| CARVIEW |
May 26, 2022 | Dublin
Welcome to 2nd DialDoc Workshop
May 26, 2022 at Dublin
DialDoc Workshop focuses on building and scaling up document-grounded dialogue and conversational question answering systems for various domains.
Join our Google Group for important updates.
Read MoreThemes
Our special theme aim to support building scalable document-grounded conversational systems, especially for low-resource domains including applications in low-resource languages or unforeseen emerging domains (e.g., social well-being during pandemic).
Read MoreShared Task
We propose a shared task to invite different research perspectives towards building goal-oriented document-grounded dialogue system on open-book multi-domain documents.
Read MoreAbout
DialDoc Workshop
Welcome to 2nd DialDoc Workshop co-located with ACL 2022!
DialDoc Workshop focuses on document-grounded dialogue and conversational question answering. There is a vast amount of document content created every day by human writers to communicate with human readers for sharing knowledge, ranging from encyclopedias to social benefits. Making the document content accessible to users via conversational systems and scaling it to various domains could be a meaningful yet challenging task. There are significant individual research threads that show promise in handling heterogeous knowledge embedded in documents for building conversational systems, including (1) unstructured content, such as text passages; (2) semi-structured content, such as tables or lists; (3) multi-modal content, such as images and videos along with text descriptions, and so on. The purpose of this workshop is to invite researchers and practitioners to bring their individual perspectives on the subject of document-grounded dialogue and conversational question answering to advance the field in a joint effort.
Topics of interests include but not limited to
- document-grounded dialogue and conversational machine reading, such as CMUDoG, DREAM, ShARC and Doc2Dial;
- open domain dialogue and conversational QA, such as ORConvQA, TopiOCQA, Abg-CoQA, QReCC and MultiDoc2Dial;
- conversational search among domain documents, such as MANtIS
- parsing semi-structured document content for dialogue and conversational QA, table reading, such as HybridQA and TAT-QA;
- summarization for dialogue, query-based summarization, such as AQUAMUSE;
- knowledge-grounded dialogue generation;
- evaluations for document-grounded dialogue;
- interpretability and faithfulness in dialogue modeling;
- knowledge-grounded multimodal dialogue and question answering.
Workshop Schedule
Program
PDF version of all the papers could be found here.
May 26, 2022 - Dublin local time (GMT+1)
- 09:00 - 09:05 | Opening Remark
- 09:05 - 09:40 | Invited Talk I by Siva Reddy - How Grounded is Document-grounded Conversational AI?
- 09:40 - 10:25 | Oral Presentation I
-
Conversation- and Tree-Structure Losses for Dialogue Disentanglement
Tianda Li, Jia-Chen Gu, Zhen-Hua Ling, Quan Liu -
Construction of Hierarchical Structured Knowledge-based Recommendation Dialogue Dataset and Dialogue System
Takashi Kodama, Ribeka Tanaka, Sadao Kurohashi -
Retrieval-Free Knowledge-Grounded Dialogue Response Generation with Adapters
Yan Xu, Etsuko Ishii, Samuel Cahyawijaya, Zihan Liu, Genta Indra Winata, Andrea Madotto, Dan SU, Pascale Fung
-
Conversation- and Tree-Structure Losses for Dialogue Disentanglement
- 10:25 - 10:40 | Coffee Break
- 10:40 - 11:15 | Invited Talk II by Jeff Dalton - Knowledge-Grounded Conversation Search and Understanding: Current Progress and Future Directions
- 11:15 - 11:55 | Paper Lightning Talk
-
TRUE: Re-evaluating Factual Consistency Evaluation
Or Honovich, Roee Aharoni, Jonathan Herzig, Hagai Taitelbaum, Doron Kukliansy, Vered Cohen, Thomas Scialom, Idan Szpektor, Avinatan Hassidim, Yossi Matias -
Pseudo Ambiguous and Clarifying Questions Based on Sentence Structures Toward Clarifying Question Answering System
Yuya Nakano, Seiya Kawano, Koichiro Yoshino, Katsuhito Sudoh, Satoshi Nakamura -
Parameter-Efficient Abstractive Question Answering over Tables or Text
Vaishali Pal, Evangelos Kanoulas, Maarten de Rijke -
Conversational Search with Mixed-Initiative - Asking Good Clarification Questions backed-up by Passage Retrieval
Yosi Mass, Doron Cohen, Asaf Yehudai, David Konopnicki -
Graph-combined Coreference Resolution Methods on Conversational Machine Reading Comprehension with Pre-trained Language Model
Zhaodong Wang, Kazunori Komatani -
MSAMSum: Towards Benchmarking Multi-lingual Dialogue Summarization
Xiachong Feng, Xiaocheng Feng, Bing Qin -
UniDS: A Unified Dialogue System for Chit-Chat and Task-oriented Dialogues
Xinyan Zhao, Bin He, Yasheng Wang, Yitong Li, Fei Mi, Yajiao LIU, Xin Jiang, Qun Liu, Huanhuan Chen -
G4: Grounding-guided Goal-oriented Dialogues Generation with Multiple Documents
Shiwei Zhang, Yiyang Du, Guanzhong Liu, Zhao Yan, Yunbo Cao
-
TRUE: Re-evaluating Factual Consistency Evaluation
- 11:55 - 12:55 | Poster Session
- 12:55 - 14:00 | Lunch Break
- 14:00 - 14:35 | Invited Talk III by Zhou Yu - Will Big Models Solve Dialog Systems?
