Elamur Research⌕
☰
home / papers / 2020 / Towards Conversational Recommendation over Multi-Type Dialogs

Towards Conversational Recommendation over Multi-Type Dialogs

Paper · 2020 · Zeming Liu, Haifeng Wang, Zheng-Yu Niu et al.

Summary

Earlier conversational recommender research mostly handles one kind of dialog and assumes the user knows the conversation's purpose from the start. This paper defines a new task: a bot that, without revealing its plan, gradually moves a conversation from question answering, chitchat or a task dialog toward recommending an item, while adapting to the user's interests and reactions. To support this task, the authors built DuRecDial at Baidu. It is a human-to-human Chinese corpus grounded in a knowledge graph, with explicit seeker profiles and several consecutive sessions for each seeker–recommender pair. They also propose MGCG, a multi-goal driven generation framework with two parts. A goal-planning module selects a chain of goals, each a dialog type plus a topic, that ends in a recommendation target. A goal-guided module then produces the responses. Both a retrieval variant and a generation variant clearly beat a seq2seq baseline in automatic and human evaluation, and supplying goals and knowledge improves results. The dataset gives a demanding benchmark for proactive, personalized, mixed-type dialog systems.

Key points

  • DuRecDial has about 10.2k dialogs and 156k utterances. It covers recommendation, chitchat, QA and task dialogs across domains such as movies, music, movie stars, food, restaurants, news and weather, and includes user profiles. Data and code are on GitHub (PaddlePaddle).
  • MGCG comes in two variants, MGCG_R (retrieval-based) and MGCG_G (generation-based), and is compared against S2S. MGCG_R scores better on Hits@k and DIST-2, while MGCG_G is stronger on knowledge F1.
  • Topic prediction is difficult because there are roughly 1000 candidate topics.
  • Human evaluation collected 100 dialogs per model. Overall completed goals rose from 109 (S2S) to 158 (MGCG_R) and 174 (MGCG_G), and knowledge usage rose from 14 to 46 and 57.
  • Goal completion by dialog type: chitchat goals are the easiest to complete, while QA and recommendation goals are harder. Knowledge selection appears closely tied to successful goal completion.

Our summary of the paper, in our own words — not its text. The work belongs to its authors; read it at the links above.

 Institutions

🎓Harbin Institute of TechnologyHarbin · China76 works · 27,918 citations🏢BaiduBeijing · China40 works · 14,627 citations

📚 Shelves 5The field of science this work belongs to (OpenAlex fields).Computer Science · 9,086 worksArtificial Intelligence · 8,889 worksArts and Humanities · 9,296 worksLanguage and Linguistics · 9,206 worksInformation Systems · 523 works

🧭 Topics 1Research subjects and languages the work is about.Dialogue systems · 401 works

Cite

Cite as Zeming Liu, Haifeng Wang, Zheng-Yu Niu, Hua Wu, Wanxiang Che and Ting Liu. 2020. Towards Conversational Recommendation over Multi-Type Dialogs. Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics.

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.18653/v1/2020.acl-main.98

@inproceedings{liu2020conversational,
  title = {{Towards Conversational Recommendation over Multi-Type Dialogs}},
  author = {Zeming Liu, Haifeng Wang, Zheng-Yu Niu, Hua Wu, Wanxiang Che and Ting Liu},
  booktitle = {Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics},
  year = {2020},
  doi = {10.18653/v1/2020.acl-main.98},
  url = {https://doi.org/10.18653/v1/2020.acl-main.98}
}

