Learning implicational models of universal grammar parameters

Abstract

The use of parameters in the description of natural language syntax has to balance between the need to discriminate among (sometimes subtly different) languages, which can be seen as a cross-linguistic version of Chomsky's descriptive adequacy (Chomsky, 1964), and the complexity of the acquisition task that a large number of parameters would imply, which is a problem for explanatory adequacy. Here we first present a novel approach in which machine learning is used to detect hidden dependencies in a table of parameters. The result is a dependency graph in which some of the parameters can be fully predicted from others. These findings can be then subjected to linguistic analysis, which may either refute them by providing typological counterexamples of languages not included in the original dataset, dismiss them on theoretical grounds, or uphold them as tentative empirical laws worth of further study. Machine learning is also used to explore the full sets of parameters that are sufficient to distinguish one historically established language family from others. These results provide a new type of empirical evidence about the historical adequacy of parameter theories.

Topological Analysis of Syntactic Structures

“…We analyze the clustering structure between syntactic parameters by analyzing the persistent connected components of the data at various scales and the resulting tree that follows the order in which the components merge as the scale parameter increases. We compare the detected cluster structure obtained in this way with those discussed in [18] and [31]. We also compute the persistent H 1 and we show that there are further relations between syntactic parameters corresponding to non-trivial persistent H 1 -generators that are not detectable by cluster information.…”

“…When we look at the data points as syntactic features or syntactic parameters, with coordinates given by the values of the parameter over a given set of languages, we focus on the question of identifying relations between these syntactic variables. This is a main open question already investigated by other methods in [18], [31], [33], [38]. We analyze the clustering structure between syntactic parameters by analyzing the persistent connected components of the data at various scales and the resulting tree that follows the order in which the components merge as the scale parameter increases.…”

“…With respect to the second question, the use of syntactic parameters as a method for computational reconstruction of phylogenetic trees of language families was developed in [19], [20], [22], [23] and recently also considered from the perspective of phylogenetic algebraic geometry in [37] and from the topological perspective we consider here in [35]. The first question has been approached with mathematical methods in [18], [27], [31], [33], [38], [41].…”

Phylogenetics of Indo-European Language Families via an Algebro-Geometric Analysis of Their Syntactic Structures

“…The LanGeLin data presented in [23] that we use here include 91 parameters affecting the Determiner Phrases structure. The full list of the LanGeLin syntactic parameters used in this paper is reported in Appendix D, reproduced from Appendix A of [20].…”

“…Unlike the SSWL data, which do not record any explicit relations between the variables, many explicit relations between the Longobardi syntactic parameters are recorded in the LanGeLin data. A more detailed analysis of the relations in the LanGeLin data is given in [20] and in [33]. In our analysis here we have removed those parameters in the LanGeLin data that are explicitly dependent upon the configuration of other parameters.…”