Predicting word order universals

Paola Merlo, University of Geneva, Switzerland


This paper shows a computational learning paradigm to compare and
test theories about language universals. Its main contribution lies in
the illustration of the encoding and comparison of theories about typological
universals to measure the generalisation ability of these theories.
In so doing, this method uncovers hidden dependencies between
theoretical dimensions and primitives that were considered independent
and independently motivated.


frequency; word order universals; Greenberg’s Universal 20; computational modelling; probabilistic models; Naive Bayes

Full Text:



ISSN of the paper edition: 2299-856X