explosion/spaCy
1
1
Fork 0
mirror of https://github.com/explosion/spaCy.git synced 2024-11-11 04:08:09 +03:00
Code Issues Packages Projects Releases Wiki Activity
9,851 Commits 68 Branches 175 Tags 279 MiB
d811c97da1
Commit Graph

430 Commits

Author SHA1 Message Date
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
Matthew Honnibal
fa62427300 Remove lookup-based lemmatization 2017-11-23 12:32:22 +00:00
Matthew Honnibal
fb26b2cb12 Use lookup lemmatizer if lemma unset 2017-11-23 12:31:58 +00:00
Burton DeWilde
a5c6869b2d Fix bug where span.orth_ != span.text (see #1612) 2017-11-20 12:05:43 -06:00
Motoki Wu
a52e195a0a Fixes Issue #1207 where noun_chunks of Span gives an error. 
Make sure to reference `self.doc` when getting the noun chunks.

Same fix as 9750a0128c
 2017-11-17 17:16:20 -08:00
ines
1c218397f6 Ensure path in Doc.to_disk/from_disk (resolves ##1521) 
Also add Doc serialization tests with both Path and string path options
 2017-11-09 02:29:03 +01:00
Matthew Honnibal
144a93c2a5 Back-off to tensor for similarity if no vectors 2017-11-03 20:56:33 +01:00
Matthew Honnibal
62ed58935a Add Doc.extend_tensor() method 2017-11-03 11:20:31 +01:00
ines
9659391944 Update deprecated methods and add warnings 2017-11-01 16:49:42 +01:00
ines
705a4e3e4a Fix formatting 2017-11-01 16:44:08 +01:00
Matthew Honnibal
9e0ebee81c Add Token.is_sent_start property, so can deprecate Token.sent_start 2017-11-01 13:27:14 +01:00
Matthew Honnibal
7e7116cdf7 Fix Doc.to_array when only one string attr provided 2017-11-01 13:26:43 +01:00
Matthew Honnibal
301fb2bb60 Implement Span.n_lefts and Span.n_rights 2017-11-01 13:25:12 +01:00
Matthew Honnibal
86eba61fae Fix token.vector when vectors are missing 2017-11-01 00:47:35 +01:00
ines
d96e72f656 Tidy up rest 2017-10-27 21:07:59 +02:00
ines
d2df81d907 Fix not implemented Span getters 2017-10-27 18:09:28 +02:00
ines
544a407b93 Tidy up Doc, Token and Span and add missing docs 2017-10-27 17:07:26 +02:00
ines
6a0483b7aa Tidy up and document Doc, Token and Span 2017-10-27 15:41:45 +02:00
ines
1a559d4c95 Remove old, unused file 2017-10-27 15:34:35 +02:00
ines
ea4a41c8fb Tidy up util and helpers 2017-10-27 14:39:09 +02:00
Matthew Honnibal
b66b8f028b Fix #1375 -- out-of-bounds on token.nbor() 2017-10-24 12:10:39 +02:00
Matthew Honnibal
ccd2ab1a62 Merge pull request #1443 from ramananbalakrishnan/develop-get-lca-matrix 
Add LCA matrix for spans and docs
 2017-10-24 11:22:46 +02:00
Matthew Honnibal
fdf25d10ba Merge pull request #1440 from ramananbalakrishnan/develop 
Support single value for attribute list in doc.to_array
 2017-10-24 10:23:12 +02:00
ines
a31f048b4d Fix formatting 2017-10-23 10:38:06 +02:00
Ramanan Balakrishnan
d2fe56a577
Add LCA matrix for spans and docs 2017-10-20 23:58:00 +05:30
Ramanan Balakrishnan
0726946563
cleanup to_array implementation using fixes on master 2017-10-20 17:09:37 +05:30
Ramanan Balakrishnan
b3ab124fc5
Support strings for attribute list in doc.to_array 2017-10-20 11:46:57 +05:30
Ramanan Balakrishnan
7b9b1be44c
Support single value for attribute list in doc.to_array 2017-10-19 17:00:41 +05:30
Matthew Honnibal
394633efce Make doc pickling support hooks 2017-10-17 19:44:09 +02:00
Matthew Honnibal
cdb0c426d8 Improve deserialization of user_data, esp. for Underscore 2017-10-17 19:29:20 +02:00
Matthew Honnibal
32a8564c79 Fix doc pickling 2017-10-17 18:20:24 +02:00
Matthew Honnibal
