對話系統，如客服系統、聊天機器人、智慧音箱等，是人工智能領域中受到廣泛期待的應用。任務型導向對話系統的挑戰包括訓練語料的收集、標記，以及模型的架構及訓練。在語料蒐集上，過去常使用 Wizard-of-Oz(WOZ) 方法，透過人與人之間的互動對話來進行語料的收集和標記。但由於使用 WOZ 方法時，標記人員須同時標記對話狀態及槽值，也必須留意先前對話歷史中是否提及某些實體，這會大大影響整體資料集的品質，且也無法迅速
建立對話語料。為了能加速資料的蒐集，部分研究人員也採用綱要引導對話
(Schema-Guided Dialogue，SGD) 來自動蒐集語料，透過定義清楚的綱要使對話代理來自動並蒐集對話。但由於綱要引導對話在自動對話蒐集前須做詳細的定義，且目前尚未有任何原始碼的公佈，使得研究人員現今仍普遍使用 WOZ來模擬人機互動的對話情境。
本文參照 Schema-Guided Dialogue(SGD) 的語料收集方法，透過使用者、助理兩個對話代理程式建立對話模擬器，建立一個能夠處理電子郵件、管理行事曆、以及傳遞訊息等三種服務的對話語料集。SGD 的做法為依據任務綱要(Schema) 來讓使用者代理能依據綱要提出需求，助理代理再查看綱要來向對方詢問相關資訊，或呼叫 API 來進行查詢。模擬器經過對話模擬後會產生由一連串對話行為組成的對話大綱，我們將對話大綱通過預先定義好的對話模板轉為樣版式句子，最後標記人員僅須結合多個單一句子轉成自然的對話，即可獲得訊息領域相關的對話語料 messageSGD。
基於本研究所建立的 messageSGD，我們以 T5 為基底套用至 TOD 對話
系統中的四個任務上，並在訓練時應用 Instruction Prompt 方法，在各個任務的輸入中添加任務敘述及特殊標記。經過如此的設計，使系統中各個模型能更理解四項任務的輸入輸出，在對話理解任務中整體準確度由 76.03 到 83.36，提升 7.33 個百分點，對話生成中的 BLEU-Score 也從 25.89 提升至 32.43。在對話狀態追蹤和對話決策任務上，我們也透過調整對話歷史的多寡，使狀態追蹤的準確度從 41.65 提升至 51.69，上升了 10.04 個百分點。

Dialogue systems, such as customer service systems, chatbots, and smart speakers, are widely anticipated applications in the field of artificial intelligence. The challenges of task-oriented dialogue systems include the collection and annotation of training data and the architecture and training of models. In terms of data collection, the Wizard-of-Oz (WOZ) method has often been used in the past, where conversational data is collected and annotated through interactive dialogues between humans. However, when using the WOZ method, annotators must pay attention to whether certain entities have been mentioned in the previous dialogue history. This significantly affects the overall quality of the dataset and hinders the rapid establishment of dialogue corpora.
In this paper, we refer to the data collection method of Schema-Guided Dialogue (SGD), which involves building dialogue simulators using two dialogue agents. These simulators are designed to handle three types of services: email
processing, calendar management, and message delivery. The SGD approach
generates dialogue outlines based on task schema and then converts these outlines into templated sentences using pre-defined dialogue templates. As a result,
annotators only need to combine multiple single sentences to create natural dialogues.
Based on the messageSGD developed in this study, we apply the T5 model
to four tasks in the TOD dialogue system. During training, we employ the
Instruction Prompt method by adding task descriptions and special markers to
the inputs of each task. Through this design, the models in the system gain
a better understanding of the input and output for the four tasks, leading to
improved performance.

[1] Stephanie Seneff and Joseph Polifroni. Dialogue management in the mercury flight reservation system. In ANLP-NAACL 2000 Workshop: Conversational Systems, 2000.
[2] Antoine Raux, Brian Langner, Dan Bohus, Alan W Black, and Maxine Eskenazi. Let’s go public! taking a spoken dialog system to the real world. In in Proc. of Interspeech 2005. Citeseer, 2005.
[3] Hongshen Chen, Xiaorui Liu, Dawei Yin, and Jiliang Tang. A survey on dialogue systems: Recent advances and new frontiers. Acm Sigkdd Explorations Newsletter, 19(2):25–35, 2017.
[4] Long Ouyang, Jeffrey Wu, Xu Jiang, Diogo Almeida, Carroll Wainwright, Pamela Mishkin, Chong Zhang, Sandhini Agarwal, Katarina Slama, Alex Gray, John Schulman, Jacob Hilton, Fraser Kelton, Luke Miller, Maddie Simens, Amanda Askell, Peter Welinder, Paul Christiano, Jan Leike, and Ryan Lowe. Training language models to follow instructions with human feedback. In Alice H. Oh, Alekh Agarwal, Danielle Belgrave, and Kyunghyun Cho, editors, Advances in Neural Information Processing Systems, 2022.
[5] Charles T Hemphill, John J Godfrey, and George R Doddington. The atis spoken language systems pilot corpus. In Speech and Natural Language: Proceedings of a Workshop Held at Hidden Valley, Pennsylvania, June 24- 27, 1990, 1990.
[6] Pararth Shah, Dilek Hakkani-Tür, Gokhan Tür, Abhinav Rastogi, Ankur Bapna, Neha Nayak, and Larry Heck. Building a conversational agent overnight with dialogue self-play. arXiv preprint arXiv:1801.04871, 2018.
