nlup contains some base libraries I use in natural language processing projects. Some highlights:

• confusion.py: classifier evaluation objects
• decorators.py: clever decorators for various purposes
• jsonable.py: a mix-in which allows the state of most objects to be serialized to (and deserialized from) compressed JSON
• perceptron.py: perceptron-like classifiers (binary and multiclass), including some forms of structured prediction
• reader.py: classes and readers for tagged and dependency-parsed data
• timer.py: a with-block that logs wall clock time elapsed

All have been tested on CPython 3.4.1 and PyPy 3.2.5 (PyPy version 2.3.1). They will not work on Python 2 without modification.

Some projects using nlup:

• Detector Morse: simple sentence boundary detection
• Perceptronix Point Never: simple part of speech tagging
• Where's Yr Head At: simple transition-based dependency parsing