Towards Empathetic Human-Robot Interactions

Fung, Pascale; Bertero, Dario; Wan, Yan; Dey, Anik; Chan, Ricky Ho Yin; Siddique, Farhad Bin; Yang, Yang; Wu, Chien-Sheng; Lin, Ruixi

doi:10.48550/arxiv.1605.04072

Cited by 3 publications

(2 citation statements)

References 17 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Unlike modularized systems that use a pipeline to connect each module, end-to-end neural systems are designed to train directly on text transcripts and the KB information. Inspired by the success of end-to-end training of chit-chat dialogue systems [32][33][34][35], researchers have recently started exploring end-to-end solutions for task-oriented dialogue systems. RNNs, such as long short-term memory networks [36] (LSTMs) and gated recurrent units [37] (GRUs), play important roles in this direction due to their ability to create latent representations, avoiding the need for artificial labels.…”

Section: Overviewmentioning

confidence: 99%

Learning to Memorize in Neural Task-Oriented Dialogue Systems

Wu¹

2019

Preprint

Self Cite

View full text Add to dashboard Cite

show abstract

Section: Overviewmentioning

confidence: 99%

Learning to Memorize in Neural Task-Oriented Dialogue Systems

Wu¹

2019

Preprint

Self Cite

View full text Add to dashboard Cite

show abstract

“…There has been only minimal computational work on empathy, with different definitions and operationalizations: (Litvak et al 2016;Fung et al 2016;Collins 2014) focus on 'empathetic concern' or the sharing others' emotions in a conversation, and (Xiao et al 2015;Gibson et al 2016;Xiao et al 2016) work on modeling 'empathy' in psychotherapy sessions, defined as the ability of a therapist to attune to the emotions of their clients. Our work is different in that we focus on a specific type of empathy, i.e., pathogenic empathy (Buffone et al in prep) and hence we ground our work in the psychological literature.…”

Section: Background and Related Workmentioning

confidence: 99%

Recognizing Pathogenic Empathy in Social Media

Abdul-Mageed

Buffone

Peng

et al. 2017

ICWSM

View full text Add to dashboard Cite

Empathy is an integral part of human social life, as people care about and for others who experience adversity. However, a specific “pathogenic” form of empathy, marked by automatic contagion of negative emotions, can lead to stress and burnout. This is particularly detrimental for individuals in caregiving professions who experience empathic states more frequently, because it can result in illness and high costs for health systems. Automatically recognizing pathogenic empathy from text is potentially valuable to identify at-risk individuals and monitor burnout risk in caregiving populations. We build a model to predict this type of empathy from social media language on a data set we collected of users’ Facebook posts and their answers to a new questionnaire measuring empathy. We obtain promising results in identifying individuals' empathetic states from their social media (Pearson r = 0.252, p <0.003).

show abstract

Learning to memorize in neural task-oriented dialogue systems

Wu¹

Self Cite

View full text Add to dashboard Cite

Dialogue act annotations are important to improve response generation quality in taskoriented dialogue systems. However, it can be challenging to use dialogue acts to control response generation in a generalizable way because different datasets and tasks may have incompatible annotations. While alternative methods that utilize latent action spaces or reinforcement learning do not require explicit annotations, they may lack interpretability or face difficulties defining task-specific rewards. In this work, we present a novel end-to-end latent dialogue act model (DiactTOD) that represents dialogue acts in a latent space. Diact-TOD, when pre-trained on a large corpus, is able to predict and control dialogue acts to generate controllable responses using these latent representations in a zero-shot fashion. Our approach demonstrates state-of-the-art performance across a wide range of experimental settings on the MultiWOZ dataset, including zeroshot, few-shot, and full data fine-tuning with both end-to-end and policy optimization configurations.

show abstract

Towards Empathetic Human-Robot Interactions

Cited by 3 publications

References 17 publications

Learning to Memorize in Neural Task-Oriented Dialogue Systems

Learning to Memorize in Neural Task-Oriented Dialogue Systems

Recognizing Pathogenic Empathy in Social Media

Learning to memorize in neural task-oriented dialogue systems

Contact Info

Product

Resources

About