v2.1.5: Base support for Marathi and Korean, better pretraining, scores per entity and bug fixes

✨ New features and improvements

• NEW: Base language data for Marathi and Korean (via mecab-ko, mecab-ko-dic and natto-py).
• Improve language data for Lithuanian, Spanish, Kannada, French, Norwegian and Hindi.
• Add evaluation metrics per entity type.
• Add resume logic to spacy pretrain.
• Add optional id property to EntityRuler patterns.
• Better introspection and IDE automcomplete for custom extension attributes.
• Make Doc.is_sentenced always return True for single-token docs.

🔴 Bug fixes

• Fix issue #3490: Add evaluation metrics per entity type to Scorer.
• Fix issue #3526: Serialize EntityRuler settings correctly.
• Fix issue #3558: Improve E024 error message for incorrect GoldParse.
• Fix issue #3611: Fix bug when setting ngram parameter in text classifier.
• Fix issue #3625: Improve default punctuation rules for Hindi.
• Fix issue #3707: Improve introspection of custom attributes.
• Fix issue #3737: Check if component is callable in Language.replace_pipe.
• Fix issue #3743: Fix documentation of lex_id.
• Fix issue #3749: Change vector training script to work with latest Gensim.
• Fix issue #3762, #3934: Make Doc.is_sentenced default to True for single-token Docs.
• Fix issue #3802: Fix typo in docs example.
• Fix issue #3811: Fix type of --seed option in spacy pretrain.
• Fix issue #3822: Allow passing PhraseMatcher arguments to EntityRuler.
• Fix issue #3839: Ensure the Matcher returns correct match IDs when used with operators.
• Fix issue #3840: Improve error messages in spacy pretrain.
• Fix issue #3853: Rename vectors if multiple models are loaded to prevent clashes.
• Fix issue #3859: Update pretrain to prevent unintended overwriting of weight files.
• Fix issue #3862: Fix matcher callback example.
• Fix issue #3868: Add "v.s." to English tokenizer exceptions.
• Fix issue #3869: Make Doc.count_by work as expected.
• Fix issue #3880: Fix unflatten padding in Thinc when last element is empty.
• Fix issue #3882: Exclude user_data when copying doc in displaCy.
• Fix issue #3892: Update Tokenizer initialization docs.
• Fix issue #3912: Make text classifier raise more friendly errors.

📖 Documentation and examples

• Add documentation for Scorer, Language.evaluate and gold.docs_to_json.
• Fix various typos and inconsistencies.

👥 Contributors

Thanks to @BreakBB, @ujwal-narayan, @estr4ng7d, @maknotavailable, @ramananbalakrishnan, @nipunsadvilkar, @NirantK, @munozbravo, @intrafindBreno, @Azagh3l, @jarib, @tokestermw, @polm, @skrcode, @kabirkhan, @demongolem, @elbaulp, @clarus, @BramVanroy, @rokasramas, @askhogan, @khellan, @kognate, @cedar101 and @yash1994 for the pull requests and contributions.

v2.1.4: Training improvements and bug fixes

✨ New features and improvements

• NEW: util.filter_spans helper to filter duplicates and overlaps from a list of Span objects.
• Improve language data for Thai, Japanese, Indonesian and Dutch.
• Add --n-save-every to spacy pretrain and rename --nr-iter to --n-iter for consistency.
• Add --return-scores flag to spacy evaluate to return a dict.
• Add --n-early-stopping option to spacy train to define maximum number of iterations without dev accuracy improvements.

🔴 Bug fixes

• Fix issue #3307: Fix symlink creation to show error on Windows.
• Fix issue #3473: Fix GPU training for text classification.
• Fix issue #3475: Change favicon.
• Fix issue #3482: Add Estonian base support to documentation.
• Fix issue #3484: Ensure lemmatization is always consistent between sessions.
• Fix issue #3521: Add variations of contractions to English stop words.
• Fix issue #3523: Make spacy convert correctly default to json.
• Fix issue #3525, #3551, #3572: Fix problem that'd cause lemmas to not be lowercase.
• Fix issue #3531: Don't make "settings" or "title" required in displaCy data.
• Fix issue #3533: Remove non-existent example from docs.
• Fix issue #3546: Make sure path in GoldParse.__del__ is a string.
• Fix issue #3549: Ensure match pattern error isn't raised on empty errors list.
• Fix issue #3561: Fix DependencyParser.predict docs.
• Fix issue #3598: Allow jupyter=False to override Jupyter mode in displacy.
• Fix issue #3620: Fix bug in .iob converter.
• Fix issue #3628: Relax jsonschema pin.
• Fix issue #3667: Fix offset bug in loading pre-trained word2vec.
• Fix issue #3679: Update glossary to include missing labels in spacy.explain.
• Fix issue #3680: Re-add missing universe README.
• Fix issue #3681: Rewrite information extraction example to use Doc.retokenize.
• Fix issue #3692: Fix return value in Language.update docs.
• Fix issue #3694: Make "text" in spacy pretrain optional when "tokens" is provided.
• Fix issue #3701: Improve Token.prob and Lexeme.prob docs.
• Fix issue #3708: Fix error in regex matcher examples.
• Fix issue #3713: Call rmtree and copytree with strings in spacy train.
• Fix issue #3720: Add version tag to --base-model argument in spacy train docs.

📖 Documentation and examples

• Add free interactive spaCy course.
• Fix various typos and inconsistencies.
• Add new projects to the spaCy universe.

👥 Contributors

Thanks to @svlandeg, @wannaphongcom, @Bharat123rox, @DuyguA, @SamuelLKane, @graus, @HiromuHota, @jeannefukumaru, @ivigamberdiev, @socool, @yvespeirsman, @lemontheme, @Dobita21, @w4nderlust, @pierremonico, @bryant1410, @celikomer, @xssChauhan, @kowaalczyk, @BreakBB, @fizban99, @tokestermw, @bjascob, @pickfire, @yaph, @amitness, @henry860916, @d5555, @BramVanroy, @F0rge1cE, @richardpaulhudson, @ldorigo, @aaronkub and @devforfu for the pull requests and contributions.

v2.1.3: Improve sentencizer and serialization

✨ New features and improvements

• Allow customizing punctuation characters in sentencizer and make it serializable.
• Add new "bow" architecture for TextCategorizer, to do faster bag-of-words text classification.

🔴 Bug fixes

• Fix issue #3433, #3458: Fix mismatch of classes in parser after serialization.
• Fix issue #3464: Fix training loop in train_textcat.py example.
• Fix issue #3468: Make sentencizer set Token.is_sent_start correctly.
• Fix bug in the "ensemble" TextClassifier architecture that prevented the unigram bag-of-words submodel from working properly.

