Weighted Tree Automata and Tree Transducers

Weighted tree automata and tree transducers extend the classical model of tree automata and tree transducers. An unweighted tree automaton calls a binary decision for each input tree: either the automaton accepts the tree or it rejects the tree. In the weighted case, a quantitative decision can be made (e.g. with which confidence can the input tree be accepted; how often does a certain symbol occur in the input tree; etc.). Such a quantitative decision is useful, for example, in natural language processing, where several accepted input trees can now be ordered according to their weight (quality, etc.).

We are interested in the theoretical foundations of weighted tree automata and tree transducers. In particular, we did and still do research on closure properties, determinization and minimization, learning of tree automata, and equivalent logical formalisms.

Slides about the topic

• slides presented at ATANLP 2012 by Heiko Vogler
• slides presented at Theorietag October 3, 2012 in Prague by Heiko Vogler
• slides presented at DLT 2013 in Paris by Heiko Vogler

Contact

• Prof. Dr.-Ing. habil. Dr. h.c./Univ. Szeged
  Heiko Vogler
  Phone: +49 (0) 351 463-38232