Learning Stochastic Tree Edit Distance

Marc Bernard (1) , Amaury Habrard (2) , Marc Sebban (1)

1 LHC - Laboratoire Hubert Curien
2 LIF - Laboratoire d'informatique Fondamentale de Marseille - UMR 6166

Trees provide a suited structural representation to deal with complex tasks such as web information extraction, RNA secondary structure prediction, or conversion of tree structured documents. In this context, many applications require the calculation of similarities between tree pairs. The most studied distance is likely the tree edit distance for which improvements in terms of complexity have been achieved during the last decade. However, this classic edit distance usually uses a priori fixed edit costs which are often difficult to tune, that leaves little room for tackling complex problems. In this paper, we focus on the learning of a stochastic tree edit distance. We use an adaptation of the expectation-maximization algorithm for learning the primitive edit costs. We carried out several series of experiments that confirm the interest to learn a tree edit distance rather than a priori imposing edit costs.

Mots clés: Stochastic tree edit distance; EM algorithm; generative models; discriminative models

Domaines: Apprentissage [cs.LG]

Marc Bernard, Amaury Habrard, Marc Sebban. Learning Stochastic Tree Edit Distance. 17th European Conference on Machine Learning, Sep 2006, Berlin, Germany. pp.42-53. ⟨ujm-00109696⟩