👥 Contributors

Thanks to @chkoar for the pull request!

v2.1.2: Fixes to regex handling on Python 2 and tag map

🔴 Bug fixes

• Fix issue #3356: Fix handling of unicode ranges in regular expressions on Python 2.
• Fix issue #3432: Update wasabi to better handle non-UTF-8 terminals.
• Fix issue #3445: Update docs on label argument in Span.__init__.
• Fix issue #3455: Bring English tag_map in line with UD Treebank.

📖 Documentation and examples

• Add --init-tok2vec argument to train_textcat.py example.
• Fix various typos and inconsistencies.

v2.1.1: Small GPU fixes

✨ New features and improvements

• Raise error if user is running a narrow unicode build.
• Move ud_train, ud_evaluate and other UD scripts from CLI to /bin in repo only.
• Improve accuracy of spacy pretrain by implementing cosine loss.

🔴 Bug fixes

• Fix issue #3421: Update docs and raise error for narrow unicode builds.
• Fix issue #3427: Correct mistake in French lemmatizer.
• Fix issue #3431: Make Doc.vector and Doc.vector_norm work as expected on GPU.
• Fix issue #3437: Fix installation problem on GPU.
• Fix issue #3439, #3446: Don't include UD scripts in spacy.cli.

👥 Contributors

Thanks to @mhham and @Bharat123Rox for the pull requests!

v2.1.0: New models, ULMFit/BERT/Elmo-like pretraining, faster tokenization, better Matcher, bug fixes & more

⚠️ This version of spaCy requires downloading new models. You can use the spacy validate command to find out which models need updating, and print update instructions. If you've been training your own models, you'll need to retrain them with the new version.

✨ New features and improvements

Tagger, Parser, NER and Text Categorizer

• NEW: Experimental ULMFit/BERT/Elmo-like pretraining (see #2931) via the new spacy pretrain command. This pre-trains the CNN using BERT's cloze task. A new trick we're calling Language Modelling with Approximate Outputs is used to apply the pre-training to smaller models. The pre-training outputs CNN and embedding weights that can be used in spacy train, using the new -t2v argument.
• NEW: Allow parser to do joint word segmentation and parsing. If you pass in data where the tokenizer over-segments, the parser now learns to merge the tokens.
• Make parser, tagger and NER faster, through better hyperparameters.
• Add simpler, GPU-friendly option to TextCategorizer, and allow setting exclusive_classes and architecture arguments on initialization.
• Add EntityRecognizer.labels property.
• Remove document length limit during training, by implementing faster Levenshtein alignment.
• Use Thinc v7.0, which defaults to single-thread with fast blis kernel for matrix multiplication. Parallelisation should be performed at the task level, e.g. by running more containers.

Models & Language Data

• NEW: 2-3 times faster tokenization across all languages at the same accuracy!
• NEW: Small accuracy improvements for parsing, tagging and NER for 6+ languages.
• NEW: The English and German models are now available under the MIT license.
• NEW: Statistical models for Greek.
• NEW: Alpha support for Tamil, Ukrainian and Kannada, and base language classes for Afrikaans, Bulgarian, Czech, Icelandic, Lithuanian, Latvian, Slovak, Slovenian and Albanian.
• Improve loading time of French by ~30%.
• Add Vocab.writing_system (populated via the language data) to expose settings like writing direction.

CLI

• NEW: pretrain command for ULMFit/BERT/Elmo-like pretraining (see #2931).
• NEW: New ud-train command, to train and evaluate using the CoNLL 2017 shared task data.
• Check if model is already installed before downloading it via spacy download.
• Pass additional arguments of download command to pip to customise installation.
• Improve train command by letting GoldCorpus stream data, instead of loading into memory.
• Improve init-model command, including support for lexical attributes and word-vectors, using a variety of formats. This replaces the spacy vocab command, which is now deprecated.
• Add support for multi-task objectives to train command.
• Add support for data-augmentation to train command.

Other

• NEW: Enhanced pattern API for rule-based Matcher (see #1971).
• NEW: Doc.retokenize context manager for merging and splitting tokens more efficiently.
• NEW: Add support for custom pipeline component factories via entry points (#2348).
• NEW: Implement fastText vectors with subword features.
• NEW: Built-in rule-based NER component to add entities based on match patterns (see #2513).
• NEW: Allow PhraseMatcher to match on token attributes other than ORTH, e.g. LOWER (for case-insensitive matching) or even POS or TAG.
• NEW: Replace ujson, msgpack, msgpack-numpy, pickle, cloudpickle and dill with our own package srsly to centralise dependencies and allow binary wheels.
• NEW: Doc.to_json() method which outputs data in spaCy's training format. This will be the only place where the format is hard-coded (see #2932).
• NEW: Built-in EntityRuler component to make it easier to build rule-based NER and combinations of statistical and rule-based systems.
• NEW: gold.spans_from_biluo_tags helper that returns Span objects, e.g. to overwrite the doc.ents.
• Add warnings if .similarity method is called with empty vectors or without word vectors.
• Improve rule-based Matcher and add return_matches keyword argument to Matcher.pipe to yield (doc, matches) tuples instead of only Doc objects, and as_tuples to add context to the Doc objects.
• Make stop words via Token.is_stop and Lexeme.is_stop case-insensitive.
• Accept "TEXT" as an alternative to "ORTH" in Matcher patterns.
• Use black for auto-formatting .py source and optimse codebase using flake8. You can now run flake8 spacy and it should return no errors or warnings. See CONTRIBUTING.md for details.

