About
Constraint weighting systems might be extremely promising for linguistic analysis. But the process of finding weightings by hand is extremely difficult for large systems. This is a huge obstacle to fair assessment.
HaLP can make life easier. We show, in the associated writing, that constraint weighting grammars translate into linear programs, which are solvable using computationally efficient optimization algorithms. We are at present using an implementation of the two-phase simplex method.
Versions 1.0.* of HaLP were written in Perl.
The current version, 2.0, is written in Java. It is a much faster and more robust solver than versions 1.0.* were.
Citation
Potts, Christopher; Joe Pater; Rajesh Bhatt; and Michael Becker. 2007. HaLP 2.0: Harmonic Grammar with Linear Programming. Software package: http://web.linguist.umass.edu/~halp/.
We'll soon release a downloadable version, for processing systems that are too large for the browser interface. If you have a particularly large system that you need to solve, feel free to write to us for a downloadable.
Work on HaLP is supported in part by a UMass Amherst Faculty Research grant to Joe Pater.