92c1eb2d6f Fix Doc pickling. This also removes need for Binder class 2017-10-17 16:11:13 +02:00
Matthew Honnibal
a002264fec Remove caching of Token in Doc, as caused cycle. 2017-10-16 19:34:21 +02:00
Matthew Honnibal
59c216196c Allow weakrefs on Doc objects 2017-10-16 19:22:11 +02:00
ines
e0ff145a8b Merge branch 'develop' into feature/dot-underscore 2017-10-11 11:57:05 +02:00
Matthew Honnibal
3b527fa52b Call morphology.assign_untagged when pushing token to Doc 2017-10-11 03:23:57 +02:00
Matthew Honnibal
e0a9b02b67 Merge Span._ and Span.as_doc methods 2017-10-09 22:00:15 -05:00
ines
3fc4fe61d2 Fix typo 2017-10-10 04:15:14 +02:00
ines
59c4f27499 Add get, set and has methods to Underscore 2017-10-10 04:14:35 +02:00
Matthew Honnibal
51d18937af Partially apply doc/span/token into method 
We want methods to act like they're "bound" to the object, so that you can make your method conditional on the `doc`, `span` or `token` instance --- like, well, a method. We therefore partially apply the function, which works like this:

```
def partial(unbound_method, constant_arg):
    def bound_method(*args, **kwargs):
        return unbound_method(constant_arg, *args, **kwargs)
    return bound_method
 2017-10-10 02:21:28 +02:00
Matthew Honnibal
e938bce320 Adjust parsing transition system to allow preset sentence segments. 2017-10-08 23:53:34 +02:00
Matthew Honnibal
080afd4924 Add ternary value setting to Token.sent_start 2017-10-08 23:51:58 +02:00
Matthew Honnibal
7ae67ec6a1 Add Span.as_doc method 2017-10-08 23:50:20 +02:00
Matthew Honnibal
668a0ea640 Pass extensions into Underscore class 2017-10-07 18:56:01 +02:00
Matthew Honnibal
1289129fd9 Add Underscore class 2017-10-07 18:00:14 +02:00
Matthew Honnibal
9bfd585a11 Fix parameter name in .pxd file 2017-09-26 07:28:50 -05:00
ines
2480f8f521 Add missing return in Doc.from_disk() (closes #1330) 2017-09-18 15:32:00 +02:00
Matthew Honnibal
03b5b9727a Fix Doc.vector for empty doc objects 2017-08-22 19:52:19 +02:00
Matthew Honnibal
0551b7b03a Fix doc.vector 2017-08-22 19:46:52 +02:00
Matthew Honnibal
d55d6e1cfa Fix comparison of Token from different docs. Closes #1257 2017-08-19 16:39:32 +02:00
Matthew Honnibal
dea229c634 Fix Span.to_array method 2017-08-19 16:24:28 +02:00
Matthew Honnibal
8b7ac77c23 Allow span label to be string in Doc.char_span 2017-08-19 16:18:09 +02:00
Matthew Honnibal
80236116a6 Add Doc.char_span method, to get a span by character offset 2017-08-19 12:21:09 +02:00
Matthew Honnibal
482bba1722 Add Span.to_array method 2017-08-19 12:20:45 +02:00
Matthew Honnibal
a6a2159969 Add slot for text categories to Doc 2017-07-22 00:34:15 +02:00
Matthew Honnibal
2a3bd5ee90 Fix fetching of noun chunk iterator 2017-06-04 15:53:05 -05:00
Matthew Honnibal
92ae36f84e Improve way noun chunks iterator is looked up 2017-06-04 21:53:39 +02:00
Matthew Honnibal
675f448313 Fix vector linkage on Doc 2017-06-04 14:25:30 -05:00
Matthew Honnibal
f4662e9218 Fix vector linkage for token 2017-06-04 14:19:58 -05:00
ines
459a1e8470 Fix whitespace 2017-06-03 11:31:18 +02:00
ines
5109bba910 Port over fix from #1070 2017-06-03 11:31:11 +02:00
Matthew Honnibal
498ad85309 Try using tensor for vector/similarity methdos 2017-05-30 23:35:17 +02:00
Matthew Honnibal
4ddff020c3 Fix compile error 2017-05-28 23:30:40 +02:00
Matthew Honnibal
6d3caeadd2 Fix type check for long 2017-05-28 23:22:45 +02:00
Matthew Honnibal