🔴 Bug fixes

• Fix issue #795: Fix behaviour of Token.conjuncts.
• Fix issue #1487: Add Doc.retokenize() context manager.
• Fix issue #1537: Make Span.as_doc return a copy, not a view.
• Fix issue #1574: Make sure stop words are available in medium and large English models.
• Fix issue #1585: Prevent parser from predicting unseen classes.
• Fix issue #1642: Replace regex with re and speed up tokenization.
• Fix issue #1665: Correct typos in symbol Animacy_inan and add Animacy_nhum.
• Fix issue #1748, #1798, #2756, #2934: Add simpler GPU-friendly option to TextCategorizer.
• Fix issue #1773: Prevent tokenizer exceptions from setting POS but not TAG.
• Fix issue #1782, #2343: Fix training on GPU.
• Fix issue #1816: Allow custom Language subclasses via entry points.
• Fix issue #1865: Correct licensing of it_core_news_sm model.
• Fix issue #1889: Make stop words case-insensitive.
• Fix issue #1903: Add relcl dependency label to symbols.
• Fix issue #1963: Resize Doc.tensor when merging spans.
• Fix issue #1971: Update Matcher engine to support regex, extension attributes and rich comparison.
• Fix issue #2014: Make Token.pos_ writeable.
• Fix issue #2091: Fix displacy support for RTL languages.
• Fix issue #2203, #3268: Prevent bad interaction of lemmatizer and tokenizer exceptions.
• Fix issue #2329: Correct TextCategorizer and GoldParse API docs.
• Fix issue #2369: Respect pre-defined warning filters.
• Fix issue #2390: Support setting lexical attributes during retokenization.
• Fix issue #2396: Fix Doc.get_lca_matrix.
• Fix issue #2464, #3009: Fix behaviour of Matcher's ? quantifier.
• Fix issue #2482: Fix serialization when parser model is empty.
• Fix issue #2512, #2153: Fix issue with deserialization into non-empty vocab.
• Fix issue #2603: Improve handling of missing NER tags.
• Fix issue #2644: Add table explaining training metrics to docs.
• Fix issue #2648: Fix KeyError in Vectors.most_similar.
• Fix issue #2671, #2675: Fix incorrect match ID on some patterns.
• Fix issue #2693: Only use 'sentencizer' as built-in sentence boundary component name.
• Fix issue #2728: Fix HTML escaping in displacy NER visualization and correct API docs.
• Fix issue #2740: Add ability to pass additional arguments to pipeline components.
• Fix issue #2754, #3028: Make NORM a Token attribute instead of a Lexeme attribute to allow setting context-specific norms in tokenizer exceptions.
• Fix issue #2769: Fix issue that'd cause segmentation fault when calling EntityRecognizer.add_label.
• Fix issue #2772: Fix bug in sentence starts for non-projective parses.
• Fix issue #2779: Fix handling of pre-set entities.
• Fix issue #2782: Make like_num work with prefixed numbers.
• Fix issue #2833: Raise better error if Token or Span are pickled.
• Fix issue #2838: Add Retokenizer.split method to split one token into several.
• Fix issue #2869: Make doc[0].is_sent_start == True.
• Fix issue #2870: Make it illegal for the entity recognizer to predict whitespace tokens as B, L or U.
• Fix issue #2871: Fix vectors for reserved words.
• Fix issue #2901: Fix issue with first call of nlp in Japanese (MeCab).
• Fix issue #2924: Make IDs of displaCy arcs more unique to avoid clashes.
• Fix issue #3012: Fix clobber of Doc.is_tagged in Doc.from_array.
• Fix issue #3027: Allow Span to take unicode value for label argument.
• Fix issue #3036: Support mutable default arguments in extension attributes.
• Fix issue #3048: Raise better errors for uninitialized pipeline components.
• Fix issue #3064: Allow single string attributes in Doc.to_array.
• Fix issue #3093, #3067: Set vectors.name correctly when exporting model via CLI.
• Fix issue #3112: Make sure entity types are added correctly on GPU.
• Fix issue #3191: Fix pickling of Japanese.
• Fix issue #3122: Correct docs of Token.subtree and Span.subtree.
• Fix issue #3128: Improve error handling in converters.
• Fix issue #3248: Fix PhraseMatcher pickling and make __len__ consistent.
• Fix issue #3274: Make Token.sent work as expected without the parser.
• Fix issue #3277: Add en/em dash to tokenizer prefixes and suffixes.
• Fix issue #3346: Expose Japanese stop words in language class.
• Fix issue #3357: Update displaCy examples in docs to correctly show Token.pos_.
• Fix issue #3345: Fix NER when preset entities cross-sentence boundaries.
• Fix issue #3348: Don't use numpy directly for similarity.
• Fix issue #3366: Improve converters, training data formats and docs.
• Fix issue #3369: Fix #egg fragments in direct downloads.
• Fix issue #3382: Make Doc.from_array consistent with Doc.to_array.
• Fix issu...

v2.1.0a13: New models, ULMFit/BERT/Elmo-like pretraining, joint word segmentation and parsing, better Matcher, bug fixes & more

🌙 This is an alpha pre-release of spaCy v2.1.0 and available on pip as spacy-nightly. It's not intended for production use. See here for the updated nightly docs.

pip install -U spacy-nightly

If you want to test the new version, we recommend using a new virtual environment. Also make sure to download the new models – see below for details and benchmarks.

⚠️ This nightly release currently doesn't work on Python 2.7 on Windows, due to difficulties compiling our new matrix multiplication dependency blis in that environment. We expect this can be corrected in future.

✨ New features and improvements

Tagger, Parser, NER and Text Categorizer

• NEW: Experimental ULMFit/BERT/Elmo-like pretraining (see #2931) via the new spacy pretrain command. This pre-trains the CNN using BERT's cloze task. A new trick we're calling Language Modelling with Approximate Outputs is used to apply the pre-training to smaller models. The pre-training outputs CNN and embedding weights that can be used in spacy train, using the new -t2v argument.
• NEW: Allow parser to do joint word segmentation and parsing. If you pass in data where the tokenizer over-segments, the parser now learns to merge the tokens.
• Make parser, tagger and NER faster, through better hyperparameters.
• Add simpler, GPU-friendly option to TextCategorizer, and allow setting exclusive_classes and architecture arguments on initialization.
• Add EntityRecognizer.labels property.
• Remove document length limit during training, by implementing faster Levenshtein alignment.
• Use Thinc v7.0, which defaults to single-thread with fast blis kernel for matrix multiplication. Parallelisation should be performed at the task level, e.g. by running more containers.

Models & Language Data

• NEW: 2-3 times faster tokenization across all languages at the same accuracy!
• NEW: Small accuracy improvements for parsing, tagging and NER for 6+ languages.
• NEW: The English and German models are now available under the MIT license.
• NEW: Statistical models for Greek.
• NEW: Alpha support for Tamil, Ukrainian and Kannada, and base language classes for Afrikaans, Bulgarian, Czech, Icelandic, Lithuanian, Latvian, Slovak, Slovenian and Albanian.
• Improve loading time of French by ~30%.
• Add Vocab.writing_system (populated via the language data) to expose settings like writing direction.

CLI

• NEW: pretrain command for ULMFit/BERT/Elmo-like pretraining (see #2931).
• NEW: New ud-train command, to train and evaluate using the CoNLL 2017 shared task data.
• Check if model is already installed before downloading it via spacy download.
• Pass additional arguments of download command to pip to customise installation.
• Improve train command by letting GoldCorpus stream data, instead of loading into memory.
• Improve init-model command, including support for lexical attributes and word-vectors, using a variety of formats. This replaces the spacy vocab command, which is now deprecated.
• Add support for multi-task objectives to train command.
• Add support for data-augmentation to train command.

