Syntactico-semantic generalisation operators for learning large-scale usage-based construction grammars

Since its inception in the mid-eighties, the field of construction grammar has been steadily growing
and constructionist approaches to language have by now become a mainstream paradigm for
linguistic research. Constructionist theories of language consider form-meaning mappings, called
constructions, to be the fundamental building blocks of human languages. Empirical studies have
shown that constructions are learnt through communication, in particular through 'intention reading'
and 'pattern finding' processes that take place in a learner during situated communicative
interactions. This project aims to computationally model these processes and thereby provide for the
first time a methodology for bootstrapping large-scale construction grammars in a usage-based
fashion. We will design algorithms that can generalise over form-meaning pairs, creating abstract
and modular schemata that can be used for language comprehension and production. The results of
this project are expected to have important theoretical and practical implications. Theoretically, the
learnt grammars will provide a unique insight into the compositional and non-compositional aspects
of human languages. Practically, the methodology that we introduce for automatically learning largescale
construction grammars will drastically enhance the application potential of computational
construction grammar.