We augment a task-oriented visual dialogue model with a decision-making module that decides which action needs to be performed next given the current dialogue state, i.e. whether to ask a follow-up question or stop the dialogue. We show that, on the GuessWhat?! game, the new module enables the agent to succeed at the game with shorter and hence less error-prone dialogues, despite a slightly decrease in task accuracy. We argue that both dialogue quality and task accuracy are essential features to evaluate dialogue systems.

Jointly learning to see, ask, decide when to stop, and then GuessWhat / Shekhar, Ravi; Testoni, Alberto; Fernández, Raquel; Bernardi, Raffaella. - ELETTRONICO. - 2481:(2019). (Intervento presentato al convegno CLiC-it 2019 tenutosi a Bari, Italia nel 13th-15th November 2019).

Jointly learning to see, ask, decide when to stop, and then GuessWhat

Shekhar Ravi;Testoni Alberto;Bernardi Raffaella
2019-01-01

Abstract

We augment a task-oriented visual dialogue model with a decision-making module that decides which action needs to be performed next given the current dialogue state, i.e. whether to ask a follow-up question or stop the dialogue. We show that, on the GuessWhat?! game, the new module enables the agent to succeed at the game with shorter and hence less error-prone dialogues, despite a slightly decrease in task accuracy. We argue that both dialogue quality and task accuracy are essential features to evaluate dialogue systems.
2019
Proceedings of the Sixth Italian Conference on Computational Linguistics
Aachen
CEUR
Shekhar, Ravi; Testoni, Alberto; Fernández, Raquel; Bernardi, Raffaella
Jointly learning to see, ask, decide when to stop, and then GuessWhat / Shekhar, Ravi; Testoni, Alberto; Fernández, Raquel; Bernardi, Raffaella. - ELETTRONICO. - 2481:(2019). (Intervento presentato al convegno CLiC-it 2019 tenutosi a Bari, Italia nel 13th-15th November 2019).
File in questo prodotto:
File Dimensione Formato  
clic-it19.pdf

accesso aperto

Tipologia: Versione editoriale (Publisher’s layout)
Licenza: Creative commons
Dimensione 553.19 kB
Formato Adobe PDF
553.19 kB Adobe PDF Visualizza/Apri

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione

Utilizza questo identificativo per citare o creare un link a questo documento: https://hdl.handle.net/11572/250553
Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus 0
  • ???jsp.display-item.citation.isi??? ND
  • OpenAlex ND
social impact