDs via environment variable

Usage: SPACY_WARNING_EXCLUDE=W001,W007

* Add warnings for empty vectors

* Add warning if no word vectors are used in .similarity methods

For example, if only tensors are available in small models – should hopefully clear up some confusion around this

* Capture warnings in tests

* Rename SPACY_WARNING_EXCLUDE to SPACY_WARNING_IGNORE
2018-05-21 01:22:38 +02:00
ines
5401c55c75 Merge branch 'master' into develop 2018-05-20 16:49:40 +02:00
ines
b59e3b157f Don't require attrs argument in Doc.retokenize and allow both ints and unicode (resolves #2304) 2018-05-20 15:15:37 +02:00
Matthew Honnibal
546dd99cdf Merge master into develop -- mostly Arabic and website 2018-05-15 18:14:28 +02:00
Douglas Knox
9b49a40f4e Test and fix for Issue #2219 (#2272)
Test and fix for Issue #2219: Token.similarity() failed if single letter
2018-05-03 18:40:46 +02:00
Matthew Honnibal
9d147e12c4 Merge remote-tracking branch 'origin/master' into develop 2018-05-01 18:18:51 +02:00
Mr Roboto
6f5ccda19c Addresses Issue #2228 - Deserialization fails when using tensor=False or sentiment=False (#2230)
* Fixes issue #2228

* Adds a new contributor
2018-05-01 13:40:22 +02:00
Matthew Honnibal
a8bc947fd4 Fix Token.set_extension 2018-04-29 15:48:19 +02:00
Matthew Honnibal
2c4a6d66fa Merge master into develop. Big merge, many conflicts -- need to review 2018-04-29 14:49:26 +02:00
ines
1c6d77610c Add remove_extension method on Doc, Token and Span (closes #2242) 2018-04-28 23:33:09 +02:00
ines
9632595fb4 Use correct, non-deprecated merge syntax (resolves #2226) 2018-04-18 18:28:28 -04:00
Suraj Rajan
5957f15227 Fixed typos for #2222,#2223 (#2233) (closes #2222, closes #2223) 2018-04-18 14:55:26 -07:00
Xiaoquan Kong
e2f13ec722 bugfix: Doc.noun_chunks call Doc.noun_chunks_iterator without checking (closes #2194) 2018-04-08 23:44:05 +02:00
ines
e5f47cd82d Update errors 2018-04-03 21:40:29 +02:00
ines
62b4b527d7 Don't raise error if set_extension has getter and setter (closes #2177)
Improve error messages, raise error if setter is specified without a getter and compare against _unset to allow default=None. Also add more tests.
2018-04-03 18:30:17 +02:00
ines
ee3082ad29 Fix whitespace 2018-04-03 18:29:53 +02:00
Ines Montani
3141e04822
💫 New system for error messages and warnings (#2163)
* Add spacy.errors module

* Update deprecation and user warnings

* Replace errors and asserts with new error message system

* Remove redundant asserts

* Fix whitespace

* Add messages for print/util.prints statements

* Fix typo

* Fix typos

* Move CLI messages to spacy.cli._messages

* Add decorator to display error code with message

An implementation like this is nice because it only modifies the string when it's retrieved from the containing class – so we don't have to worry about manipulating tracebacks etc.

* Remove unused link in spacy.about

* Update errors for invalid pipeline components

* Improve error for unknown factories

* Add displaCy warnings

* Update formatting consistency

* Move error message to spacy.errors

* Update errors and check if doc returned by component is None
2018-04-03 15:50:31 +02:00
Matthew Honnibal
abf8b16d71
Add doc.retokenize() context manager (#2172)
This patch takes a step towards #1487 by introducing the
doc.retokenize() context manager, to handle merging spans, and soon
splitting tokens.

The idea is to do merging and splitting like this:

with doc.retokenize() as retokenizer:
    for start, end, label in matches:
        retokenizer.merge(doc[start : end], attrs={'ent_type': label})

The retokenizer accumulates the merge requests, and applies them
together at the end of the block. This will allow retokenization to be
more efficient, and much less error prone.

