Dialogue-based Navigation with Multivariate Adaptation driven by Politeness

Chandrakant Bothe, Fernando Garcia, Arturo Cruz Maya, Amit Kumar Pandey, Stefan Wermter

Click here to see the log of dialogues in the demo video

in_utt = input utterance;
in_slu = input utterance spoken language understanding output (dialogua act and slot-value pairs);
in_dop = input degree of politeness;
op_utt = output utterance (spoken by a robot);
INPUT_UTT = processed input utterance number (it is the utterance number in the graph);
NOT_INPUT = not processed input utterance (it is been used as a context only)
Polite Dialogue Impolite Dialogue
---------- dialogue.log_on_14_08_18_at_15_50_07 ----------
-<>- turn : 1
---- in_utt : "hello how are you?" -->INPUT_UTT == 1
x--- in_slu : ('Greeting', 1.0, {})
x--- in_dop : 1 accumulative DoP : 1
---> op_utt : Hello, I am fine, thanks, how can I help you?
-<>- turn : 2
---- in_utt : "could you please show me the education" -->INPUT_UTT == 2
x--- in_slu : ('TakeToPlace', 1.0, {'slot_value': ['room', 'education']})
x--- in_dop : 1 accumulative DoP : 2
---> op_utt : Please follow me, I can show you the education
-<>- turn : 3
---- in_utt : could you please show me the education ---NOT_INPUT
x--- in_slu : ('FinishedOne', 1.0, {'slot_value': ['room', 'education']})
x--- in_dop : 0 accumulative DoP : 2
---> op_utt : We arrived at education department, here you can find the education related stuff. Would you like to visit next department?
-<>- turn : 4
---- in_utt : "yes I would like to visit" -->INPUT_UTT == 3
x--- in_slu : ('Accept', 0.5653610591806222, {})
x--- in_dop : 1 accumulative DoP : 3
---> op_utt : Okay lets go to the next department.
-<>- turn : 5
---- in_utt : yes I would like to visit ---NOT_INPUT
x--- in_slu : ('TakeToPlace', 1.0, {'slot_value': ['room', 'tourism']})
x--- in_dop : 0 accumulative DoP : 3
---> op_utt : Please follow me, I can take you to the tourism
-<>- turn : 6
---- in_utt : yes I would like to visit ---NOT_INPUT
x--- in_slu : ('FinishedOne', 1.0, {'slot_value': ['room', 'tourism']})
x--- in_dop : 0 accumulative DoP : 3
---> op_utt : We arrived at tourism department, here you can find the tourism related stuff. Would you like to visit next department?
-<>- turn : 7
---- in_utt : "no thank you" -->INPUT_UTT == 4
x--- in_slu : ('Thanking', 1.0, {})
x--- in_dop : 0 accumulative DoP : 3
---> op_utt : You are welcome, thanks for bearing with me.
---------- dialogue.log_on_14_08_18_at_16_24_24 ----------
-<>- turn : 1
---- in_utt : "hello" -->INPUT_UTT == 1
x--- in_slu : ('Greeting', 1.0, {})
x--- in_dop : 0 accumulative DoP : 0
---> op_utt : Hello, how can I help you?
-<>- turn : 2
---- in_utt : "show me education department" -->INPUT_UTT == 2
x--- in_slu : ('TakeToPlace', 1.0, {'slot_value': ['room', 'education']})
x--- in_dop : -1 accumulative DoP : -1
---> op_utt : Please follow me.
-<>- turn : 3
---- in_utt : show me education department ---NOT_INPUT
x--- in_slu : ('FinishedOne', 1.0, {'slot_value': ['room', 'education']})
x--- in_dop : 0 accumulative DoP : -1
---> op_utt : Here is the education department, here you can find the stuff related to education. Do you wish to visit next department?
-<>- turn : 4
---- in_utt : "show me the tourism" -->INPUT_UTT == 3
x--- in_slu : ('TakeToPlace', 1.0, {'slot_value': ['room', 'tourism']})
x--- in_dop : -1 accumulative DoP : -2
---> op_utt : Please follow me.
-<>- turn : 5
---- in_utt : show me the tourism ---NOT_INPUT
x--- in_slu : ('FinishedOne', 1.0, {'slot_value': ['room', 'tourism']})
x--- in_dop : 0 accumulative DoP : -2
---> op_utt : We arrived at tourism department, here you can find the tourism related stuff. Would you like to visit next department?
-<>- turn : 6
---- in_utt : "no" -->INPUT_UTT == 4
x--- in_slu : ('Reject', 1.0, {})
x--- in_dop : -1 accumulative DoP : -3
---> op_utt : Okay.
-<>- turn : 7
---- in_utt : "bye" -->INPUT_UTT == 5
x--- in_slu : ('ConvClosing', 1.0, {})
x--- in_dop : 0 accumulative DoP : -3
---> op_utt : Bye bye, have a nice day.

SECURE Secondment Project by Chandrakant Bothe

Towards Dialogue-based Navigation with Multivariate Adaptation based on Intention and Politeness for Social Robots
arXiv:1809.07269
Accepted at the International Conference on Social Robotics (ICSR 2018)

The secondment is accomplished with extensive experiments during July-August '18 in collaboration with the industrial partner SoftBank Robotics in Paris, France with their semi-humanoid robot Pepper.

Any further deatails/updates/news about this project will be posted on this page.