disable ALL pipeline components for tokenizing, even unknown ones

writer / replaCy

spaCy match and replace, maintaining conjugation

https://pypi.org/project/replacy/

MIT License

34 stars 8 forks source link

Closed sam-writer closed 4 years ago

sam-writer commented 4 years ago

Without pulling out scorer, this is the only way I can make a neuralcoref replaCy extension play nice with kenLM scoring

because neuralcoref is a pipeline component, and it freaks out if parser is disabled.