Skip to Main content Skip to Navigation

Contribution à l'apprentissage dans le dialogue homme-machine

Abstract : The main goal of the learning process in the human-machine dialog is to increase the interaction and adaptation capabilities of the computer to the user, mainly with respect to the acquisition of new tasks. In the framework of the finalized cooperative dialog, the task gives structure to the dialog and it is through this dialog that the what-to-do and how-to-do are established. In this thesis, we consider the computer as a collaborator in the completion of a task, capable to learn and recognize later the learned situations. We present a dialog model centered on the learning notion of knowing-how-to-do a task. We propose a dialog model independent of the task, in which the context is defined by the application, but where the system does not know a-priori all the possible tasks that under this context could be performed. The machine learns how to perform a task and from this new knowledge, can participate when other analog situations are found. The proposed model is based on the mental states of the user. These mental states express the beliefs and intentions of the user, and through them the machine can interpret and produce its answers. The interchanges between user and machine are formalized in a modal logic framework based on action. This logic contains elements of an epistemic logic (from knowledge representation), a logic of action (to describe actions and their effects), and a dialog logic (to express commitments, dialog interruptions and misunderstandings). In this way, our model integrates (i) the management of interchanges during the dialog (performed either with the help of the language or by direct manipulation), (ii) the coordination of actions obtained from plan recognition, and (iii) the construction of common knowledge through new task acquisition. From a computational point of view, we implemented an adaptive ICPdraw system that has enabled us to validate our methodology.
Document type :
Complete list of metadata

Cited literature [68 references]  Display  Hide  Download
Contributor : Thèses Imag <>
Submitted on : Wednesday, February 18, 2004 - 6:07:06 PM
Last modification on : Friday, November 6, 2020 - 4:13:02 AM
Long-term archiving on: : Friday, April 2, 2010 - 7:27:30 PM


  • HAL Id : tel-00004861, version 1




Luis Villasenor-Pineda. Contribution à l'apprentissage dans le dialogue homme-machine. Autre [cs.OH]. Université Joseph-Fourier - Grenoble I, 1999. Français. ⟨tel-00004861⟩



Record views


Files downloads