Clone wiki

upparse / Home

Welcome to the UPParse Wiki

This software contains efficient implementations of hidden Markov models (HMMs) and probablistic right linear grammars (PRLGs) for unsupervised partial parsing (or UPP, also known as: unsupervised chunking, unsupervised NP identification, unsupervsed phrasal segmentation). These models are particularly effective at noun phrase identification, and have been evaluated at that task using corpora in English, German and Chinese.

In addition, this software package provides a driver script to manage a cascade of chunkers to create full (unlabeled) constituent trees. This strategy produces state-of-the-art unsupervised constituent parsing results when evaluated using labeled constituent trees in English, German and Chinese -- possibly others, those are just the ones we tried.

A description of the methods implemented in this project can be found in the paper "Simple Unsupervised Grammar Induction from Raw Text with Cascaded Finite State Models" by Elias Ponvert, Jason Baldridge and Katrin Erk, to appear in Proceedings of the 49th Annual Meeting of the Association for Computational Linguistics: Human Language Technologies, June, 2011.

The song goes: You down with UPP? YEAH YOU KNOW ME!

Elias Ponvert