Structured Prediction Models via the Matrix-Tree Theorem

This paper provides an algorithmic framework for learning statistical models involving directed spanning trees, or equivalently non-projective dependency structures. We show how partition functions and marginals for directed spanning trees can be computed by an adaptation of Kirchhoff’s Matrix-Tree Theorem. To demonstrate an application of the method, we perform experiments which use the algorithm in training both log-linear and max-margin dependency parsers. The new training methods give improvements in accuracy over perceptron-trained models.

• Type: Conference
• Year: 2007
• Where: EMNLP
• Authors: Terry Koo, Amir Globerson, Xavier Carreras, Michael Collins
• Keywords: Directed Spanning Trees, EMNLP 2007, Kirchhoff’s Matrix-Tree Theorem, Natural Language Processing, Non-projective Dependency Structures