Can Greenbergian universals be induced from language networks?
Kartik Sharma, Kaivalya Swami, Aditya Shete, Samar Husain · 2019
Language networks have been proposed to be the underlying representation for syntactic knowledge (Roelofs, 1992;Pickering and Branigan, 1998).Such networks are known to explain various word order related priming effects in psycholinguistics.Under the assumption that word order information is encoded in these networks, we explore if Greenbergian word order universals (Greenberg, 1963) can be induced from such networks.Language networks for 34 languages were constructed from the Universal Dependencies Treebank (Nivre et al., 2016) based on the assumptions in Roelofs (1992); Pickering and Branigan (1998).We conducted a series of experiments to investigate if certain network parameters can be used to cluster various languages based on the word order typology proposed by Greenberg.Our results show that some network parameters robustly cluster the languages correctly, thereby providing some support for language network as a valid representation for such linguistic generalizations.