- 14:35 - 15:05 | Oral Presentation II
- Low-Resource Adaptation of Open-Domain Generative Chatbots
Greyson Gerhard-Young, Raviteja Anantha, Srinivas Chappidi, Bjorn Hoffmeister -
Task2Dial: A Novel Task and Dataset for Commonsense-enhanced Task-based Dialogue Grounded in Documents
Carl Strathearn, Dimitra Gkatzia
- Low-Resource Adaptation of Open-Domain Generative Chatbots
- 15:05 - 15:20 | Coffee Break
- 15:20 - 15:35 | Shared Task Results
- presentation by Team CPII-NLP
- 15:35 - 15:50 | Shared Task Prizes and Best Paper Awards presented by Luis Lastras, Director, IBM Research
- 15:50 - 16:25 | Invited Talk IV by Michel Galley - Interactive Document Generation
- 16:25 - 17:00 | Invited Talk V by Mari Ostendorf - Understanding Conversation Context is Central to Conversational AI
- 17:00 - 17:05 | Ending Remark
Calls
Call for Papers
We welcome submissions of original work as long or short papers, as well as non-archival papers. We also accept paper submissions via ARR. All accepted papers will be presented at the workshop.
We will have the Best Paper Award and Best Student Paper Award, which will be announced during the workshop.
Submission Instructions
Formatting Guidelines:
We accept long (eight pages plus unlimited references) and short (four pages plus unlimited references) papers, which should
conform to ARR CFP guidelines.
Non-Archival Submissions:
The accepted papers can opt to be
non-archival.
Submit your paper:
- For submitting papers directly to our workshop, use button "ACL 2022 Workshop DialDoc Submission" on the submission site.
- For submitting papers with ARR reviews, use button "ACL 2022 Workshop DialDoc Commitment Submission" on the submission site.
Review Process
All submissions will be peer-reviewed by at least two reviewers. The reviewing process will be two-way anonymized. Authors are responsible for anonymizing the submissions.
Important Dates
- For regular workshop paper submissions (regular/non-archival long/short tracks)
- Paper Due Date:
February 28, 2022 (AoE)→ March 7, 2022 (AoE) - Notification of Acceptance: March 26, 2022
- Paper Due Date:
- For paper submissions with ARR reviews
- Paper Due Date:
February 28, 2022 (AoE)→ March 24, 2022 (AoE) - Notification of Acceptance: March 26, 2022
- Paper Due Date:
- For technical paper submissions (Shared Task track)
- Paper Due Date: March 27, 2022 (AoE)
- Notification of Acceptance: April 3, 2022
- For all
- Camera-ready Paper Due Date: April 10, 2022 (AoE)
- Workshop Date: May 26, 2022
Special Theme
This workshop focuses on scaling up document-grounded dialogue systems especially for low-resource domains, e.g., the applications in low-resource languages or emerging unforseen situations such as COVID-19 pandemic. We seek submissions that tackles the challenge on different aspects, including but not limited to
- effective adaption of pre-trained models;
- data augmentation and data generation for dialogue models;
- domain adaption and meta learning for knowledge-grounded dialogue;
We also maintain a list of resources on COVID-19 datasets in different languages. Please contact us if you would like to suggest a related dataset.
