, G is more specific than all the other hypotheses G 0 that also cover the data

, G forbids structures S that are substructures of structures S 0 forbidden by other grammars G 0 that also satisfy (1) and (2), i.e

, Conclusion Feature-based phonological constraints are naturally structured in a way that provides a generality relation which learners can use to effectively prune and search the hypothesis space

. Luc-de-raedt, Logical and Relational Learning, 2008.

B. Hayes and C. Wilson, A maximum entropy model of phonotactics and phonotactic learning, Linguistic Inquiry, vol.39, pp.379-440, 2008.

J. Heinz, Learning long-distance phonotactics, Linguistic Inquiry, vol.41, issue.4, pp.623-661, 2010.

J. Heinz, String extension learning, Proceedings of the 48th Annual Meeting of the Association for Computational Linguistics, pp.897-906, 2010.

M. H. Vu, A. Zehfroosh, K. Strother-garcia, M. Sebok, H. G. Tanner et al., Statistical relational learning with unconventional string models. Frontiers in Robotics and AI, 2018.

C. Wilson and G. Gallagher, Constraint complexity in surface-based phonotactics: A case study of South Bolivian Quechua, Linguistic Inquiry, 2018.