A retokenizer.split() function will then be added, to handle splitting a
single token into multiple tokens. These methods take `Span` and `Token`
objects; if the user wants to go directly from offsets, they can append
to the .merges and .splits lists on the retokenizer.

The doc.merge() method's behaviour remains unchanged, so this patch
should be 100% backwards incompatible (modulo bugs). Internally,
doc.merge() fixes up the arguments (to handle the various deprecated styles),
opens the retokenizer, and makes the single merge.

We can later start making deprecation warnings on direct calls to doc.merge(),
to migrate people to use of the retokenize context manager.
2018-04-03 14:10:35 +02:00
Matthew Honnibal
0b375d50c8 Fix ent_iob tags in doc.merge to avoid inconsistent sequences 2018-03-28 18:39:03 +02:00
Matthew Honnibal
e807f88410 Resolve merge when cherry-picking ent iob patches from develop 2018-03-28 18:38:13 +02:00
Matthew Honnibal
99fbc7db33 Improve error message when entity sequence is inconsistent 2018-03-28 18:36:53 +02:00
ines
9e83513004 Add position of invalid token to error message 2018-03-27 23:56:59 +02:00
ines
693971dd8f Improve error message if token text is empty string (see #2101) 2018-03-27 22:25:40 +02:00
ines
0c829e6605 Fix whitespace 2018-03-27 22:20:59 +02:00
Matthew Honnibal
de9fd091ac Fix #2014: token.pos_ not writeable 2018-03-27 21:21:11 +02:00
Matthew Honnibal
1f7229f40f Revert "Merge branch 'develop' of https://github.com/explosion/spaCy into develop"
This reverts commit c9ba3d3c2d, reversing
changes made to 92c26a35d4.
2018-03-27 19:23:02 +02:00
Matthew Honnibal
d2118792e7 Merge changes from master 2018-03-27 13:38:41 +02:00
Matthew Honnibal
63a267b34d Fix #2073: Token.set_extension not working 2018-03-27 13:36:20 +02:00
Matthew Honnibal
a3d0cb15d3 Fix ent_iob tags in doc.merge to avoid inconsistent sequences 2018-03-26 07:16:06 +02:00
Matthew Honnibal
514d89a3ae Set missing label for non-specified entities when setting doc.ents 2018-03-26 07:14:16 +02:00
Matthew Honnibal
54d7a1c916 Improve error message when entity sequence is inconsistent 2018-03-26 07:13:34 +02:00
Matthew Honnibal
8e08c378fe Fix entity IOB and tag in span merging 2018-03-25 22:16:01 +02:00
Matthew Honnibal
bede11b67c
Improve label management in parser and NER (#2108)
This patch does a few smallish things that tighten up the training workflow a little, and allow memory use during training to be reduced by letting the GoldCorpus stream data properly.

Previously, the parser and entity recognizer read and saved labels as lists, with extra labels noted separately. Lists were used becaue ordering is very important, to ensure that the label-to-class mapping is stable.

We now manage labels as nested dictionaries, first keyed by the action, and then keyed by the label. Values are frequencies. The trick is, how do we save new labels? We need to make sure we iterate over these in the same order they're added. Otherwise, we'll get different class IDs, and the model's predictions won't make sense.

To allow stable sorting, we map the new labels to negative values. If we have two new labels, they'll be noted as having "frequency" -1 and -2. The next new label will then have "frequency" -3. When we sort by (frequency, label), we then get a stable sort.

Storing frequencies then allows us to make the next nice improvement. Previously we had to iterate over the whole training set, to pre-process it for the deprojectivisation. This led to storing the whole training set in memory. This was most of the required memory during training.

To prevent this, we now store the frequencies as we stream in the data, and deprojectivize as we go. Once we've built the frequencies, we can then apply a frequency cut-off when we decide how many classes to make.

Finally, to allow proper data streaming, we also have to have some way of shuffling the iterator. This is awkward if the training files have multiple documents in them. To solve this, the GoldCorpus class now writes the training data to disk in msgpack files, one per document. We can then shuffle the data by shuffling the paths.

This is a squash merge, as I made a lot of very small commits. Individual commit messages below.

