Enhancing Dialog Coherence with Event Graph Grounded Content Planning

IJCAI, pp. 3941-3947, 2020.

Cited by: 0|Bibtex|Views25|Links
EI
Keywords:
multi turn open domainopen domain conversation generationcoherent dialogexperimental resultevent graphMore(11+)
Weibo:
Experimental results on two datasets show that event graph grounded RL framework can generate a more coherent dialog with appropriate content ordering when compared with baselines

Abstract:

How to generate informative, coherent and sustainable open-domain conversations is a non-trivial task. Previous work on knowledge grounded conversation generation focus on improving dialog informativeness with little attention on dialog coherence. In this paper, to enhance multi-turn dialog coherence, we propose to leverage event chains t...More

Code:

Data:

0
Introduction
  • One of the key goals of AI is to build a machine that can converse with humans by generating informative, coherent and sustainable open-domain conversations
  • To achieve this goal, end-to-end neural generative models have been studied [Ritter et al, 2011; Shang et al, 2015].
  • These models tend to produce generic responses
  • To address this issue, some work propose to generate responses by grounding on external knowledge [Dinan et al, 2019; Ghazvininejad et al, 2018; Zhou et al, 2018].
  • The authors make a step towards coherent and informative multi-turn open-domain conversation generation
Highlights
  • One of the key goals of AI is to build a machine that can converse with humans by generating informative, coherent and sustainable open-domain conversations
  • We present a novel event graph grounded Reinforcement Learning framework (EGRL)
  • Event graph grounded RL framework significantly outperforms all baselines in terms of all the metrics except for length-of-dialog
  • Results on Twitter Corpus As shown in Table 3, our model significantly outperforms the baselines in terms of dialog coherence, engagement and informativeness except for length-of-dialog
  • We present an event graph grounded Reinforcement Learning framework (EGRL) to demonstrate how the knowledge of event chains can help plan a multi-turn open-domain dialog
  • If two events share no less than 80% words, they will be merged into one event
  • Experimental results on two datasets show that event graph grounded RL framework can generate a more coherent dialog with appropriate content ordering when compared with baselines
Methods
  • Informativeness Info.* Dist-1/2# 0.18 0.07/0.21 0.56 0.12/0.44 0.47 0.10/0.41 0.48 0.12/0.47 0.81 0.27/0.69 0.79 0.23/0.61 0.59 0.17/0.51 0.74 0.25/0.66 Overall Quality.
  • The authors use Dist-1 and Dist-2 to measure the diversity of generated responses.
  • (8) User-interests consistency (User-Cons.) for overall quality: The metric is used to evaluate if a model can follow a new topic mentioned by a user.
  • A dialog will be rated “1” if the model follows the user’s new topic, otherwise “0”
Results
  • The authors invite three annotators to evaluate each dialog from each model. System identifiers are masked during evaluation.
  • As shown in Table 1, EGRL significantly outperforms all baselines in terms of all the metrics except for length-of-dialog
  • It demonstrates that EGRL can effectively foster a more coherent, informative, engaging conversation.
  • The authors' model obtains the best user-interests consistency result compared with baselines
  • It indicates that with EGRL, the model avoids one-sided conversation while focusing on dialog coherence.
  • As shown in Table 3, the model significantly outperforms the baselines in terms of dialog coherence, engagement and informativeness except for length-of-dialog
  • It is consistent with Table 1.
  • The ordering of the dialog content by the model is more appropriate, e.g., “tooth decay” → “emergency surgery” → “recover”
Conclusion
  • The authors present an event graph grounded RL framework (EGRL) to demonstrate how the knowledge of event chains can help plan a multi-turn open-domain dialog.
  • Experimental results on two datasets show that EGRL can generate a more coherent dialog with appropriate content ordering when compared with baselines.
  • Integrating text knowledge directly for modeling multiturn dialog logic paves the way for developing models in lowresource domains that lack sufficient dialog corpus.
  • The authors will adapt EGRL for low-resource domains
Summary
  • Introduction:

    One of the key goals of AI is to build a machine that can converse with humans by generating informative, coherent and sustainable open-domain conversations
  • To achieve this goal, end-to-end neural generative models have been studied [Ritter et al, 2011; Shang et al, 2015].
  • These models tend to produce generic responses
  • To address this issue, some work propose to generate responses by grounding on external knowledge [Dinan et al, 2019; Ghazvininejad et al, 2018; Zhou et al, 2018].
  • The authors make a step towards coherent and informative multi-turn open-domain conversation generation
  • Methods:

    Informativeness Info.* Dist-1/2# 0.18 0.07/0.21 0.56 0.12/0.44 0.47 0.10/0.41 0.48 0.12/0.47 0.81 0.27/0.69 0.79 0.23/0.61 0.59 0.17/0.51 0.74 0.25/0.66 Overall Quality.
  • The authors use Dist-1 and Dist-2 to measure the diversity of generated responses.
  • (8) User-interests consistency (User-Cons.) for overall quality: The metric is used to evaluate if a model can follow a new topic mentioned by a user.
  • A dialog will be rated “1” if the model follows the user’s new topic, otherwise “0”
  • Results:

