Building machines that learn and use language in a similar way than we do

Description

As human beings, we acquire our native languages by taking part in communicative interactions that are meaningful in the environment in which we grow up.

On a fundamental level, there are two main cognitive mechanisms at play during language acquisition: intention reading and pattern finding. Intention reading refers to the capacity to hypothesise about the intended meaning of an observed utterance based on the situational context it is uttered in. Pattern finding refers to the ability to generalise over pairs of observed utterances and their hypothesised meaning, yielding form‐meaning mappings of varying degrees of abstraction. The mechanisms through which humans acquire language are in sharp contrast with the methodologies that currently dominate the field of natural language processing (NLP). Indeed, unlike the linguistic systems of humans, large language models are learnt with an exclusive focus on linguistic forms (characters, words and sentences), in the absence of meaningful and intentional communicative interactions that take place in situated environments.

In this talk, I will discuss recent advances in modelling language acquisition in a more human-like manner, and show how we can operationalise the processes of intention reading and pattern finding in artificial agents.

Période	9 févr. 2023
Titre de l'événement	Women and Girls in Science
Type d'événement	Une conférence
Lieu	Namur, BelgiqueAfficher sur la carte
Niveau de reconnaissance	Local