Creativity from averages: the view for AI, with maths from middle school
TLDR: Averaging may produce novel results, and even extreme results
I am a researcher at CNRS, the French center for research (more on institutions). I primarily work in linguistics and psycholinguistics: using language is an every day achievement, and we want to understand this human capacity. The field of linguistics is composed of interacting subfields, such as phonology, morphology, syntax; I specialize in semantics and pragmatics: what are the elementary operations that we deploy to construct the meaning of a series of words? How do we manage to compute and combine these operations with no apparent effort? How do children come to master this complex system? What do other species do of this? How do machines perform?
On a daily basis, we develop formal models of this human system: we use tools from mathematics and computer science to emulate parts of this system. This is called formal semantics, and it inherits from the formal rigor of a century long tradition in philosophy and logic. We test the predictions of these simulations against what humans actually do in the relevant situations. This is psycholinguistics, and it offers numerous methods, from the collection of mere introspective judgments from native speakers, on a small or large scale, to the inspection of what requires what effort (processing time, even if by a couple of hundreds of milliseconds difference, disruption of linguistic abilities by non-linguistic tasks, brain activity, etc.).
Scientific results are evaluated through and distributed in scientific publications. The list of publications below and its informal list of keywords give an idea of what specific linguistic phenomena I have been interested in, and what connex domains beyond traditional linguistics I have worked on.
Links in the entries may lead to more material: scripts for
computations or simulations, experimental material (stimuli, data,
scripts for analyses, pre-registration, etc.), often hosted on
precious scientific archives such as
,
LingBuzz,
,
.
You may pick keywords below to display all (loosely) relevant entries: one of the keywords (or a close associate) may appear in the title, abstract, or metadata of the entry.
Emmanuel Chemla
Ecole Normale Supérieure
Laboratoire de Sciences Cognitives et Psycholinguistique
29 rue d'Ulm
75005 Paris
France
last name at ens.fr
My office is in Batiment Jaurès at Ecole Normale Supérieure.
We often use a meeting room in the Pavillon Jardin, which hosts IJN (see research environment).
My research environment creates unique opportunities to work with numerous people, on numerous questions, using numerous methods. But it is also complicated. Here is my attempt to describe it - still simplifying a bit, apologies.