Description du projet
Apprendre aux machines à interagir avec les humains
La notion de conversation fait référence à des communications interactives entre deux ou plusieurs personnes. Il s’agit d’interactions individuelles et uniques. Les gens adaptent le contenu et la forme des énoncés à leurs différents interlocuteurs (étudiants, parents, collègues, grands-parents) et surveillent leur niveau de compréhension. Dans le domaine de l’informatique et de l’IA, les systèmes de traitement automatique du langage naturel (TALN) permettent aux humains de parler aux machines, mais les interactions de ces systèmes TALN manquent encore de naturel. C’est dans ce contexte que le projet IDDISC, financé par l’UE, ouvrira la voie à des interactions linguistiques individualisées avec les systèmes informatiques. Il s’intéressera en particulier aux différences de compréhension individuelles au niveau pragmatique et discursif. Le projet développera des modèles innovants pour prédire les déductions basées sur les propriétés cognitives spécifiques d’un individu et sa connaissance du domaine en question.
Objectif
Humans adapt the content and form of their utterances to different interlocutors (students vs. colleagues vs. granny), and monitor the level of understanding in their conversational partner. Today's NLP systems are however largely blind with respect to individual variation in language comprehension, which in turn leads to misunderstandings and lack of naturalness in the interaction.
The vision of IDDISC is to enable individualised language interaction with computer systems, such that information or explanations generated by a system will fit the user and the situation, by explicitly modelling their state of understanding. This project will break completely new ground by addressing individual differences in comprehension at the pragmatics and discourse level, i.e. with respect to the inferred meaning that goes beyond the literal meaning of an utterance.
This vision requires ground-breaking contributions at the intersection of individual differences research, language processing models and statistical methods: (a) we will develop innovative models that can predict inferences based on specific cognitive properties of an individual and their domain knowledge; (b) we will undertake foundational research on the factors that lead to individual differences in pragmatic inferences; (c) we will contribute to new statistical modelling techniques for quantifying similarities between individuals, as well as new methods for modelling inference variation in the NLP pipeline; (d) we will test the success of adaptation to individual characteristics in practical applications.
This project will make it possible to reduce the risk of misunderstandings, and enable adaptation of automatically generated language (e.g. explanations, summaries) to specific users. The new statistical methods and crowd-sourcing paradigms developed as part of this project will open the door to other researchers for investigating individual differences in all areas of language processing.
Champ scientifique
Programme(s)
Thème(s)
Régime de financement
ERC-STG - Starting GrantInstitution d’accueil
66123 Saarbrucken
Allemagne