References

43 references, 12 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Paweł Budzianowski, Tsung-Hsien Wen, Bo-Hsiang Tseng, Iñigo Casanueva, Stefan Ultes, Osman Ramadan, and Milica Gašić. 2018. MultiWOZ - a large-scale multi-domain wizard-of-Oz dataset for task-oriented dialogue modelling. In EMNLP. find
  2. Qibin Chen, Junyang Lin, Yichang Zhang, Ming Ding, Yukuo Cen, Hongxia Yang, and Jie Tang. 2019. Towards knowledge-based recommender dialog system. In ACL. find
  3. Konstantina Christakopoulou, Alex Beutel, Rui Li, Sagar Jain, and Ed H. Chi. 2018. Q and r: A two-stage approach toward interactive recommendation. In KDD. find
  4. Konstantina Christakopoulou, Katja Hofmann, and Filip Radlinski. 2016. Towards conversational recommender systems. In KDD. find
  5. Junyoung Chung, Caglar Gulcehre, KyungHyun Cho, and Yoshua Bengio. 2014. Empirical evaluation of gated recurrent neural networks on sequence modeling. In arXiv preprint arXiv:1412.3555. find
  6. Jacob Devlin, Ming-Wei Chang, Kenton Lee, and Kristina Toutanova. 2018. Bert: Pre-training of deep bidirectional transformers for language understanding. In arXiv preprint arXiv:1810.04805.
  7. Emily Dinan, Stephen Roller, Kurt Shuster, Angela Fan, Michael Auli, and Jason Weston. 2019. Wizard of wikipedia: knowledge-powered conversational agents. In Proceedings of ICLR. find
  8. Jiwei Li, Michel Galley, Chris Brockett, Jianfeng Gao, and Bill Dolan. 2016. A diversity-promoting objective function for neural conversation models. In NAACL-HLT, pages 110–119.
  9. Raymond Li, Samira Ebrahimi Kahou, Hannes Schulz, Vincent Michalski, Laurent Charlin, and Chris Pal. 2018. Towards deep conversational recommendations. In NIPS. find
  10. Nikita Moghe, Siddhartha Arora, Suman Banerjee, and Mitesh M. Khapra. 2018. Towards exploiting background knowledge for building conversation systems. In EMNLP.
  11. Seungwhan Moon, Pararth Shah, Anuj Kumar, and Rajen Subba. 2019. Opendialkg: Explainable conversational reasoning with attention-based walks over knowledge graphs. In ACL.
  12. Kishore Papineni, Salim Roukos, Todd Ward, and Wei-Jing Zhu. 2002. Bleu: a method for automatic evaluation of machine translation. In ACL, pages 311– 318.
  13. Ashwin Ram, Rohit Prasad, Chandra Khatri, Anu Venkatesh, Raefer Gabriel, Qing Liu, Jeff Nunn, Behnam Hedayatnia, Ming Cheng, Ashish Nagar, Eric King, Kate Bland, Amanda Wartick, Yi Pan, Han Song, Sk Jayadevan, Gene Hwang, and Art Pet-tigrue. 2018. Conversational AI: the science behind the alexa priz… find
  14. Kevin Reschke, Adam Vogel, and Daniel Jurafsky. 2013. Generating recommendation dialogs by extracting information from user reviews. In ACL. find
  15. Jesse Dodge, Andreea Gane, Xiang Zhang, Antoine Bordes, Sumit Chopra, Alexander H. Miller, and Arthur Szlam andJason Weston. 2016. Evaluating prerequisite qualities for learning end-to-end dialog systems. In ICLR. find
  16. Yueming Sun and Yi Zhang. 2018. Conversational recommender system. In SIGIR. find
  17. Marjan Ghazvininejad, Chris Brockett, Ming-Wei Chang, Bill Dolan, Jianfeng Gao, Wen tau Yih, and Michel Galley. 2018. A knowledge-grounded neural conversation model. In AAAI. find
  18. Jianheng Tang, Tiancheng Zhao, Chengyan Xiong, Xiaodan Liang, Eric P Xing, and Zhiting Hu. 2019a. Target-guided open-domain conversation. In ACL.
  19. Dongyeop Kang, Anusha Balakrishnan, Pararth Shah, Paul Crook, Y-Lan Boureau, and Jason Weston. 2019. Recommendation as a communication game: Self-supervised bot-play for goal-oriented dialogue. In EMNLP. find
  20. Yoon Kim. 2014. Convolutional neural networks for sentence classification. In Proceedings of the 2014 Conference on Empirical Methods in Natural Language Processing (EMNLP), Doha, Qatar. Association for Computational Linguistics.