Other

• NEW: Enhanced pattern API for rule-based Matcher (see #1971).
• NEW: Doc.retokenize context manager for merging and splitting tokens more efficiently.
• NEW: Add support for custom pipeline component factories via entry points (#2348).
• NEW: Implement fastText vectors with subword features.
• NEW: Built-in rule-based NER component to add entities based on match patterns (see #2513).
• NEW: Allow PhraseMatcher to match on token attributes other than ORTH, e.g. LOWER (for case-insensitive matching) or even POS or TAG.
• NEW: Replace ujson, msgpack, msgpack-numpy, pickle, cloudpickle and dill with our own package srsly to centralise dependencies and allow binary wheels.
• NEW: Doc.to_json() method which outputs data in spaCy's training format. This will be the only place where the format is hard-coded (see #2932).
• NEW: Built-in EntityRuler component to make it easier to build rule-based NER and combinations of statistical and rule-based systems.
• NEW: gold.spans_from_biluo_tags helper that returns Span objects, e.g. to overwrite the doc.ents.
• Add warnings if .similarity method is called with empty vectors or without word vectors.
• Improve rule-based Matcher and add return_matches keyword argument to Matcher.pipe to yield (doc, matches) tuples instead of only Doc objects, and as_tuples to add context to the Doc objects.
• Make stop words via Token.is_stop and Lexeme.is_stop case-insensitive.
• Accept "TEXT" as an alternative to "ORTH" in Matcher patterns.
• Use black for auto-formatting .py source and optimse codebase using flake8. You can now run flake8 spacy and it should return no errors or warnings. See CONTRIBUTING.md for details.

🔴 Bug fixes

• Fix issue #795: Fix behaviour of Token.conjuncts.
• Fix issue #1487: Add Doc.retokenize() context manager.
• Fix issue #1537: Make Span.as_doc return a copy, not a view.
• Fix issue #1574: Make sure stop words are available in medium and large English models.
• Fix issue #1585: Prevent parser from predicting unseen classes.
• Fix issue #1642: Replace regex with re and speed up tokenization.
• Fix issue #1665: Correct typos in symbol Animacy_inan and add Animacy_nhum.
• Fix issue #1748, #1798, #2756, #2934: Add simpler GPU-friendly option to TextCategorizer.
• Fix issue #1773: Prevent tokenizer exceptions from setting POS but not TAG.
• Fix issue #1782, #2343: Fix training on GPU.
• Fix issue #1816: Allow custom Language subclasses via entry points.
• Fix issue #1865: Correct licensing of it_core_news_sm model.
• Fix issue #1889: Make stop words case-insensitive.
• Fix issue #1903: Add relcl dependency label to symbols.
• Fix issue #1963: Resize Doc.tensor when merging spans.
• Fix issue #1971: Update Matcher engine to support regex, extension attributes and rich comparison.
• Fix issue #2014: Make Token.pos_ writeable.
• Fix issue #2091: Fix displacy support for RTL languages.
• Fix issue #2203, #3268: Prevent bad interaction of lemmatizer and tokenizer exceptions.
• Fix issue #2329: Correct TextCategorizer and GoldParse API docs.
• Fix issue #2369: Respect pre-defined warning filters.
• Fix issue #2390: Support setting lexical attributes during retokenization.
• Fix issue #2396: Fix Doc.get_lca_matrix.
• Fix issue #2464, #3009: Fix behaviour of Matcher's ? quantifier.
• Fix issue #2482: Fix serialization when parser model is empty.
• Fix issue #2512, #2153: Fix issue with deserialization into non-empty vocab.
• Fix issue #2603: Improve handling of missing NER tags.
• Fix issue #2644: Add table explaining training metrics to docs.
• Fix issue #2648: Fix KeyError in Vectors.most_similar.
• Fix issue #2671, #2675: Fix incorrect match ID on some patterns.
• Fix issue #2693: Only use 'sentencizer' as built-in sentence boundary component name.
• Fix issue #2728: Fix HTML escaping in displacy NER visualization and correct API docs.
• Fix issue #2740: Add ability to pass additional arguments to pipeline components.
• Fix issue #2754, #3028: Make NORM a Token attribute instead of a Lexeme attribute to allow setting context-specific norms in tokenizer exceptions.
• Fix issue #2769: Fix issue that'd cause segmentation fault when calling EntityRecognizer.add_label.
• Fix issue #2772: Fix bug in sentence starts for non-projective parses.
• Fix issue #2779: Fix handling of pre-set entities.
• Fix issue #2782: Make like_num work with prefixed numbers.
• Fix issue #2833: Raise better error if Token or Span are pickled.
• Fix issue #2838: Add Retokenizer.split method to split one token into several.
• Fix issue #2869: Make doc[0].is_sent_start == True.
• Fix issue #2870: Make it illegal for the entity recognizer to predict whitespace tokens as B, L or U.
• Fix issue #2871: Fix vectors for reserved words.
• Fix issue #2901: Fix issue with first call of nlp in Japanese (MeCab).
• Fix issue #2924: Make IDs of displaCy arcs more unique to avoid clashes.
• Fix issue #3012: Fix clobber of Doc.is_tagged in Doc.from_array.
• Fix issue #3027: Allow Span to take unicode value for label argument.
• Fix issue #3036: Support mutable default arguments in extension attributes.
• Fix issue #3048: Raise better errors for uninitialized pipeline components.
• Fix issue #3064: Allow single string attributes in Doc.to_array.
• Fix issue #3093, #3067: Set vectors.name correctly when exporting model via CLI.
• Fix issue #3112: Make sure entity types are added correctly on GPU.
• Fix issue #3191: Fix pickling of Japanese.
• Fix issue #3122: Correct docs of Token.subtree and Span.subtree.
• Fix issue #3128: Improve error handling in converters.
• Fix issue #3248: Fix PhraseMatcher pickling and make __len__ consistent.
• Fix issue #3274: Make Token.sent work as expected without the parser.
• Fix issue #3277: Add en/em dash to tokenizer prefixes and suffixes.
• Fix issue #3346: Expose Japanese stop words in language class.
• Fix issue #3357: Update displ...

v2.1.0a12: New models, ULMFit/BERT/Elmo-like pretraining, joint word segmentation and parsing, better Matcher, bug fixes & more

🌙 This is an alpha pre-release of spaCy v2.1.0 and available on pip as spacy-nightly. It's not intended for production use. See here for the updated nightly docs.

pip install -U spacy-nightly

If you want to test the new version, we recommend using a new virtual environment. Also make sure to download the new models – see below for details and benchmarks.

