Originally published on March 2nd, 2021.
Given you survived Wednesday evening’s snow blitz, welcome to the third issue of Updates in the CS World (really though, 9 inches of snow overnight?). As promised, this issue we’ll touch on spaCy, a natural language processing library for Python, given its major v3.0 release.
Edu-macating your Computer
For the uninitiated, natural language processing (or NLP) is a procedure in which a computer can derive the meaning and subject matter of prose or other natural language documents. Teaching computers English and semantics… reminds me of Skynet or HAL-9000. Most modern NLP engines use neural networks, and such engines are typically difficult to use and resource intensive. To combat this, spaCy provides a dead-simple API along with pre-trained language models to save your poor rig from the horrors of NLP training. For basic usage examples, see their “101” page (https://spacy.io/usage/spacy-101). You’ll notice on the examples page that something like getting part of speech tags for a document can be accomplished in only a handful of lines. spaCy’s 3.0 release has a few exciting headlines, interoperability with other NLP deep learning models (i.e. Transformers), along with new supporting trained pipelines, and finally a new configuration system for running training systems.
For those even slightly interested, I highly recommend toying with spaCy, you’ll be appalled by how powerful it is despite its ease-of-use. Stay tuned for future issues! •