PASCAL - Pattern Analysis, Statistical Modelling and Computational Learning

Exponential Families for Conditional Random Fields
Yasemin Altun, Alex Smola and Thomas Hofmann
In: 21th International Conference on Machine Learning (ICML), 2004(2004).

Abstract

Many real-world classification tasks involve the prediction of multiple, inter-dependent class labels. A prototypical case of this sort deals with sequences of observations for which a corresponding sequence of labels needs to be predicted. Such problems arise naturally in the context of annotating and segmenting observation sequences. This paper generalizes Gaussian Process classification to deal with these types of problems in a way that allows to take dependencies between (neighboring) labels into account. Our approach is motivated by the desire to retain a rigorous probabilistic semantics, while overcoming limitations of parametric methods like Conditional Random Fields, which exhibit conceptual and computational difficulties in high-dimensional input spaces. Experiments on Named Entity Recognition and Pitch Accent prediction tasks demonstrate the competitiveness of our approach.

PDF - Requires Adobe Acrobat Reader or other PDF viewer.
EPrint Type:Conference or Workshop Item (Talk)
Project Keyword:Project Keyword UNSPECIFIED
Subjects:Learning/Statistics & Optimisation
Natural Language Processing
Theory & Algorithms
ID Code:710
Deposited By:Adam Kowalczyk
Deposited On:02 January 2005