⚠️ Due to difficulties linking our new blis for faster platform-independent matrix multiplication, this nightly release currently doesn't work on Python 2.7 on Windows. We expect this to be corrected in the future.

✨ New features and improvements

Tagger, Parser, NER and Text Categorizer

• NEW: Experimental ULMFit/BERT/Elmo-like pretraining (see #2931) via the new spacy pretrain command. This pre-trains the CNN using BERT's cloze task. A new trick we're calling Language Modelling with Approximate Outputs is used to apply the pre-training to smaller models. The pre-training outputs CNN and embedding weights that can be used in spacy train, using the new -t2v argument.
• NEW: Allow parser to do joint word segmentation and parsing. If you pass in data where the tokenizer over-segments, the parser now learns to merge the tokens.
• Make parser, tagger and NER faster, through better hyperparameters.
• Add simpler, GPU-friendly option to TextCategorizer, and allow setting exclusive_classes and architecture arguments on initialization.
• Add EntityRecognizer.labels property.
• Remove document length limit during training, by implementing faster Levenshtein alignment.
• Use Thinc v7.0, which defaults to single-thread with fast blis kernel for matrix multiplication. Parallelisation should be performed at the task level, e.g. by running more containers.

Models & Language Data

• NEW: 2-3 times faster tokenization across all languages at the same accuracy!
• NEW: Small accuracy improvements for parsing, tagging and NER for 6+ languages.
• NEW: The English and German models are now available under the MIT license.
• NEW: Statistical models for Greek.
• NEW: Alpha support for Tamil, Ukrainian and Kannada, and base language classes for Afrikaans, Bulgarian, Czech, Icelandic, Lithuanian, Latvian, Slovak, Slovenian and Albanian.
• Improve loading time of French by ~30%.
• Add Vocab.writing_system (populated via the language data) to expose settings like writing direction.

CLI

• NEW: pretrain command for ULMFit/BERT/Elmo-like pretraining (see #2931).
• NEW: New ud-train command, to train and evaluate using the CoNLL 2017 shared task data.
• Check if model is already installed before downloading it via spacy download.
• Pass additional arguments of download command to pip to customise installation.
• Improve train command by letting GoldCorpus stream data, instead of loading into memory.
• Improve init-model command, including support for lexical attributes and word-vectors, using a variety of formats. This replaces the spacy vocab command, which is now deprecated.
• Add support for multi-task objectives to train command.
• Add support for data-augmentation to train command.

Other

• NEW: Enhanced pattern API for rule-based Matcher (see #1971).
• NEW: Doc.retokenize context manager for merging and splitting tokens more efficiently.
• NEW: Add support for custom pipeline component factories via entry points (#2348).
• NEW: Implement fastText vectors with subword features.
• NEW: Built-in rule-based NER component to add entities based on match patterns (see #2513).
• NEW: Allow PhraseMatcher to match on token attributes other than ORTH, e.g. LOWER (for case-insensitive matching) or even POS or TAG.
• NEW: Replace ujson, msgpack, msgpack-numpy, pickle, cloudpickle and dill with our own package srsly to centralise dependencies and allow binary wheels.
• NEW: Doc.to_json() method which outputs data in spaCy's training format. This will be the only place where the format is hard-coded (see #2932).
• NEW: Built-in EntityRuler component to make it easier to build rule-based NER and combinations of statistical and rule-based systems.
• NEW: gold.spans_from_biluo_tags helper that returns Span objects, e.g. to overwrite the doc.ents.
• Add warnings if .similarity method is called with empty vectors or without word vectors.
• Improve rule-based Matcher and add return_matches keyword argument to Matcher.pipe to yield (doc, matches) tuples instead of only Doc objects, and as_tuples to add context to the Doc objects.
• Make stop words via Token.is_stop and Lexeme.is_stop case-insensitive.
• Accept "TEXT" as an alternative to "ORTH" in Matcher patterns.
• Use black for auto-formatting .py source and optimse codebase using flake8. You can now run flake8 spacy and it should return no errors or warnings. See CONTRIBUTING.md for details.

🔴 Bug fixes

• Fix issue #795: Fix behaviour of Token.conjuncts.
• Fix issue #1487: Add Doc.retokenize() context manager.
• Fix issue #1537: Make Span.as_doc return a copy, not a view.
• Fix issue #1574: Make sure stop words are available in medium and large English models.
• Fix issue #1585: Prevent parser from predicting unseen classes.
• Fix issue #1642: Replace regex with re and speed up tokenization.
• Fix issue #1665: Correct typos in symbol Animacy_inan and add Animacy_nhum.
• Fix issue #1748, #1798, #2756, #2934: Add simpler GPU-friendly option to TextCategorizer.
• Fix issue #1773: Prevent tokenizer exceptions from setting POS but not TAG.
• Fix issue #1782, #2343: Fix training on GPU.
• Fix issue #1816: Allow custom Language subclasses via entry points.
• Fix issue #1865: Correct licensing of it_core_news_sm model.
• Fix issue #1889: Make stop words case-insensitive.
• Fix issue #1903: Add relcl dependency label to symbols.
• Fix issue #1963: Resize Doc.tensor when merging spans.
• Fix issue #1971: Update Matcher engine to support regex, extension attributes and rich comparison.
• Fix issue #2014: Make Token.pos_ writeable.
• Fix issue #2091: Fix displacy support for RTL languages.
• Fix issue #2203, #3268: Prevent bad interaction of lemmatizer and tokenizer exceptions.
• Fix issue #2329: Correct TextCategorizer and GoldParse API docs.
• Fix issue #2369: Respect pre-defined warning filters.
• Fix issue #2390: Support setting lexical attributes during retokenization.
• Fix issue #2396: Fix Doc.get_lca_matrix.
• Fix issue #2464, #3009: Fix behaviour of Matcher's ? quantifier.
• Fix issue #2482: Fix serialization when parser model is empty.
• Fix issue #2512, #2153: Fix issue with deserialization into non-empty vocab.
• Fix issue #2603: Improve handling of missing NER tags.
• Fix issue #2644: Add table explaining training metrics to docs.
• Fix issue #2648: Fix KeyError in Vectors.most_similar.
• Fix issue #2671, #2675: Fix incorrect match ID on some patterns.
• Fix issue #2693: Only use 'sentencizer' as built-in sentence boundary component name.
• Fix issue #2728: Fix HTML escaping in displacy NER visualization and correct API docs.
• Fix issue #2740: Add ability to pass additional arguments to pipeline components.
• Fix issue #2754, #3028: Make NORM a Token attribute instead of a Lexeme attribute to allow setting context-specific norms in tokenizer exceptions.
• Fix issue #2769: Fix issue that'd cause segmentation fault when calling EntityRecognizer.add_label.
• Fix issue #2772: Fix bug in sentence starts for non-projective parses.
• Fix issue #2779: Fix handling of pre-set entities.
• Fix issue #2782: Make like_num work with prefixed numbers.
• Fix issue #2833: Raise better error if Token or Span are pickled.
• Fix issue #2838: Add Retokenizer.split method to split one token into several.
• Fix issue #2869: Make doc[0].is_sent_start == True.
• Fix issue #2870: Make it illegal for the entity recognizer to predict whitespace tokens as B, L or U.
• Fix issue #2871: Fix vectors for reserved words.
• Fix issue #2901: Fix issue with first call of nlp in Japanese (MeCab).
• Fix issue #2924: Make IDs of displaCy arcs more unique to avoid clashes.
• Fix issue #3012: Fix clobber of Doc.is_tagged in Doc.from_array.
• Fix issue #3027: Allow Span to take unicode value for label argument.
• Fix issue #3036: Support mutable default arguments in extension attributes.
• Fix issue #3048: Raise better errors for uninitialized pipeline components.
• Fix issue #3064: Allow single string attributes in Doc.to_array.
• Fix issue #3093, #3067: Set vectors.name correctly when exporting model via CLI.
• Fix issue #3112: Make sure entity types are added correctly on GPU.
• Fix issue #3191: Fix pickling of Japanese.
• Fix issue #3122: Correct docs of Token.subtree and Span.subtree.
• Fix issue #3128: Improve error handling in converters.
• Fix issue #3248: Fix PhraseMatcher pickling and make __len__ consistent.
• Fix issue #3274: Make Token.sent work as expected without the parser.
• Fix issue #3277: Add en/em dash to tokenizer prefixes and suffixes.
• Fix issue #3346: Expose Japanese stop words in language class.
• Fix issue #3357: Update dis...

