vpyp

An implementation of Pitman-Yor processes in Python, inspired by cpyp.

Implemented models:

• an n-gram model (Teh, 2006)
• a topic model (LDA with a Pitman-Yor prior instead of a Dirichlet prior)
• an alignment model (variant of the IBM 2 model)

Requirements:

• Python 2.7 (pypy is highly recommended for speed)
• the kenlm module, if you want to use character language models as base distributions