    The authors invite three annotators to evaluate each dialog from each model. System identifiers are masked during evaluation.
  • As shown in Table 1, EGRL significantly outperforms all baselines in terms of all the metrics except for length-of-dialog
  • It demonstrates that EGRL can effectively foster a more coherent, informative, engaging conversation.
  • The authors' model obtains the best user-interests consistency result compared with baselines
  • It indicates that with EGRL, the model avoids one-sided conversation while focusing on dialog coherence.
  • As shown in Table 3, the model significantly outperforms the baselines in terms of dialog coherence, engagement and informativeness except for length-of-dialog
  • It is consistent with Table 1.
  • The ordering of the dialog content by the model is more appropriate, e.g., “tooth decay” → “emergency surgery” → “recover”
  • Conclusion:

    The authors present an event graph grounded RL framework (EGRL) to demonstrate how the knowledge of event chains can help plan a multi-turn open-domain dialog.
  • Experimental results on two datasets show that EGRL can generate a more coherent dialog with appropriate content ordering when compared with baselines.
  • Integrating text knowledge directly for modeling multiturn dialog logic paves the way for developing models in lowresource domains that lack sufficient dialog corpus.
  • The authors will adapt EGRL for low-resource domains
Tables
  • Table1: Results for dialogs with user simulator on Weibo corpus. * denotes human evaluation metrics and # denotes automatic metrics
  • Table2: Results for dialogs with human on Weibo corpus. * denotes human evaluation metrics and # denotes automatic metrics
  • Table3: Results for dialogs with user simulator on Twitter corpus. * denotes human evaluation metrics and # denotes automatic metrics
Download tables as Excel
Related work
  • Knowledge-grounded Conversation Generation. There are growing interests in leveraging external knowledge for generation of more informative responses [Ghazvininejad et al, 2018; Moghe et al, 2018; Zhou et al, 2018; Bao et al, 2019; Moon et al, 2019; Liu et al, 2019; Dinan et al, 2019; Xu et al, 2020]. Different from those work, we put more efforts on dialog coherence in the setting of multi-turn dialogs. In particular, we care about the ordering of selected knowledge, which is less studied in previous work.