  21. Sunhwan Lee, Robert Moore, Guang-Jie Ren, Raphael Arar, and Shun Jiang. 2018. Making personalized recommendation through conversation: Architecture design and recommendation methods. In AAAI Workshops. find
  22. Ilya Sutskever, Oriol Vinyals, and Quoc V Le. 2014. Sequence to sequence learning with neural networks. In NIPS. find
  23. Jianheng Tang, Tiancheng Zhao, Chenyan Xiong, Xiaodan Liang, Eric P. Xing, and Zhiting Hu. 2019b. Target-guided open-domain conversation. In Proceedings of ACL.
  24. Zhuoran Wang, Hongliang Chen, Guanchun Wang, Hao Tian, Hua Wu, and Haifeng Wang. 2014. Policy learning for domain selection in an extensible multi-domain spoken dialogue system. In EMNLP. find
  25. Pontus Warnestal. 2005. Modeling a dialogue strategy for personalized movie recommendations. In The Beyond Personalization Workshop. find
  26. Wenquan Wu, Zhen Guo, Xiangyang Zhou, Hua Wu, Xiyuan Zhang, Rongzhong Lian, and Haifeng Wang. 2019. Proactive human-machine conversation with explicit conversation goal. In ACL.
  27. Gender: We randomely select ”male” or ”female” as the seeker’s gender. find
  28. Jun Xu, Haifeng Wang, Zheng-Yu Niu, Hua Wu, and Wanxiang Che. 2020. Knowledge graph grounded goal planning for open-domain conversation generation. In AAAI. find
  29. Age range: We randomly choose one from the 5 age ranges. find
  30. Lili Yao, Yaoyuan Zhang, Yansong Feng, Dongyan Zhao, and Rui Yan. 2017. Towards implicit content-introducing for generative short-text conversation systems. In EMNLP. find
  31. Residential city: We randomly choose one from 55 cities in China as the seeker’s residential city. find
  32. Zhou Yu, Alexander I. Rudnicky, and Alan W. Black. 2017. Learning conversational systems that interleave task and non-task content. In IJCAI. find
  33. Occupation status: We randomly choose one from ”student”, ”worker” and ”retirement” based on above age range. find
  34. Saizheng Zhang, Emily Dinan, Jack Urbanek, Arthur Szlam, Douwe Kiela, and Jason Weston. 2018a. Personalizing dialogue agents: I have a dog, do you have pets too? In ACL.
  35. Domain preference: We randomly select one or two domains as ones that the seeker likes (e.g., movie, food), and one domain as the one that the seeker dislikes (e.g., news). It will affect the setting of task templates for this seeker. find
  36. Yongfeng Zhang, Xu Chen, Qingyao Ai, Liu Yang, and W. Bruce Croft. 2018b. Towards conversational search and recommendation: System ask, user respond. In CIKM. find
  37. Tiancheng Zhao, Ran Zhao, and Maxine Eskenazi. 2017. Learning discourse-level diversity for neural dialog models using conditional variational autoencoders. In ACL.
  38. Seed entity preference: We randomly select one or two entities from KG entities of the domains preferred by the seeker as his/her preference at entity level. It will affect the setting of task templates for this seeker. find
  39. Hao Zhou, Chujie Zheng, Kaili Huang, Minlie Huang, and Xiaoyan Zhu. 2020. Kdconv: A chinese multi-domain dialogue dataset towards multi-turn knowledge-driven conversation. In Proceedings of ACL. find
  40. Rejected entity list and accepted entity list: Both of them are empty at the beginning, and they will be updated as the conversation progresses; the two lists will affect the recommendation results of subsequent conversations to some extent. find
  41. Kangyan Zhou, Shrimai Prabhumoye, and Alan W Black. 2018a. A dataset for document grounded conversations. In Proceedings of EMNLP.
  42. Li Zhou, Jianfeng Gao, Di Li, and Heung-Yeung Shum. 2018b. The design and implementation of xiaoice, an empathetic social chatbot. In CoRR, abs/1812.08989. find
  43. Collection of Knowledge graph (KG) The domain of knowledge graph include stars, movies, music, news, food, POI(Point of Interest), weather. find

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/liu-2020-conversational-recommendation-over-multi-type/