[7] Alan Ritter, Colin Cherry, and Bill Dolan. Unsupervised modeling of Twitter conversations. In Human Language Technologies: The 2010 Annual Conference of the North American Chapter of the Association for Computational Linguistics, pages 172–180, Los Angeles, California, June 2010. Association for Computational Linguistics. 29
[8] Ryan Lowe, Nissan Pow, Iulian Serban, and Joelle Pineau. The Ubuntu dialogue corpus: A large dataset for research in unstructured multi-turn dialogue systems. In Proceedings of the 16th Annual Meeting of the Special Interest Group on Discourse and Dialogue, pages 285–294, Prague, Czech Republic, September 2015. Association for Computational Linguistics.
[9] Colin Raffel, Noam Shazeer, Adam Roberts, Katherine Lee, Sharan Narang, Michael Matena, Yanqi Zhou, Wei Li, and Peter J. Liu. Exploring the limits of transfer learning with a unified text-to-text transformer. J. Mach. Learn. Res., 21(1), jan 2020.
[10] J. F. Kelley. An iterative design methodology for user-friendly natural language office information applications. ACM Trans. Inf. Syst., 2(1):26–41, jan 1984.
[11] Tsung-Hsien Wen, David Vandyke, Nikola Mrkšić, Milica Gašić, Lina M. Rojas-Barahona, Pei-Hao Su, Stefan Ultes, and Steve Young. A networkbased end-to-end trainable task-oriented dialogue system. In Proceedings of the 15th Conference of the European Chapter of the Association for Computational Linguistics: Volume 1, Long Papers, pages 438–449, Valencia, Spain, April 2017. Association for Computational Linguistics.
[12] Osman Ramadan, Paweł Budzianowski, and Milica Gašić. Large-scale multidomain belief tracking with knowledge sharing. In Proceedings of the 56th Annual Meeting of the Association for Computational Linguistics (Volume 2: Short Papers), pages 432–437, Melbourne, Australia, July 2018. Association for Computational Linguistics.
[13] Paweł Budzianowski, Tsung-Hsien Wen, Bo-Hsiang Tseng, Iñigo Casanueva, Stefan Ultes, Osman Ramadan, and Milica Gašić. MultiWOZ - a large-scale multi-domain Wizard-of-Oz dataset for task-oriented dialogue modelling. In Proceedings of the 2018 Conference on Empirical Methods in Natural Language Processing, pages 5016–5026, Brussels, Belgium, October-November 2018. Association for Computational Linguistics.
[14] Qi Zhu, Kaili Huang, Zheng Zhang, Xiaoyan Zhu, and Minlie Huang. CrossWOZ: A large-scale Chinese cross-domain task-oriented dialogue dataset. Transactions of the Association for Computational Linguistics, 8:281–295, 2020.
[15] Abhinav Rastogi, Xiaoxue Zang, Srinivas Sunkara, Raghav Gupta, and Pranav Khaitan. Towards scalable multi-domain conversational agents: The 30 schema-guided dialogue dataset. Proceedings of the AAAI Conference on Artificial Intelligence, 34(05):8689–8696, Apr. 2020.
[16] Ssu Chiu, Maolin Li, Yen-Ting Lin, and Yun-Nung Chen. SalesBot: Transitioning from chit-chat to task-oriented dialogues. In Proceedings of the 60th Annual Meeting of the Association for Computational Linguistics (ACL), 2022.
[17] Stephen Roller, Emily Dinan, Naman Goyal, Da Ju, Mary Williamson, Yinhan Liu, Jing Xu, Myle Ott, Eric Michael Smith, Y-Lan Boureau, and Jason Weston. Recipes for building an open-domain chatbot. In Proceedings of the 16th Conference of the European Chapter of the Association for Computational Linguistics: Main Volume, pages 300–325, Online, April 2021. Association for Computational Linguistics.
[18] 李聿鎧和張嘉惠葉丞鴻. 多領域任務導向用戶語音助理對話收集系統. TAAI, 2022.
[19] Colin Raffel, Noam Shazeer, Adam Roberts, Katherine Lee, Sharan Narang, Michael Matena, Yanqi Zhou, Wei Li, and Peter J. Liu. Exploring the limits of transfer learning with a unified text-to-text transformer. Journal of Machine Learning Research, 21(140):1–67, 2020.
[20] Qi Zhu, Christian Geishauser, Hsien-chin Lin, Carel van Niekerk, Baolin Peng, Zheng Zhang, Michael Heck, Nurul Lubis, Dazhen Wan, Xiaochen Zhu, et al. Convlab-3: A flexible dialogue system toolkit based on a unified data format. arXiv preprint arXiv:2211.17148, 2022.
[21] Linting Xue, Noah Constant, Adam Roberts, Mihir Kale, Rami Al-Rfou, Aditya Siddhant, Aditya Barua, and Colin Raffel. mt5: A massively multilingual pre-trained text-to-text transformer. arXiv preprint arXiv:2010.11934, 2020.
[22] Prakhar Gupta, Cathy Jiao, Yi-Ting Yeh, Shikib Mehri, Maxine Eskenazi, and Jeffrey Bigham. InstructDial: Improving zero and few-shot generalization in dialogue through instruction tuning. In Proceedings of the 2022 Conference on Empirical Methods in Natural Language Processing, pages 505–525, Abu Dhabi, United Arab Emirates, December 2022. Association for Computational Linguistics.
[23] ClueAI. Promptclue：全中文任務零樣本學習模型, 2022.
[24] Kishore Papineni, Salim Roukos, Todd Ward, and Wei-Jing Zhu. Bleu: a method for automatic evaluation of machine translation. In Proceedings of 31 the 40th annual meeting of the Association for Computational Linguistics, pages 311–318, 2002.
[25] Tianyi Zhang*, Varsha Kishore*, Felix Wu*, Kilian Q. Weinberger, and Yoav Artzi. Bertscore: Evaluating text generation with bert. In International Conference on Learning Representations, 2020.