Natural-Language-Processing on Ryan Orban

Parsing English with 500 Lines of Python

me@ryanorban.com (Ryan Orban) — Mon, 28 Apr 2014 00:00:00 +0000

Summary

Matthew Honnibal (later the creator of spaCy) wrote this post to demonstrate that dependency parsing — extracting the grammatical structure of a sentence — doesn’t require a massive industrial system. The algorithm he describes is an arc-eager transition-based parser using a perceptron for scoring transitions. It produces good accuracy on standard benchmarks while fitting in a few hundred lines of readable Python.

Natural Language Processing for the Working Programmer

me@ryanorban.com (Ryan Orban) — Mon, 11 Feb 2013 00:00:00 +0000

Summary

“Natural Language Processing for the Working Programmer” was a free online book at nlpwp.org that taught NLP concepts using Haskell as the implementation language. This made it distinctive in a field dominated by Python tutorials: using a functional programming language to teach NLP aligned naturally with the mathematical structures underlying the field (probability distributions, transformations, composition of text processing steps).