Spacy-universal-sentence-encoder

Pipeline names now include the `use` prefix, in order to avoid clashes with other libraries
`overwrite_vectors` --> `use_overwrite_vectors`
`save_tfhub_model_url` --> `use_add_model_to_doc`: from containing the URL, to inserting in the doc attributes the wrapper

Extensions:
- `tfhub_model_url` removed
- `use_model` extension holds the tfhub wrapper, no static method and singleton required

Models are built from blank, not from existing ones
- lighter
- can be loaded on pip (https://pypi.org/project/spacy-universal-sentence-encoder/)
- can be composed with any model by using `spacy_universal_sentence_encoder.create_from`

`create_nlp` now takes a config dict and an optional `nlp`. If present, it is extended otherwise a new blank `nlp` object is created.

xx_use_md-0.2.3


xx_use_lg-0.2.3


en_use_md-0.2.3


en_use_lg-0.2.3


xx_use_md-0.2.2


xx_use_lg-0.2.2


en_use_md-0.2.2


en_use_lg-0.2.2


xx_use_md-0.2.1


xx_use_lg-0.2.1


en_use_md-0.2.1


en_use_lg-0.2.1


xx_use_md-0.2.0
SpaCy wrapper for https://tfhub.dev/google/universal-sentence-encoder-multilingual/3

xx_use_lg-0.2.0
SpaCy wrapper for https://tfhub.dev/google/universal-sentence-encoder-multilingual-large/3

en_use_md-0.2.0
SpaCy wrapper for https://tfhub.dev/google/universal-sentence-encoder/4

en_use_lg-0.2.0
SpaCy wrapper for https://tfhub.dev/google/universal-sentence-encoder-large/5

en_use-0.1.3

- version match between wrapper (`en_use` and `universal_sentence_encoder`)
- avoid loading of `tf_wrapper` just when `spacy.load` is called on other models. Now it is instantiated as singleton just when `spacy.load('en_use')` is called
- fix the `Token.vector` that was raising an exception

en_use-0.1.2
added tagger, parser, ner from en_core_web_sm

en_use-0.1.1


en_use-0.1.0
Initial version of English Universal Sentence Encoder