Research Topics

My research is dedicated to one of the most fascinating mysteries of humankind: the origins and evolution of language. In my work, I try to piece together new information of this puzzle by combining techniques from computational linguistics and artificial intelligence that allow me to 'bring back alive' language systems that have disappeared or that have unrecognizably changed over time, and that allow me to investigate how a new language can develop from scratch. Within this broad research context, I spend my time on four concrete research topics:

The Evolution of German Case

German articles

The German case system is often considered to be a "historical accident" because it evolved from a transparent system of form-to-meaning mappings to a seemingly unsystematic and opaque paradigm. But there are many fundamental questions that remain unanswered in this hypothesis. My experiments on the evolution of German case try to shed a new light on this decades-long mystery in linguistics by 'bringing back from the dead' the Old High German case system in a computational model, and then test how that system performs for communication. The surprising answer is that the Modern High German system actually outperforms its predecessor in various linguistic criteria, such as processing and articulatory effort.

Click here to read more...

The Origins of Case Systems

Two embodied agents playing a language game.

There are few linguistic phenomena that have seduced linguists so skilfully as grammatical case has done. Case has claimed a central role in linguistics for centuries, without however revealing its most important secrets. In my case experiments, I try to answer the question how a speech community can self-organize a brand new case system from scratch. These experiments implement usage-based models of language through embodied autonomous agents (robots) that communicate with each other about events in a real-world setting.

Click here to read more...

Fluid Construction Grammar

Grammar

Fluid Construction Grammar (FCG) is a computational grammar formalism for deep language processing that aims at capturing the 'living' aspects of language. It is well-suited for conducting experiments on robust and open-ended language processing, language variation and evolution, and usage-based models of language. Together with my colleagues at Sony CSL Paris and the VUB Artificial Intelligence Laboratory I have been an active developer of the formalism since 2005, with a special focus on argument structure and agreement.

Click here to read more...

Robust Language Learning and Processing

Diagnostics and Repairs

Language users do not follow a rule book. Especially in spoken dialog, linguistic utterances are full of errors (such as hesitations, false starts and disconnected phrases) and novel expressions (such as word play, new or borrowed words and other innovations). Despite these challenges, natural language communication is remarkably fast and robust. My work on robust language processing and learning investigates how language users manage to cope with the open-ended and ever-changing dynamics of language.

Click here to read more...