spaCy v2.1 has focussed primarily on stability and performance, solidifying the design changes introduced in v2.0. As well as smaller models, faster runtime, and many bug fixes, v2.1 also introduces experimental support for some exciting new NLP innovations. For the full changelog, see the release notes on GitHub. For more details and a behind-the-scenes look at the new release, see our blog post.

BERT/ULMFit/Elmo-style pre-training experimental

spaCy v2.1 introduces a new CLI command, spacy pretrain, that can make your models much more accurate. It’s especially useful when you have limited training data. The spacy pretrain command lets you use transfer learning to initialize your models with information from raw text, using a language model objective similar to the one used in Google’s BERT system. We’ve taken particular care to ensure that pretraining works well even with spaCy’s small default architecture sizes, so you don’t have to compromise on efficiency to use it.

Extended match pattern API

Instead of mapping to a single value, token patterns can now also map to a dictionary of properties. For example, to specify that the value of a lemma should be part of a list of values, or to set a minimum character length. It now also supports a REGEX property, as well as set membership via IN and NOT_IN, custom extension attributes via _ and rich comparison for numeric values.

Easy rule-based entity recognition

The EntityRuler is an exciting new component that lets you add named entities based on pattern dictionaries, and makes it easy to combine rule-based and statistical named entity recognition for even more powerful models. Entity rules can be phrase patterns for exact string matches, or token patterns for full flexibility.

Phrase matching with other attributes

By default, the PhraseMatcher will match on the verbatim token text, e.g. Token.text. By setting the attr argument on initialization, you can change which token attribute the matcher should use when comparing the phrase pattern to the matched Doc. For example, LOWER for case-insensitive matches or POS for finding sequences of the same part-of-speech tags.

Retokenizer for merging and splitting

The new Doc.retokenize context manager allows merging spans of multiple tokens into one single token, and splitting single tokens into multiple tokens. Modifications to the Doc’s tokenization are stored, and then made all at once when the context manager exits. This is much more efficient, and less error-prone. Doc.merge and Span.merge still work, but they’re considered deprecated.

Components and languages via entry points

Using entry points, model packages and extension packages can now define their own "spacy_factories" and "spacy_languages", which will be added to the built-in factories and languages. If a package in the same environment exposes spaCy entry points, all of this happens automatically and no further user action is required.

Improved documentation

Although it looks pretty much the same, we’ve rebuilt the entire documentation using Gatsby and MDX. It’s now an even faster progressive web app and allows us to write all content entirely in Markdown, without having to compromise on easy-to-use custom UI components. We’re hoping that the Markdown source will make it even easier to contribute to the documentation. For more details, check out the styleguide and source. While converting the pages to Markdown, we’ve also fixed a bunch of typos, improved the existing pages and added some new content:

• Usage Guide: Rule-based Matching
  How to use the Matcher, PhraseMatcher and the new EntityRuler, and write powerful components to combine statistical models and rules.
• Usage Guide: Saving and Loading
  Everything you need to know about serialization, and how to save and load pipeline components, package your spaCy models as Python modules and use entry points.
• **Usage Guide: ** Merging and Splitting
  How to retokenize a Doc using the new retokenize context manager and merge spans into single tokens and split single tokens into multiple.
• Universe: Videos and Podcasts
• API: EntityRuler
• API: Sentencizer
• API: Pipeline functions

Backwards incompatibilities

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

• While the Matcher API is fully backwards compatible, its algorithm has changed to fix a number of bugs and performance issues. This means that the Matcher in v2.1.x may produce different results compared to the Matcher in v2.0.x.

• The deprecated Doc.merge and Span.merge methods still work, but you may notice that they now run slower when merging many objects in a row. That’s because the merging engine was rewritten to be more reliable and to support more efficient merging in bulk. To take advantage of this, you should rewrite your logic to use the Doc.retokenize context manager and perform as many merges as possible together in the with block.

• The serialization methods to_disk, from_disk, to_bytes and from_bytes now support a single exclude argument to provide a list of string names to exclude. The docs have been updated to list the available serialization fields for each class. The disable argument on the Language serialization methods has been renamed to exclude for consistency.

• The .pos value for several common English words has changed, due to corrections to long-standing mistakes in the English tag map (see issue #593 and issue #3311 for details).

• For better compatibility with the Universal Dependencies data, the lemmatizer now preserves capitalization, e.g. for proper nouns. See issue #3256 for details.

• The built-in rule-based sentence boundary detector is now only called "sentencizer" – the name "sbd" is deprecated.

• The is_sent_start attribute of the first token in a Doc now correctly defaults to True. It previously defaulted to None.

• The keyword argument n_threads on the .pipe methods is now deprecated, as the v2.x models cannot release the global interpreter lock. (Future versions may introduce a n_process argument for parallel inference via multiprocessing.)

• The Doc.print_tree method is now deprecated. If you need a custom nested JSON representation of a Doc object, you might want to write your own helper function. For a simple and consistent JSON representation of the Doc object and its annotations, you can now use the Doc.to_json method. Going forward, this method will output the same format as the JSON training data expected by spacy train.

• The spacy train command now lets you specify a comma-separated list of pipeline component names, instead of separate flags like --no-parser to disable components. This is more flexible and also handles custom components out-of-the-box.

• The spacy init-model command now uses a --jsonl-loc argument to pass in a a newline-delimited JSON (JSONL) file containing one lexical entry per line instead of a separate --freqs-loc and --clusters-loc.

• Also note that some of the model licenses have changed: it_core_news_sm is now correctly licensed under CC BY-NC-SA 3.0, and all English and German models are now published under the MIT license.