Deliverables

The first public project deliverables are now available.
The currently planned milestones and deliverables are listed below. Non-restricted documents will be made available after approval by the Commission.

CLASSIC Project Milestones and Deliverables

Month 6

•   Requirements analysis and theory for statistical learning approaches in automata-based dialogue management:  D1.1.1 (report)

•   Initial communication architecture, protocols, and module interfaces:  D5.1.1 (report, restricted to programme participants)

Month12

• Training tools and semantic decoder for the TownInfo task: D2.1  (prototype)

• Adaptive automata-based DM component, first prototype:  D1.1.2 (prototype, restricted to programme participants)

• Improved POMDP-based dialogue system:  D1.4.1 (prototype)

• Semantic decoder which exploits syntactic-semantic parsing, for the TownInfo task: D2.2 (prototype)

• Shared dialogue context model (XML schema):  D3.1 (report)

• Re-ranking alternative inputs to the speech synthesizer, Self-Help system:  D4.1 (prototype, confidential)

• Initial CLASSIC “TownInfo” system (“System 1”) :  D5.2.1 (prototype, restricted to programme participants)

• Initial France Telecom industrial plat form adapted to CLASSIC architecture (“System 3”): D5.3.1 (prototype, restricted to programme participants)

Month 15

M13 Initial evaluation of CLASSIC TownInfo and Self-Help systems:  D6.3 (report)

Month 18

M14 “Self-help” Data Collection for Task Oriented Dialogue:  D6.1 (report)

M15 New frameworks for generalization, with implementation in DIPPER ISU DM system:  D1.2 (prototype)

M16 Semantic parser for French built using one for English:  D2.3 (prototype)

M17 User simulations for training in the (PO)MDP framework including the simulation of grounding: D3.2 (prototype)

M18 Optimal Wizard NLG behaviours in context:  D4.2 (report)

M19 Generating expressive TTS variants, Self-Help system:  D4.3 (prototype)

Month 24

M20 POMDP learning for ISU dialogue management: D1.3 (prototype) 

M21 Summary-based policy optimisation: D1.4.2 (report)

M22 Agenda-based user simulations and EM training tools for TownInfo and Self-Help domains:  D3.4 (prototype)

M23 Final communication architecture, protocols, and module interfaces:  D5.1.2 (report)

Month 30

M24 Three trained semantic decoders for the Self-Help task:  D2.3 (prototype)

M25 Proof-of-concept CLASSIC Self-Help system (“System 2”):  D5.4 (prototype)

D3.6 User simulations of different types of Self-Help user

Month 36

D1.5 Report on online adaptation of dialogue systems

D2.5 Improved models of syntactic-semantic parsing, for English and French

D4.5 Testing learned NLG and TTS policies with real users, in Self-Help systems

D6.4 Final evaluation of CLASSIC TownInfo and Self-Help systems

D6.5 Annotated data archive