7996d21717 Fixes for new StringStore 2017-05-28 11:09:27 -05:00
Matthew Honnibal
fe11564b8e Finish stringstore change. Also xfail vectors tests 2017-05-28 15:10:22 +02:00
Matthew Honnibal
84e66ca6d4 WIP on stringstore change. 27 failures 2017-05-28 14:06:40 +02:00
Matthew Honnibal
39293ab2ee Merge branch 'develop' of https://github.com/explosion/spaCy into develop 2017-05-28 11:46:57 +02:00
Matthew Honnibal
2445707f3c Re-delegate vectors to vocab 2017-05-28 11:46:10 +02:00
ines
66088851dc Add Doc.to_disk() and Doc.from_disk() methods 2017-05-24 11:58:17 +02:00
Matthew Honnibal
d44b1eafc4 Fix conflict artefacts 2017-05-23 18:47:11 +02:00
Matthew Honnibal
01e59e4e6e * Add Token.sent_start property, re Issue #235 2017-05-23 18:41:11 +02:00
Matthew Honnibal
d68dd1f251 Add SENT_START attribute, for custom sentence boundary detection 2017-05-23 18:37:58 +02:00
ines
7ed8a92ed1 Update docstrings and API docs for Token 2017-05-20 15:13:33 +02:00
ines
a804045597 Use is_ancestor instead of deprecated is_ancestor_of 2017-05-19 20:23:40 +02:00
ines
e9e62b01b0 Update docstrings and API docs for Token 2017-05-19 18:47:56 +02:00
ines
62ceec4fc6 Update docstrings and API docs for Span 2017-05-19 18:47:46 +02:00
ines
23f9a3ccc8 Update docstrings and API docs for Doc 2017-05-19 18:47:39 +02:00
ines
0791f0aae6 Update docstrings and API docs for Span class 2017-05-19 00:31:31 +02:00
ines
8455cb1327 Update docstring for Doc.__getitem__ 2017-05-19 00:30:51 +02:00
ines
b687ad109d Update docstrings and API docs for Doc class 2017-05-18 23:59:44 +02:00
ines
593361ee3c Update docstrings for Span class 2017-05-18 22:17:41 +02:00
ines
b87066ff10 Update docstrings and API docs for Doc class 2017-05-18 22:17:41 +02:00
Matthew Honnibal
4b9d69f428 Merge branch 'v2' into develop 
* Move v2 parser into nn_parser.pyx
* New TokenVectorEncoder class in pipeline.pyx
* New spacy/_ml.py module

Currently the two parsers live side-by-side, until we figure out how to
organize them.
 2017-05-14 01:10:23 +02:00
ines
9d85cda8e4 Fix models error message and use about.__docs_models__ (see #1051) 2017-05-13 13:05:47 +02:00
ines
6b942763f0 Tidy up imports 2017-05-13 13:04:40 +02:00
ines
6129016e15 Replace deepcopy 2017-05-13 12:32:37 +02:00
ines
df68bf45ce Set defaults for light and flat kwargs 2017-05-13 12:32:23 +02:00
ines
b9dea345e5 Remove old import 2017-05-13 12:32:11 +02:00
ines
293ee359c5 Fix formatting 2017-05-13 12:32:06 +02:00
Matthew Honnibal
ee1d35bdb0 Fix merge conflict 2017-05-13 03:20:19 +02:00
Matthew Honnibal
b2540d2379 Merge Kengz's tree_print patch 2017-05-13 03:18:49 +02:00
Matthew Honnibal
4efb391994 Fix serializer 2017-05-09 18:45:18 +02:00
Matthew Honnibal
1166b0c491 Implement Doc.to_bytes and Doc.from_bytes methods 2017-05-09 18:11:34 +02:00
Matthew Honnibal
9e167b7bb6 Strip serializer from code 2017-05-09 17:28:50 +02:00
Matthew Honnibal
62ecdea9f2 Add binder class for document serialization 2017-05-09 17:21:00 +02:00
Matthew Honnibal
6782eedf9b Tmp GPU code 2017-05-07 11:04:24 -05:00
Matthew Honnibal
4d98511db7 Make Span hashable. Closes #1019 2017-04-26 19:01:05 +02:00
Matthew Honnibal
6a4221a6de Allow lemma to be set from Python. Re #973 2017-04-16 18:07:53 +02:00
ines
0739ae7b76 Tidy up and fix formatting and imports 2017-04-15 13:05:15 +02:00
ines
3b667a24d4 Remove whitespace 2017-04-01 10:21:08 +02:00
ines
e71a1f4bd0 Fix download commands in error messages (see #946) 2017-04-01 10:20:57 +02:00
