Despite important progress, conversational systems often generate dialogues that sound unnatural to humans. We conjecture that the reason lies in their different training and testing conditions: agents are trained in a controlled “lab” setting but tested in the “wild”. During training, they learn to generate an utterance given the human dialogue history. On the other hand, during testing, they must interact with each other, and hence deal with noisy data. We propose to fill this gap by training the model with mixed batches containing both samples of human and machinegenerated dialogues. We assess the validity of the proposed method on GuessWhat?!, a visual referential game.

Overprotective Training Environments Fall Short at Testing Time: Let Models Contribute to Their Own Training / Testoni, Alberto; Bernardi, Raffaella. - ELETTRONICO. - 2769:(2020). (Intervento presentato al convegno CLiC-it 2020 tenutosi a Bologna, Online nel 1-3 Marzo 2021).

Overprotective Training Environments Fall Short at Testing Time: Let Models Contribute to Their Own Training

Testoni, Alberto;Bernardi, Raffaella
2020-01-01

Abstract

Despite important progress, conversational systems often generate dialogues that sound unnatural to humans. We conjecture that the reason lies in their different training and testing conditions: agents are trained in a controlled “lab” setting but tested in the “wild”. During training, they learn to generate an utterance given the human dialogue history. On the other hand, during testing, they must interact with each other, and hence deal with noisy data. We propose to fill this gap by training the model with mixed batches containing both samples of human and machinegenerated dialogues. We assess the validity of the proposed method on GuessWhat?!, a visual referential game.
2020
Proceedings of the Seventh Italian Conference on Computational Linguistics
Aachen
CEUR
Testoni, Alberto; Bernardi, Raffaella
Overprotective Training Environments Fall Short at Testing Time: Let Models Contribute to Their Own Training / Testoni, Alberto; Bernardi, Raffaella. - ELETTRONICO. - 2769:(2020). (Intervento presentato al convegno CLiC-it 2020 tenutosi a Bologna, Online nel 1-3 Marzo 2021).
File in questo prodotto:
File Dimensione Formato  
paper_24.pdf

accesso aperto

Tipologia: Versione editoriale (Publisher’s layout)
Licenza: Creative commons
Dimensione 506.22 kB
Formato Adobe PDF
506.22 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/11572/288538
Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus 0
  • ???jsp.display-item.citation.isi??? ND
social impact