Kontinuální plánování pro zajištění vzájemného porozumění v situovaném dialogu

• Thesis title in English: CONTINUOUS PLANNING FOR COMMUNICATIVE GROUNDING IN SITUATED DIALOGUE
• Academic year of topic announcement: 2008/2009
• Thesis type: diploma thesis
• Thesis language: angličtina
• Department: Institute of Formal and Applied Linguistics (32-UFAL)
• Supervisor: Ing. Ivana Kruijffová, Dr.
• Author: hidden - assigned and confirmed by the Study Dept.
• Date of registration: 13.11.2008
• Date of assignment: 21.11.2008
• Date and time of defence: 10.05.2010 00:00
• Date of electronic submission: 10.05.2010
• Date of proceeded defence: 10.05.2010
• Opponents: prof. PhDr. Eva Hajičová, DrSc.

Guidelines

Teoretická část:

Získat základní přehled v oblasti modelování dialogu, zejména co se týče
přístupu k managementu dialogu a reprezentaci kontextu. Získat podrobný
přehled v problematice zajištění vzájemného porozumění a modelování
tohoto procesu prostřednictvím verifikačních a klarifikačních sub-dialogů v
existujících dialogových systémech. Vyhodnotit vhodnost a použitelnost
stávajících modelů pro použití v situovaném systému. Navrhnout potřebné
úpravy, případně rozšířit stávající modely o aspekty situované
komunikace.

Praktická část:

Seznámit se s architekturou systemu COSY-Explorer a zde používanými
nástroji. Zejména CAST (The CoSy Architecture Schema Toolkit), DFKI
modul pro komunikaci v přirozeném jazyce (COMSYS) a Brennerův plánovač včetně
simulačního prostředí (MAPSIM). Integrovat plánovač v rámci modulu
COMSYS.
Implementovat verifikační a klarifikační strategie prostřednictvím
plánovacích operátorů, s využitím možnosti kontinuálního plánování.
Realizovat rozhraní mezi plánovačem a ostatními částmi modulu COMSYS
zajišťujícími interpretaci vstupu a generování výstupu. Podle potřeby
realizovat aktualizaci reprezentace kontextu.

References

Cahn, J. E., & Brennan, S. E. (1999). A psychological model of
grounding and repair in dialog. Proceedings, AAAI Fall Symposium on
Psychological
Models of Communication in Collaborative Systems (pp. 25-33). North
Falmouth, MA: American Association for Artificial Intelligence.

David R. Traum Computational Models of Grounding in Collaborative
Systems, in working notes of AAAI Fall Symposium on Psychological
Models of
Communication, p. 124-131, November, 1999. PS

Matthew Purver. CLARIE: Handling Clarification Requests in a Dialogue
System. Research on Language and Computation 4(2-3), pages 259-288,
October 2006.

Michael Brenner and Kruijff-Korbayova I. 2008. A Continual Multiagent
Planning Approach to Situated Dialogue. Proceedings of LonDial'08 -- the
12th Workshop on the Semantics and Pragmatics of Dialogue (SEMDIAL
series). London, UK.

Preliminary scope of work

The domain for the thesis is spoken situated dialogue for human-robot
interaction. Characteristic for such dialogue, as with any spoken
dialogue, is that the involved dialogue participants may fail to understand each
other. This may regard what is being said, or how what is said refers
to the situation(s) the dialogue is about. To resolve uncertainty and/or
break-downs in conversation, participants typically resort to phrasing
clarification requests.

The goal of the thesis is to develop a continuous planning-based
approach to handle clarification for communicative grounding. Planning will
enable the formulation of a dialogue plan to address a clarification need, and
monitor the execution of that plan. (We assume this need may be raised by the
robot or by a human user.) Should the plan (or the dialogue based on it)
fail to fully address the need, the continuous nature of the approach will
make it possible to adapt the plan until the intended goal has been achieved.
The resulting steps in the dialogue plan should provide input to a existing
utterance content planner. Within the scope of this thesis, resources
will be formulated for this content planner to turn dialogue plan action
steps into logical forms of utterance content.

The work for this thesis will be based in existing approaches to
clarification and communicative grounding. Theoretically, the thesis
is to extend this work to cover situated dialogue. Practically, the thesis
work is to yield an implementation integrated into an existing system for spoken
situated dialogue in human-robot interaction.

Preliminary scope of work in English

The domain for the thesis is spoken situated dialogue for human-robot
interaction. Characteristic for such dialogue, as with any spoken
dialogue, is that the involved dialogue participants may fail to understand each
other. This may regard what is being said, or how what is said refers
to the situation(s) the dialogue is about. To resolve uncertainty and/or
break-downs in conversation, participants typically resort to phrasing
clarification requests.

The goal of the thesis is to develop a continuous planning-based
approach to handle clarification for communicative grounding. Planning will
enable the formulation of a dialogue plan to address a clarification need, and
monitor the execution of that plan. (We assume this need may be raised by the
robot or by a human user.) Should the plan (or the dialogue based on it)
fail to fully address the need, the continuous nature of the approach will
make it possible to adapt the plan until the intended goal has been achieved.
The resulting steps in the dialogue plan should provide input to a existing
utterance content planner. Within the scope of this thesis, resources
will be formulated for this content planner to turn dialogue plan action
steps into logical forms of utterance content.

The work for this thesis will be based in existing approaches to
clarification and communicative grounding. Theoretically, the thesis
is to extend this work to cover situated dialogue. Practically, the thesis
work is to yield an implementation integrated into an existing system for spoken
situated dialogue in human-robot interaction.