* Simplify label management for TransitionSystem and its subclasses

* Fix serialization for new label handling format in parser

* Simplify and improve GoldCorpus class. Reduce memory use, write to temp dir

* Set actions in transition system

* Require thinc 6.11.1.dev4

* Fix error in parser init

* Add unicode declaration

* Fix unicode declaration

* Update textcat test

* Try to get model training on less memory

* Print json loc for now

* Try rapidjson to reduce memory use

* Remove rapidjson requirement

* Try rapidjson for reduced mem usage

* Handle None heads when projectivising

* Stream json docs

* Fix train script

* Handle projectivity in GoldParse

* Fix projectivity handling

* Add minibatch_by_words util from ud_train

* Minibatch by number of words in spacy.cli.train

* Move minibatch_by_words util to spacy.util

* Fix label handling

* More hacking at label management in parser

* Fix encoding in msgpack serialization in GoldParse

* Adjust batch sizes in parser training

* Fix minibatch_by_words

* Add merge_subtokens function to pipeline.pyx

* Register merge_subtokens factory

* Restore use of msgpack tmp directory

* Use minibatch-by-words in train

* Handle retokenization in scorer

* Change back-off approach for missing labels. Use 'dep' label

* Update NER for new label management

* Set NER tags for over-segmented words

* Fix label alignment in gold

* Fix label back-off for infrequent labels

* Fix int type in labels dict key

* Fix int type in labels dict key

* Update feature definition for 8 feature set

* Update ud-train script for new label stuff

* Fix json streamer

* Print the line number if conll eval fails

* Update children and sentence boundaries after deprojectivisation

* Export set_children_from_heads from doc.pxd

* Render parses during UD training

* Remove print statement

* Require thinc 6.11.1.dev6. Try adding wheel as install_requires

* Set different dev version, to flush pip cache

* Update thinc version

* Update GoldCorpus docs

* Remove print statements

* Fix formatting and links [ci skip]
2018-03-19 02:58:08 +01:00
Thomas Opsomer
fbf48b3f9f lemma property to return hash instead of unicode 2018-03-14 17:03:00 +01:00
Matthew Honnibal
a1be01185c Fix array out of bounds error in Span 2018-02-28 12:27:09 +01:00
Thomas Opsomer
8df9e52829 lemma property to return hash instead of unicode 2018-02-27 19:50:01 +01:00
Matthew Honnibal
cf0e320f2b Add doc.is_sentenced attribute, re #1959 2018-02-18 14:16:55 +01:00
Matthew Honnibal
1e5aeb4eec
Merge pull request #1987 from thomasopsomer/span-sent
Make span.sent work when only manual / custom sbd
2018-02-18 14:05:37 +01:00
Thomas Opsomer
deab391cbf correct check on sent_start & raise if no boundaries 2018-02-15 16:58:30 +01:00
Thomas Opsomer
b902731313 Find span sentence when only sentence boundaries (no parser) 2018-02-14 22:18:54 +01:00
4altinok
ca8728035d added new lex feat to token 2018-02-11 18:55:48 +01:00
Thomas Opsomer
515e25910e fix sent_start in serialization 2018-01-28 19:50:42 +01:00
Matthew Honnibal
56164ab688 Set l_edge and r_edge correctly for non-projective parses. Fixes #1799 2018-01-22 20:18:04 +01:00
Matthew Honnibal
ccb51a9f36 Make .similarity() return 1.0 if all orth attrs match 2018-01-15 16:29:48 +01:00
Matthew Honnibal
b904d81e9a Fix rich comparison against None objects. Closes #1757 2018-01-15 15:51:25 +01:00
Matthew Honnibal
ab7c45b12d Fix error message and handling of doc.sents 2018-01-15 15:21:11 +01:00
Matthew Honnibal
465a6f6452 Add missing Span.vocab property. Closes #1633 2018-01-14 15:06:30 +01:00
Matthew Honnibal
0cb090e526 Fix infinite recursion in token.sent_start. Closes #1640 2018-01-14 15:02:15 +01:00
Matthew Honnibal
5cbe913b6f Don't raise deprecation warning in property. Closes #1813, #1712 2018-01-14 14:55:58 +01:00
Matthew Honnibal
e10e9ad2c5 Improve efficiency of Doc.to_array 2017-11-23 12:33:27 +00:00