Minimum Error Rate Training (MERT) is a method for training the parameters of a loglinear model. One advantage of this method of training is that it can use the large number of hypotheses encoded in a translation lattice as training data. We demonstrate that the MERT line optimisation can be modelled as computing the shortest distance in a weighted finite-state transducer using a tropical polyn...

The purpose of this project is to show observations taken from the Iterative Proportional Scaling algorithm when used with data taken from Hierarchical LogLinear Models. We show with empirical proof that the difference between decomposable and non-decomposable graph models used is not exceedingly different. For the purpose of this project we looked at the properties of the two, three, and four ...

This paper presents the modules in the statistical package ViSta related to categorical data analysis. These modules are: visualization of frequency data with mosaic and bar plots, correspondence analysis, multiple correspondence analysis and loglinear analysis. All these methods are implemented in ViSta with a big emphasis on plots and graphical representations of data, as well as interactivit...

This paper reports the development of loglinear models for the disambiguation in wide-coverage HPSG parsing. The estimation of log-linear models requires high computational cost, especially with widecoverage grammars. Using techniques to reduce the estimation cost, we trained the models using 20 sections of Penn Treebank. A series of experiments empirically evaluated the estimation techniques, ...

An interaction table is a summable square matrix emerging in analysis of inter-citation, international trade, brand-switching, mobility, or input-output industrial data. Three approaches to aggregation of interaction data are theoretically compared: (i) loglinear modeling, (ii) aggregation of Markov chains, and (iii) principle of equivalence in the correspondence analysis. This way an empirical...

Overview: Categorical Data and Graphics Methods for discrete distributions – testing goodness of fit Hanging rootograms Robust distribution plots Methods for two-way frequency tables – understanding association Fourfold displays Sieve diagrams Mosaic displays and loglinear models for n-way tables Mosaic displays Mosaic matrices Correspondence analysis and MCA Logistic and logit regression Logit...