GLU - A Gradient Model for Language Universals. A Computational Revisitation of Greenberg Universals

Identificador:  TDX:4402

Handle: https://hdl.handle.net/20.500.11797/TDX4402

Autores:  Brosa Rodríguez, Antoni

Resumen:
This doctoral thesis is situated in Linguistic Typology and proposes an innovative approach to classify typological Universals, focusing on the presence of exceptions. Fuzzy Logic is used to understand Universality as a continuum, and a computational approach to Greenberg Universals is presented. The problem addressed concerns exceptions to Universals and the need for more clarity in relevance and gradual transition in statistical Universals. The justification for the research is based on the revolution in Linguistic Typology introduced by Greenberg. Linguists tried to improve on Greenberg's results, but interest waned at the end of the 20th century. Recently, however, the rise of Computational Linguistics and access to linguistic corpora have revitalized the study of typological Universals. The methodology adopted involves formalizing Greenberg Universals and verifying them in 143 languages using Universal Dependencies. Fuzzy Logic is employed for a classification of Universality. In addition, Grew-Match syntax and UD labels are used for formalization and structure search in linguistic corpora.