Betekenis is van belang: Syntactico-semantische generalisatieoperatoren voor het leren van computationele constructiegrammatica's

Children acquire their native languages in a constructivist manner. As they take part in communicative interactions with their caregivers, they gradually build up an inventory of productive schemata, called constructions, that capture all linguistic knowledge that is needed for language comprehension and production. In this dissertation, I introduce a computational methodology through which artificial agents can build up such an inventory of constructions. This objective is divided into two essential stages. In the first stage, I present a range of generalisation operators that can distil construction grammars from utterance-meaning pairs. In the second stage, the meaning representations of utterances are no longer directly observable. Instead, an agent needs to abductively construct the interlocutor's intention from the communicative situation that it takes part in. Here, I demonstrate how the generalisation operators from the first stage serve to reduce the search space that is faced by this intention reading process to a computationally tractable size. The work that I present in this dissertation contributes to the broader goal of modelling human-like language acquisition in machines. The scientific contribution of this research is twofold. On the one hand, it provides computational evidence for the cognitive plausibility of usage-based theories of language acquisition by introducing a mechanistic model of the acquisition of construction grammars from scratch. On the other hand, the techniques that I introduce pave the way for learning computationally tractable, large-scale, usage-based construction grammars that facilitate both language comprehension and production. Apart from their theoretical importance, such grammars are also highly valuable for a large range of application domains, including intelligent tutoring systems, conversational agents and the semantic analysis of discourse.