v2.1.0a11: New models, ULMFit/BERT/Elmo-like pretraining, joint word segmentation and parsing, better Matcher, bug fixes & more

🌙 This is an alpha pre-release of spaCy v2.1.0 and available on pip as spacy-nightly. It's not intended for production use. See here for the updated nightly docs.

pip install -U spacy-nightly

If you want to test the new version, we recommend using a new virtual environment. Also make sure to download the new models – see below for details and benchmarks.

⚠️ Due to difficulties linking our new blis for faster platform-independent matrix multiplication, this nightly release currently doesn't work on Python 2.7 on Windows. We expect this to be corrected in the future.

✨ New features and improvements

Tagger, Parser, NER and Text Categorizer

• NEW: Experimental ULMFit/BERT/Elmo-like pretraining (see #2931) via the new spacy pretrain command. This pre-trains the CNN using BERT's cloze task. A new trick we're calling Language Modelling with Approximate Outputs is used to apply the pre-training to smaller models. The pre-training outputs CNN and embedding weights that can be used in spacy train, using the new -t2v argument.
• NEW: Allow parser to do joint word segmentation and parsing. If you pass in data where the tokenizer over-segments, the parser now learns to merge the tokens.
• Make parser, tagger and NER faster, through better hyperparameters.
• Add simpler, GPU-friendly option to TextCategorizer, and allow setting exclusive_classes and architecture arguments on initialization.
• Add EntityRecognizer.labels property.
• Remove document length limit during training, by implementing faster Levenshtein alignment.
• Use Thinc v7.0, which defaults to single-thread with fast blis kernel for matrix multiplication. Parallelisation should be performed at the task level, e.g. by running more containers.

Models & Language Data

• NEW: 2-3 times faster tokenization across all languages at the same accuracy!
• NEW: Small accuracy improvements for parsing, tagging and NER for 6+ languages.
• NEW: The English and German models are now available under the MIT license.
• NEW: Statistical models for Greek.
• NEW: Alpha support for Tamil, Ukrainian and Kannada, and base language classes for Afrikaans, Bulgarian, Czech, Icelandic, Lithuanian, Latvian, Slovak, Slovenian and Albanian.
• Improve loading time of French by ~30%.

CLI

• NEW: pretrain command for ULMFit/BERT/Elmo-like pretraining (see #2931).
• NEW: New ud-train command, to train and evaluate using the CoNLL 2017 shared task data.
• Check if model is already installed before downloading it via spacy download.
• Pass additional arguments of download command to pip to customise installation.
• Improve train command by letting GoldCorpus stream data, instead of loading into memory.
• Improve init-model command, including support for lexical attributes and word-vectors, using a variety of formats. This replaces the spacy vocab command, which is now deprecated.
• Add support for multi-task objectives to train command.
• Add support for data-augmentation to train command.

Other

• NEW: Enhanced pattern API for rule-based Matcher (see #1971).
• NEW: Doc.retokenize context manager for merging and splitting tokens more efficiently.
• NEW: Add support for custom pipeline component factories via entry points (#2348).
• NEW: Implement fastText vectors with subword features.
• NEW: Built-in rule-based NER component to add entities based on match patterns (see #2513).
• NEW: Allow PhraseMatcher to match on token attributes other than ORTH, e.g. LOWER (for case-insensitive matching) or even POS or TAG.
• NEW: Replace ujson, msgpack, msgpack-numpy, pickle, cloudpickle and dill with our own package srsly to centralise dependencies and allow binary wheels.
• NEW: Doc.to_json() method which outputs data in spaCy's training format. This will be the only place where the format is hard-coded (see #2932).
• NEW: Built-in EntityRuler component to make it easier to build rule-based NER and combinations of statistical and rule-based systems.
• NEW: gold.spans_from_biluo_tags helper that returns Span objects, e.g. to overwrite the doc.ents.
• Add warnings if .similarity method is called with empty vectors or without word vectors.
• Improve rule-based Matcher and add return_matches keyword argument to Matcher.pipe to yield (doc, matches) tuples instead of only Doc objects, and as_tuples to add context to the Doc objects.
• Make stop words via Token.is_stop and Lexeme.is_stop case-insensitive.
• Accept "TEXT" as an alternative to "ORTH" in Matcher patterns.
• Use black for auto-formatting .py source and optimse codebase using flake8. You can now run flake8 spacy and it should return no errors or warnings. See CONTRIBUTING.md for details.

