Language Universals: Their Discovery and Verbalization by Machine

Abstract

The task of discovering language universals from data is computationally complex and hence difficult to solve manually. To handle this problem I introduce a computational system that can discover all universals in a given database and verbalize them in English. Additionally, if this database has been previously analyzed by a human, the system can compare the discoveries of the human and machine agent and produce a whole article with its results. The system is illustrated on Greenberg's data on word order and the phonological UPSID-451 database. Finally, the system is evaluated