    RL based Models for Conversation Generation. Previous work adopt RL based frameworks to learn dialog strategies merely from dialog corpora, which fall into two categories: (1) word-level methods with words as RL actions [Li et al, 2016b; Zhang et al, 2018]; (2) utterance-level methods with high-level utterance representations as RL actions, e.g., latent variables or keywords [Zhao et al, 2019; Yao et al, 2018]. In this work, we investigate how to leverage external knowledge to explicitly plot a dialog by RL based content planning.
Funding
  • This work is supported by the National Key Research and Development Project of China (No.2018AAA0101900) and the National Natural Science Foundation of China (NSFC) via grant 61976072
Reference
  • [Bao et al., 2019] Siqi Bao, Huang He, Fan Wang, Rongzhong Lian, and Hua Wu. Know more about each other: Evolving dialogue strategy via compound assessment. In Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics, pages 5382–5391, 2019.
    Google ScholarLocate open access versionFindings
  • [Bordes et al., 2013] Antoine Bordes, Nicolas Usunier, Alberto Garcia-Duran, Jason Weston, and Oksana Yakhnenko. Translating embeddings for modeling multi-relational data. In NIPS, 2013.
    Google ScholarFindings
  • [Chambers and Jurafsky, 2008] Nathanael Chambers and Dan Jurafsky. Unsupervised learning of narrative event chains. In ACL, 2008.
    Google ScholarLocate open access versionFindings
  • [Dinan et al., 2019] Emily Dinan, Stephen Roller, Kurt Shuster, Angela Fan, Michael Auli, and Jason Weston. Wizard of wikipedia: Knowledge-powered conversational agents. In ICLR, 2019.
    Google ScholarLocate open access versionFindings
  • [Ghazvininejad et al., 2018] Marjan Ghazvininejad, Chris Brockett, Ming-Wei Chang, Bill Dolan, Jianfeng Gao, Wen tau Yih, and Michel Galley. A knowledge-grounded neural conversation model. In AAAI, 2018.
    Google ScholarLocate open access versionFindings
  • [Jang et al., 2016] Eric Jang, Shixiang Gu, and Ben Poole. Categorical reparameterization with gumbel-softmax. arXiv preprint arXiv:1611.01144, 2016.
    Findings
  • [Kadlec et al., 2015] Rudolf Kadlec, Martin Schmid, and Jan Kleindienst. Improved deep learning baselines for ubuntu corpus dialogs. arXiv preprint arXiv:1510.03753, 2015.
    Findings
  • [Li et al., 2016a] Jiwei Li, Michel Galley, Chris Brockett, Jianfeng Gao, and Bill Dolan. A diversity-promoting objective function for neural conversation models. In NAACL-HLT, 2016.
    Google ScholarLocate open access versionFindings
  • [Li et al., 2016b] Jiwei Li, Will Monroe, Alan Ritter, Michel Galley, Jianfeng Gao, and Dan Jurafsky. Deep reinforcement learning for dialogue generation. In EMNLP, 2016.
    Google ScholarLocate open access versionFindings
  • [Li et al., 2018] Zhongyang Li, Xiao Ding, and Ting Liu. Constructing narrative event evolutionary graph for script event prediction. 2018.
    Google ScholarFindings
  • [Li et al., 2019] Zhongyang Li, Xiao Ding, and Ting Liu. Story ending prediction by transferable bert. In IJCAI, 2019.
    Google ScholarLocate open access versionFindings
  • [Liu et al., 2019] Zhibin Liu, Zheng-Yu Niu, Hua Wu, and Haifeng Wang. Knowledge aware conversation generation with explainable reasoning over augmented graphs. In EMNLP-IJCNLP, 2019.
    Google ScholarLocate open access versionFindings
  • [Luong et al., 2015] Minh-Thang Luong, Hieu Pham, and Christopher D Manning. Effective approaches to attentionbased neural machine translation. In EMNLP, 2015.
    Google ScholarLocate open access versionFindings
  • [Moghe et al., 2018] Nikita Moghe, Siddhartha Arora, Suman Banerjee, and Mitesh M. Khapra. Towards exploiting background knowledge for building conversation systems. In EMNLP, 2018.
    Google ScholarLocate open access versionFindings
  • [Moon et al., 2019] Seungwhan Moon, Pararth Shah, Anuj Kumar, and Rajen Subba. Opendialkg: Explainable conversational reasoning with attention-based walks over knowledge graphs. In ACL, 2019.
    Google ScholarLocate open access versionFindings
  • [Mostafazadeh et al., 2016] Nasrin
    Google ScholarFindings
  • James Allen. A corpus and cloze evaluation for deeper understanding of commonsense stories. In NAACL, 2016.
    Google ScholarLocate open access versionFindings
  • [Qin et al., 2019] Lianhui Qin, Michel Galley, Chris Brockett, Xiaodong Liu, Xiang Gao, Bill Dolan, Yejin Choi, and Jianfeng Gao. Conversing by reading: Contentful neural conversation with on-demand machine reading. In ACL, 2019.
    Google ScholarLocate open access versionFindings
  • [Ramage et al., 2009] Daniel Ramage, David Hall, Ramesh Nallapati, and Christopher D Manning. Labeled lda: A supervised topic model for credit attribution in multi-labeled corpora. In ACL, 2009.
    Google ScholarLocate open access versionFindings
  • [Ritter et al., 2011] Alan Ritter, Colin Cherry, and William B Dolan. Data-driven response generation in social media. In EMNLP, 2011.
    Google ScholarLocate open access versionFindings
  • [Shang et al., 2015] Lifeng Shang, Zhengdong Lu, and Hang Li. Neural responding machine for short-text conversation. In ACL, 2015.
    Google ScholarLocate open access versionFindings
  • [Sutton and Barto, 2018] Richard S Sutton and Andrew G Barto. Reinforcement learning: An introduction. MIT Press, 2018.
    Google ScholarFindings
  • [Tang et al., 2019] Jianheng Tang, Tiancheng Zhao, Chengyan Xiong, Xiaodan Liang, Eric P Xing, and Zhiting Hu. Target-guided open-domain conversation. In ACL, 2019.
    Google ScholarLocate open access versionFindings
  • [Vaswani et al., 2017] Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszkoreit, Llion Jones, Aidan N Gomez, Łukasz Kaiser, and Illia Polosukhin. Attention is all you need. In NIPS, 2017.
    Google ScholarLocate open access versionFindings
  • [Xu et al., 2020] Jun Xu, Haifeng Wang, Zhengyu Niu, Hua Wu, and Wanxiang Che. Knowledge graph grounded goal planning for open-domain conversation generation. In Thirty-Fourth AAAI Conference on Artificial Intelligence, 2020.
    Google ScholarLocate open access versionFindings
  • [Yao et al., 2018] Lili Yao, Ruijian Xu, Chao Li, Dongyan Zhao, and Rui Yan. Chat more if you like: Dynamic cue words planning to flow longer conversations. arXiv preprint arXiv:1811.07631, 2018.
    Findings
  • [Zhang et al., 2018] Wei-Nan Zhang, Lingzhi Li, Dongyan Cao, and Ting Liu. Exploring implicit feedback for open domain conversation generation. In Thirty-Second AAAI Conference on Artificial Intelligence, 2018.
    Google ScholarLocate open access versionFindings
  • [Zhao et al., 2019] Tiancheng Zhao, Kaige Xie, and Maxine Eskenazi. Rethinking action spaces for reinforcement learning in end-to-end dialog agents with latent variable models. In NAACL-HLT, 2019.
    Google ScholarLocate open access versionFindings
  • [Zhou et al., 2018] Hao Zhou, Tom Young, Minlie Huang, Haizhou Zhao, Jingfang Xu, and Xiaoyan Zhu. Commonsense knowledge aware conversation generation with graph attention. In IJCAI, 2018.
    Google ScholarLocate open access versionFindings
Your rating :
0

 

Tags
Comments