🔴 Bug fixes

• Fix issue #1487: Add Doc.retokenize() context manager.
• Fix issue #1537: Make Span.as_doc return a copy, not a view.
• Fix issue #1574: Make sure stop words are available in medium and large English models.
• Fix issue #1585: Prevent parser from predicting unseen classes.
• Fix issue #1642: Replace regex with re and speed up tokenization.
• Fix issue #1665: Correct typos in symbol Animacy_inan and add Animacy_nhum.
• Fix issue #1748, #1798, #2756, #2934: Add simpler GPU-friendly option to TextCategorizer.
• Fix issue #1773: Prevent tokenizer exceptions from setting POS but not TAG.
• Fix issue #1782, #2343: Fix training on GPU.
• Fix issue #1816: Allow custom Language subclasses via entry points.
• Fix issue #1865: Correct licensing of it_core_news_sm model.
• Fix issue #1889: Make stop words case-insensitive.
• Fix issue #1903: Add relcl dependency label to symbols.
• Fix issue #1963: Resize Doc.tensor when merging spans.
• Fix issue #1971: Update Matcher engine to support regex, extension attributes and rich comparison.
• Fix issue #2014: Make Token.pos_ writeable.
• Fix issue #2203, #3268: Prevent bad interaction of lemmatizer and tokenizer exceptions.
• Fix issue #2329: Correct TextCategorizer and GoldParse API docs.
• Fix issue #2369: Respect pre-defined warning filters.
• Fix issue #2390: Support setting lexical attributes during retokenization.
• Fix issue #2396: Fix Doc.get_lca_matrix.
• Fix issue #2464, #3009: Fix behaviour of Matcher's ? quantifier.
• Fix issue #2482: Fix serialization when parser model is empty.
• Fix issue #2512, #2153: Fix issue with deserialization into non-empty vocab.
• Fix issue #2603: Improve handling of missing NER tags.
• Fix issue #2644: Add table explaining training metrics to docs.
• Fix issue #2648: Fix KeyError in Vectors.most_similar.
• Fix issue #2671, #2675: Fix incorrect match ID on some patterns.
• Fix issue #2693: Only use 'sentencizer' as built-in sentence boundary component name.
• Fix issue #2728: Fix HTML escaping in displacy NER visualization and correct API docs.
• Fix issue #2740: Add ability to pass additional arguments to pipeline components.
• Fix issue #2754, #3028: Make NORM a Token attribute instead of a Lexeme attribute to allow setting context-specific norms in tokenizer exceptions.
• Fix issue #2769: Fix issue that'd cause segmentation fault when calling EntityRecognizer.add_label.
• Fix issue #2772: Fix bug in sentence starts for non-projective parses.
• Fix issue #2779: Fix handling of pre-set entities.
• Fix issue #2782: Make like_num work with prefixed numbers.
• Fix issue #2833: Raise better error if Token or Span are pickled.
• Fix issue #2838: Add Retokenizer.split method to split one token into several.
• Fix issue #2869: Make doc[0].is_sent_start == True.
• Fix issue #2870: Make it illegal for the entity recognizer to predict whitespace tokens as B, L or U.
• Fix issue #2871: Fix vectors for reserved words.
• Fix issue #2901: Fix issue with first call of nlp in Japanese (MeCab).
• Fix issue #2924: Make IDs of displaCy arcs more unique to avoid clashes.
• Fix issue #3012: Fix clobber of Doc.is_tagged in Doc.from_array.
• Fix issue #3027: Allow Span to take unicode value for label argument.
• Fix issue #3048: Raise better errors for uninitialized pipeline components.
• Fix issue #3064: Allow single string attributes in Doc.to_array.
• Fix issue #3093, #3067: Set vectors.name correctly when exporting model via CLI.
• Fix issue #3112: Make sure entity types are added correctly on GPU.
• Fix issue #3122: Correct docs of Token.subtree and Span.subtree.
• Fix issue #3128: Improve error handling in converters.
• Fix issue #3248: Fix PhraseMatcher pickling and make __len__ consistent.
• Fix issue #3274: Make Token.sent work as expected without the parser.
• Fix issue #3277: Add en/em dash to tokenizer prefixes and suffixes.
• Fix issue #3346: Expose Japanese stop words in language class.
• Fix issue #3357: Update displaCy examples in docs to correctly show Token.pos_.
• Fix issue #3345: Fix NER when preset entities cross-sentence boundaries.
• Fix issue #3348: Don't use numpy directly for similarity.
• Fix issue #3366: Improve converters, training data formats and docs.
• Fix issue #3369: Fix #egg fragments in direct downloads.
• Fix issue #3382: Mak...

v2.1.0a10: New models, ULMFit/BERT/Elmo-like pretraining, joint word segmentation and parsing, better Matcher, bug fixes & more

🌙 This is an alpha pre-release of spaCy v2.1.0 and available on pip as spacy-nightly. It's not intended for production use. See here for the updated nightly docs.

pip install -U spacy-nightly

If you want to test the new version, we recommend using a new virtual environment. Also make sure to download the new models – see below for details and benchmarks.

⚠️ Due to difficulties linking our new blis for faster platform-independent matrix multiplication, this nightly release currently doesn't work on Python 2.7 on Windows. We expect this to be corrected in the future.

✨ New features and improvements

Tagger, Parser, NER and Text Categorizer

• NEW: Experimental ULMFit/BERT/Elmo-like pretraining (see #2931) via the new spacy pretrain command. This pre-trains the CNN using BERT's cloze task. A new trick we're calling Language Modelling with Approximate Outputs is used to apply the pre-training to smaller models. The pre-training outputs CNN and embedding weights that can be used in spacy train, using the new -t2v argument.
• NEW: Allow parser to do joint word segmentation and parsing. If you pass in data where the tokenizer over-segments, the parser now learns to merge the tokens.
• Make parser, tagger and NER faster, through better hyperparameters.
• Add simpler, GPU-friendly option to TextCategorizer, and allow setting exclusive_classes and architecture arguments on initialization.
• Add EntityRecognizer.labels property.
• Remove document length limit during training, by implementing faster Levenshtein alignment.
• Use Thinc v7.0, which defaults to single-thread with fast blis kernel for matrix multiplication. Parallelisation should be performed at the task level, e.g. by running more containers.

Models & Language Data

• NEW: 2-3 times faster tokenization across all languages at the same accuracy!
• NEW: Small accuracy improvements for parsing, tagging and NER for 6+ languages.
• NEW: The English and German models are now available under the MIT license.
• NEW: Statistical models for Greek.
• NEW: Alpha support for Tamil, Ukrainian and Kannada, and base language classes for Afrikaans, Bulgarian, Czech, Icelandic, Lithuanian, Latvian, Slovak, Slovenian and Albanian.
• Improve loading time of French by ~30%.

