This paper discusses aspects of ordering, more specifically probabilistic ordering. An algorithm is presented for adaptive learning an optimal ordering based on data, such that the resulting ordering follows the statistical properties in the input data. The properties to be ordered can be phonological rewrite rules, pronunciation rules inferred from observed phonetic data, grammar rules, etc. The novelty in the proposed model is the combination of well-established methods from three different disciplines: parsing, voting theory, and statistics.

The set-up of this paper is mainly theoretical, and we follow a quite formal approach. There is a close link with techniques that are currently being developed in the Optimality Theory, one of the mainstream approaches in phonology. The resulting techniques, however, can be applied in a more general domain.