Learning from Easy to Complex: Adaptive Multi-Curricula Learning for Neural Dialogue Generation

Cai, Hengyi; Chen, Hongshen; Zhang, Cheng; Song, Yujiang; Zhao, Xianfeng; Li, Yangxi; Duan, Dongsheng; Yin, Dawei

doi:10.1609/aaai.v34i05.6244

Cited by 25 publications

(25 citation statements)

References 16 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Emotion regulation is an influential factor in many problems that blind people suffer from [28]. Interventions that target emotion regulation strategies would be useful [29].…”

Section: Emotion Regulation Strategymentioning

confidence: 99%

“Avatar to Person” (ATP) Virtual Human Social Ability Enhanced System for Disabled People

Guo

Wang

Jin

2021

Wireless Communications and Mobile Computing

View full text Add to dashboard Cite

How to make communication more effective has been underlined unprecedentedly in the artificial intelligence (AI) era. Nowadays, with the improvement of affective computing and big data, people have generally adapted to construct social networks relying on social robots and smartphones. Although the technologies above have been widely discussed and used, researches on disabled people in the social field are still very limited. In particular, facial disabled people, deaf-mutes, and autistic patients are still meeting great difficulty when interacting with strangers using online video technology. This project creates a virtual human social system called “Avatar to Person” (ATP) based on artificial intelligence and three-dimensional (3D) simulation technology, with which disabled people can complete tasks such as “virtual face repair” and “simulated voice generation,” in order to conduct face-to-face video communication freely and confidently. The system has been proven effective in the enhancement of the sense of online social participation for people with disabilities through user tests. ATP is certain to be a unique area of inquiry and design for disabled people that is categorically different from other types of human-robot interaction.

show abstract

“…Emotion regulation is an influential factor in many problems that blind people suffer from [28]. Interventions that target emotion regulation strategies would be useful [29].…”

Section: Emotion Regulation Strategymentioning

confidence: 99%

“Avatar to Person” (ATP) Virtual Human Social Ability Enhanced System for Disabled People

Guo

Wang

Jin

2021

Wireless Communications and Mobile Computing

View full text Add to dashboard Cite

show abstract

“…Shang et al (2018) likewise design a matching network to calibrate the dialogue model training through instance weighting. Cai et al (2020) investigate curriculum learning to adapt the instance effect on dialogue model training according to the sample complexity. Whereas our proposed framework learns to reweight not only the original training examples but also the augmented examples.…”

Section: Related Workmentioning

confidence: 99%

Data Manipulation: Towards Effective Instance Learning for Neural Dialogue Generation via Learning to Augment and Reweight

Cai¹,

Chen²,

Song³

et al. 2020

Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics

Self Cite

View full text Add to dashboard Cite

Current state-of-the-art neural dialogue models learn from human conversations following the data-driven paradigm. As such, a reliable training corpus is the crux of building a robust and well-behaved dialogue model. However, due to the open-ended nature of human conversations, the quality of user-generated training data varies greatly, and effective training samples are typically insufficient while noisy samples frequently appear. This impedes the learning of those data-driven neural dialogue models. Therefore, effective dialogue learning requires not only more reliable learning samples, but also fewer noisy samples. In this paper, we propose a data manipulation framework to proactively reshape the data distribution towards reliable samples by augmenting and highlighting effective learning samples as well as reducing the effect of inefficient samples simultaneously. In particular, the data manipulation model selectively augments the training samples and assigns an importance weight to each instance to reform the training data. Note that, the proposed data manipulation framework is fully data-driven and learnable. It not only manipulates training samples to optimize the dialogue generation model, but also learns to increase its manipulation skills through gradient descent with validation samples. Extensive experiments show that our framework can improve the dialogue generation performance with respect to various automatic evaluation metrics and human judgments. * Work done at Data Science Lab, JD.com. ine ffic ien t co nv er sa tio ns augmentation reweighting eff ec tiv e co nv er sa tio ns au gm en te d eff ec tiv e co nv er sa tio ns

show abstract

“…At training step t, a batch of training samples is obtained from the top f (t) portions of the entire sorted training samples. Following Platanios et al ( 2019) and Cai et al (2020), we define the function f (t) as:…”

Section: Curriculum Plausibilitymentioning

confidence: 99%

CDL: Curriculum Dual Learning for Emotion-Controllable Response Generation

Shen

Feng

2020

Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics

View full text Add to dashboard Cite

Emotion-controllable response generation is an attractive and valuable task that aims to make open-domain conversations more empathetic and engaging. Existing methods mainly enhance the emotion expression by adding regularization terms to standard cross-entropy loss and thus influence the training process. However, due to the lack of further consideration of content consistency, the common problem of response generation tasks, safe response, is intensified. Besides, query emotions that can help model the relationship between query and response are simply ignored in previous models, which would further hurt the coherence. To alleviate these problems, we propose a novel framework named Curriculum Dual Learning (CDL) which extends the emotion-controllable response generation to a dual task to generate emotional responses and emotional queries alternatively. CDL utilizes two rewards focusing on emotion and content to improve the duality. Additionally, it applies curriculum learning to gradually generate high-quality responses based on the difficulties of expressing various emotions. Experimental results show that CDL significantly outperforms the baselines in terms of coherence, diversity, and relation to emotion factors.

show abstract

Learning from Easy to Complex: Adaptive Multi-Curricula Learning for Neural Dialogue Generation

Cited by 25 publications

References 16 publications

“Avatar to Person” (ATP) Virtual Human Social Ability Enhanced System for Disabled People

“Avatar to Person” (ATP) Virtual Human Social Ability Enhanced System for Disabled People

Data Manipulation: Towards Effective Instance Learning for Neural Dialogue Generation via Learning to Augment and Reweight

CDL: Curriculum Dual Learning for Emotion-Controllable Response Generation

Contact Info

Product

Resources

About