ADaptive And Meaning mACHines (2006-2010)

Marie Curie Excellence Grants (EXT)
The ‘Adaptive and Meaning Machines’ project aims at advancing the science of human-machine spoken communication. In state-of-the-art system, machines have been designed either with a-priori handcrafted knowledge or trained on ‘frozen’ statistical models. In retrospect this is a strong constraint, given the on-line nature of human-machine dialog which requires ontime feedback, and not a pre-defined user-independent approach. In this project the team leader plans to build a highly skilled research team that will build from state-of-the-art research technology in Spoken Dialog System (SDS) and bring innovation to human-machine communication by defining a machine learning approach to adaptive machines. This will be accomplished by defining and inventing algorithms for on-line decoding of acoustic and language models for speech recognition and understanding. The project will include theoretical and experimental research work as well as an installation of research SDS prototype that will be used as a vehicle for training, testing and evaluating the speech and language processing algorithms that will be investigated. The project will involve interdisciplinary work that will span from signal processing to computer science to machine learning.

ADAMACH, Marie Curie Fellow Team

Giuseppe Riccardi, A. Ivanov, Silvia Quarteroni, A. Sporka, Sebastian Varges
Talks