Matthew Honnibal
51882ee2b8 Fix check for setting ent_id in merge 2017-03-31 19:32:01 +02:00
Matthew Honnibal
fc3900e5b2 Allow ent_id to be set in Token 2017-03-31 14:00:14 +02:00
Matthew Honnibal
9720103428 Improve attribute handlign in doc.merge(). Still unsatisfying 2017-03-31 13:59:58 +02:00
Matthew Honnibal
0fefdfcbda Merge pull request #935 from ericzhao28/master 
Add option to use label=ent_type in doc.merge arguments (Bug fix for issue #862)
 2017-03-30 02:51:24 +02:00
Eric Zhao
aafdf6ffb8 Add option to use label karg to determine ent_type in doc.merge 2017-03-28 23:35:03 -07:00
Matthew Honnibal
28bb546939 Merge pull request #883 from ericzhao28/master 
Add `lower_` and `upper_` properties to `Span` class
 2017-03-16 23:35:47 +01:00
ines
66c1f194f9 Use consistent unicode declarations 2017-03-12 13:07:28 +01:00
Em
9c809efc25 Removed mapStr 2017-03-11 16:23:26 -08:00
Em
426d17167f Added string manipulation for spans 2017-03-10 16:50:02 -08:00
Roman Inflianskas
66e1109b53 Add support for Universal Dependencies v2.0 2017-03-03 13:17:34 +01:00
Matvey Ezhov
32a22291bc Small Doc.count_by documentation update 
Current example doesn't work
 2017-01-31 19:18:45 +03:00
Matthew Honnibal
6c665b81df Fix redundant == TAG in from_array conditional 2017-01-31 00:46:21 +11:00
Matthew Honnibal
e7f8e13cf3 Make Token hashable. Fixes #743 2017-01-16 13:27:57 +01:00
Matthew Honnibal
12cd27b821 Amend 8ae8b443f: Handle comparison with None tokens. 2017-01-11 13:03:32 +01:00
Matthew Honnibal
44e2b0100d Support TAG attribute in doc.from_array 2017-01-10 22:47:07 +01:00
Matthew Honnibal
8ae8b443f1 Add richcmp method to Token. Closes #631 2017-01-09 19:30:31 +01:00
kengz
73a38bd4d1 Merge remote-tracking branch 'upstream/master' 2016-12-30 12:19:59 -05:00
kengz
da44183ae1 move parse_tree logic to a new tokens/printers.py file 2016-12-30 12:19:18 -05:00
Matthew Honnibal
404019ad2f Fix issue #672: ent_iob_ was a string, not unicode, due to missing unicode_literals statement. 2016-12-18 22:33:53 +01:00
Matthew Honnibal
f6e356aada Add (and test) Span.sentiment attribute. By default we average token.span, but can override with custom hook. Re Issue #667 2016-12-02 11:05:50 +01:00
Matthew Honnibal
87613edf8f Add set_struct_attr staticmethod to token 2016-11-25 12:41:47 +01:00
Matthew Honnibal
fb69aa648f Merge branch 'master' of ssh://github.com/explosion/spaCy 2016-11-25 11:35:44 +01:00
Matthew Honnibal
9a03a3f85e Add get_struct_attr staticmethod to Token, to match Lexeme.get_struct_attr. 2016-11-25 11:35:17 +01:00
Pokey Rule
3e3bda142d Add noun_chunks to Span 2016-11-24 10:47:20 +00:00
tiago
b38cfd0ef9 now span.merge returns token like it says on documentation 2016-11-09 14:58:19 +00:00
Matthew Honnibal
1fb09c3dc1 Fix morphology tagger 2016-11-04 19:19:09 +01:00
Matthew Honnibal
293c79c09a Fix #595: Lemmatization was incorrect for base forms, because morphological analyser wasn't adding morphology properly. 2016-11-04 00:29:07 +01:00
Matthew Honnibal
f292f7f0e6 Fix Issue #599, by considering empty documents to be parsed and tagged. Implementation is a bit dodgy. 2016-11-02 23:48:43 +01:00
Matthew Honnibal
05a8b752a2 Fix Issue #600: Missing setters for Token attribute. 2016-11-02 23:28:59 +01:00
Matthew Honnibal
11664b9f20 Fix variable error in token 2016-11-01 13:28:00 +01:00
Matthew Honnibal
8c4d1b46ce Fix variable error in Span 2016-11-01 13:27:44 +01:00