Work progress 2006

The partners have successfully completed a range of key tasks in the 3 years of the project (2004 to 2006), both in advancing the theory and developing implementations. Many of these tasks have involved close collaboration between the sites, and a variety of new techniques and ideas have emerged as a result.

Major events and milestones in the final year (2006) were:

 prototype in-car showcase system installed in BMW test car and evaluated with users in real driving conditions 

 use of the extended information states in ISU systems for adaptive presentation planning and generation 

 development of first POMDP dialogue manager 

 testing of learned strategies with real users 

 development of the first Voice Programmable dialogue systems 

 release of TALK annotated data archive 

 TALK test car exhibited at the EC IST 2006 Conference in Helsinki 

 32 papers presented at conferences/workshops: ACL, Interspeech, SLT, EACL, BranDIAL ICSLP, HLT-NAACL, ICMI, LREC, LCM, ICFP, FinTAL, EAMT, AAAI, PAIS 

 advanced lecture course given at ESSLLI 2006 (advanced lecture course on learning and dialogue) 

 4 journal papers published: Recent Trends in Discourse and Dialogue, Research in Computing Science, Procesamiento del Lenguaje Natural 

 14 invited talks: e.g. Institute for Creative Technologies / University of Southern California, France Telecom College Scientifique

The active project tasks in 2006 were in the areas of integrating multiple modalities with SLMs and parsing the output of SLMs (D1.4), the enhanced multimodal grammar library (D1.5), proof-of-concept systems in the in-home domain (D1.6); dynamic interface reconfiguration (D2.2); interfaces for programmable devices and services (D2.3); multimodal turn planning (D3.2), and developing modality specific resources (D3.3); developing a prototype POMDP-based dialogue manager (D4.4); software infrastructure (D5.1) and system integration for the in-car and in-home showcases (D5.2); system evaluations including the showcase in-car dialogue system (D6.4), and the TALK annotated data archive D6.5).
Significant progress was made in each of these active tasks, measurable by the fact that each one has produced at least a deliverable, while many have also resulted in peer-reviewed conference publications and demonstrations. In the final year, strong progress was made in system development and integration (D5.1, D5.2, D4.4), especially for the in-car and smart-house systems and the learned policy systems, system evaluation (D6.4), and theoretical advances have been made in the areas of parsing the outputs of SLMs (D1.4), reconfigurable dialogue systems (D2.3 and D2.3), multimodal turn planning (D3.2), and novel machine learning techniques (D4.4).