Evolutionary Lexical Functional Grammar

Presenters: Steve Wechsler

BTFD1

In this workshop we will seek explanations for grammatical systems by modeling their acquisition and evolution. We will learn to build and use stochastic models of reinforcement learning, drawing upon techniques in use in psychology since the 1950s. With these models we explore the conditions under which grammatical systems are predicted to emerge and grow in complexity, and the forms of the resulting systems. We posit two types of background condition: message probabilities or patterns of preference for what sorts of messages speakers choose to express in a given context (cp. functionalism); and form probabilities or patterns of preference for how to express a given message (cp. language processing). In addition, language learning involves imitation and is therefore influenced by the learner’s similarity judgments (cp. lexical semantics). Turning to grammar emergence and evolution, the message probabilities influence the emergence of grammatical relations paired with semantic composition rules; the form probabilities influence the emergence of formal expressions of those grammatical relations; and similarity judgments influence the emergence of patterns obtaining across words, such as argument structure generalizations. These three types of grammatical structure are conveniently represented in Lexical Functional Grammar as functional structure, constituent structure, and argument structure, and so the workshop will include an introduction to LFG. We will model the emergence and evolution of a range of grammatical phenomena, including verbal argument structure and alternations, function morphemes that signal constructions, dependent and split ergative case systems, fixed word order expressing grammatical relations, complex predicates, and unbounded dependencies. The goal is for attendees to learn to apply these analytic techniques to constructions of interest to them.

Keywords: Communicative Efficiency, Language Evolution, Computational Modeling, Morphosyntax, Probabilistic Models, Semantics, Syntax, Language Change, Learning, Usage-Based Linguistics, Theoretical Frameworks

Presenters

Steve Wechsler

The University of Texas

Steve Wechsler (Stanford University PhD 1991) serves on the faculty of the University of Texas Department of Linguistics, where he works in syntax and semantics. Topics of past research include the syntax-semantics interface, self-reference by means of personal pronouns and egophoric verb marking systems, grammatical agreement systems, and the language modeling frameworks Lexical functional grammar (LFG) and Head-driven phrase structure grammar. He is a co-author of a leading textbook on LFG, Lexical-Functional Syntax (Bresnan, Asudeh, Toivonen & Wechsler 2016). In recent years he has turned his attention to usage-based and evolutionary approaches to explaining grammatical systems.