CLI

• NEW: pretrain command for ULMFit/BERT/Elmo-like pretraining (see #2931).
• NEW: New ud-train command, to train and evaluate using the CoNLL 2017 shared task data.
• Check if model is already installed before downloading it via spacy download.
• Pass additional arguments of download command to pip to customise installation.
• Improve train command by letting GoldCorpus stream data, instead of loading into memory.
• Improve init-model command, including support for lexical attributes and word-vectors, using a variety of formats. This replaces the spacy vocab command, which is now deprecated.
• Add support for multi-task objectives to train command.
• Add support for data-augmentation to train command.

Other

• NEW: Enhanced pattern API for rule-based Matcher (see #1971).
• NEW: Doc.retokenize context manager for merging and splitting tokens more efficiently.
• NEW: Add support for custom pipeline component factories via entry points (#2348).
• NEW: Implement fastText vectors with subword features.
• NEW: Built-in rule-based NER component to add entities based on match patterns (see #2513).
• NEW: Allow PhraseMatcher to match on token attributes other than ORTH, e.g. LOWER (for case-insensitive matching) or even POS or TAG.
• NEW: Replace ujson, msgpack, msgpack-numpy, pickle, cloudpickle and dill with our own package srsly to centralise dependencies and allow binary wheels.
• NEW: Doc.to_json() method which outputs data in spaCy's training format. This will be the only place where the format is hard-coded (see #2932).
• NEW: Built-in EntityRuler component to make it easier to build rule-based NER and combinations of statistical and rule-based systems.
• NEW: gold.spans_from_biluo_tags helper that returns Span objects, e.g. to overwrite the doc.ents.
• Add warnings if .similarity method is called with empty vectors or without word vectors.
• Improve rule-based Matcher and add return_matches keyword argument to Matcher.pipe to yield (doc, matches) tuples instead of only Doc objects, and as_tuples to add context to the Doc objects.
• Make stop words via Token.is_stop and Lexeme.is_stop case-insensitive.
• Accept "TEXT" as an alternative to "ORTH" in Matcher patterns.
• Use black for auto-formatting .py source and optimse codebase using flake8. You can now run flake8 spacy and it should return no errors or warnings. See CONTRIBUTING.md for details.

🔴 Bug fixes

• Fix issue #1487: Add Doc.retokenize() context manager.
• Fix issue #1537: Make Span.as_doc return a copy, not a view.
• Fix issue #1574: Make sure stop words are available in medium and large English models.
• Fix issue #1585: Prevent parser from predicting unseen classes.
• Fix issue #1642: Replace regex with re and speed up tokenization.
• Fix issue #1665: Correct typos in symbol Animacy_inan and add Animacy_nhum.
• Fix issue #1748, #1798, #2756, #2934: Add simpler GPU-friendly option to TextCategorizer.
• Fix issue #1773: Prevent tokenizer exceptions from setting POS but not TAG.
• Fix issue #1782, #2343: Fix training on GPU.
• Fix issue #1816: Allow custom Language subclasses via entry points.
• Fix issue #1865: Correct licensing of it_core_news_sm model.
• Fix issue #1889: Make stop words case-insensitive.
• Fix issue #1903: Add relcl dependency label to symbols.
• Fix issue #1963: Resize Doc.tensor when merging spans.
• Fix issue #1971: Update Matcher engine to support regex, extension attributes and rich comparison.
• Fix issue #2014: Make Token.pos_ writeable.
• Fix issue #2329: Correct TextCategorizer and GoldParse API docs.
• Fix issue #2369: Respect pre-defined warning filters.
• Fix issue #2390: Support setting lexical attributes during retokenization.
• Fix issue #2396: Fix Doc.get_lca_matrix.
• Fix issue #2464, #3009: Fix behaviour of Matcher's ? quantifier.
• Fix issue #2482: Fix serialization when parser model is empty.
• Fix issue #2603: Improve handling of missing NER tags.
• Fix issue #2644: Add table explaining training metrics to docs.
• Fix issue #2648: Fix KeyError in Vectors.most_similar.
• Fix issue #2671, #2675: Fix incorrect match ID on some patterns.
• Fix issue #2693: Only use 'sentencizer' as built-in sentence boundary component name.
• Fix issue #2728: Fix HTML escaping in displacy NER visualization and correct API docs.
• Fix issue #2754, #3028: Make NORM a Token attribute instead of a Lexeme attribute to allow setting context-specific norms in tokenizer exceptions.
• Fix issue #2769: Fix issue that'd cause segmentation fault when calling EntityRecognizer.add_label.
• Fix issue #2772: Fix bug in sentence starts for non-projective parses.
• Fix issue #2779: Fix handling of pre-set entities.
• Fix issue #2782: Make like_num work with prefixed numbers.
• Fix issue #2833: Raise better error if Token or Span are pickled.
• Fix issue #2838: Add Retokenizer.split method to split one token into several.
• Fix issue #2869: Make doc[0].is_sent_start == True.
• Fix issue #2870: Make it illegal for the entity recognizer to predict whitespace tokens as B, L or U.
• Fix issue #2871: Fix vectors for reserved words.
• Fix issue #2901: Fix issue with first call of nlp in Japanese (MeCab).
• Fix issue #2924: Make IDs of displaCy arcs more unique to avoid clashes.
• Fix issue #3012: Fix clobber of Doc.is_tagged in Doc.from_array.
• Fix issue #3027: Allow Span to take unicode value for label argument.
• Fix issue #3048: Raise better errors for uninitialized pipeline components.
• Fix issue #3064: Allow single string attributes in Doc.to_array.
• Fix issue #3093, #3067: Set vectors.name correctly when exporting model via CLI.
• Fix issue #3112: Make sure entity types are added correctly on GPU.
• Fix issue #3122: Correct docs of Token.subtree and Span.subtree.
• Fix issue #3128: Improve error handling in converters.
• Fix issue #3248: Fix PhraseMatcher pickling and make __len__ consistent.
• Fix issue #3274: Make Token.sent work as expected without the parser.
• Fix issue #3277: Add en/em dash to tokenizer prefixes and suffixes.
• Fix serialization of custom tokenizer if not all functions are defined.
• Fix bugs in beam-search training objective.
• Fix problems with model pickling.

⚠️ Backwards incompatibilities

• This version of spaCy requires downloading new models. You can use the spacy validate command to find out which models need updating, and print update instructions.
• If you've been training your own models, you'll need to retrain them with the new version.
• Due to difficulties linking our new blis for faster platform-independent matrix multiplication, v2.1.x currently **doesn't work on Python 2.7 o...