Learn More
Data Competition
Shared Task
Rewards - Task on SEEN
| 1st Prize | CPII-NLP |
| 2nd Prize | zsw_dyy_lgz |
| 3rd Prize | UGent-T2K |
Rewards - Task on UNSEEN
| 1st Prize | CMU_QA |
| 2nd Prize | CPII-NLP |
| 3rd Prize | UGent-T2K |
About
The shared task centers on building open-book goal-oriented dialogue systems, where an agent could provide an answer or ask follow-up questions for clarification or verification. The main goal is to generate grounded agent responses in natural language based on the dialogue context and domain knowledge in the documents.
Please star our leaderboard if you are interested in the task! Please see more details at Call for Participants section.
Rewards
We offer the following awards to top participating teams for each task. The prizes are sponsored by IBM Research AI.
- 1st place: $1000
- 2nd place: $500
- 3rd place: $200
Resources
The training and test data is based on MultiDoc2Dial dataset. Please check out for running baselines.
Important Dates
The final date for leaderboard submission is April 3, 2022.
The due date for techincal paper submission is March 27, 2022.
Join our Google Group for important updates! If you have any question, ask in our Google Group or email us.
Task and Data
The task is to generate grounded agent response given dialogue query and domain documents. Specifically,
- Input: latest user turn, dialogue history and all domain documents.
- Output: agent response in natural language.
The training data includes the training and validation splits from MultiDoc2Dial dataset. Given the low-resource nature of the domains in MultiDoc2Dial, we also encourage the teams to explore the approaches such as data augmentation for the tasks in two ways: (1) utilizing existing public datasets; (2) utilizing automatically generated synthetic data without involving any additional human labeled data on MultiDoc2Dial.
Evaluations
The test data is derived from the MultiDoc2Dial dataset. There are two tasks based on different settings in test data,
- MultiDoc2Dial-seen-domain (MDD-SEEN): all dialogues in the test data are grounded in the documents from the same domains as the training data.
- MultiDoc2Dial-unseen-domain (MDD-UNSEEN): all dialogues in the test data are grounded in the documents from an unseen domain.
The results will be evaluated based on both automatic metrics and human evaluations. For automatic evaluation, check out the evaluation script.
Please check out the for our Shared Task challenge.
The challenge includes leaderboards for two task settings with two phases, Dev (TestDev) and Test phase,
You can find more details regarding the phases and submission once your team is registered.Shared Task Paper Submission Guidelines
We welcome techincal paper submissions based on the Shared Task. To qualify for the competition prizes, a participating team will need to complete at least one of the two task settings and make a paper submission.
The paper should be up to four pages with unlimited references. The format should conform to ARR CFP guidelines.
Review Process
All submissions will be peer-reviewed by at least two reviewers. The reviewing process will be double-blinded at the level of the reviewers. Authors are responsible for anonymizing the submissions.
Important Dates
- Leaderboard Submission Final Date: April 3, 2022
- Techincal Paper Due Date: March 27, 2022 (AoE)
- Notification of Acceptance: April 3, 2022
- Camera-ready Paper Due: April 10, 2022 (AoE)
- Workshop Dates: May 26, 2022
Talks
Invited Speakers
Jeff Dalton
University of Glasgow
Michel Galley
Microsoft Research
Mari Ostendorf
University of Washington
Siva Reddy
MILA/MCQLL/CIFAR
Zhou Yu
Columbia UniversityOrganization
Workshop Organizers
Song Feng
Amazon - AWS AI Labs
Chengguang Tang
Tencent AI Lab
Svitlana Vakulenko
Amazon
Hui Wan
IBM Research
Zeqiu (Ellen) Wu
University of Washington
Caixia Yuan
Beijing University of Posts and TelecommunicationsProgram Committee
- Amanda Buddemeyer (University of Pittsburgh)
- Asli Celikyilmaz (Meta AI Research)
- Bowen Yu (Chinese Academy of Sciences)
- Chen Henry Wu (Carnegie Mellon University)
- Chulaka Gunasekara (IBM Research AI)
- Chun Gan (JD Research)
- Cunxiang Wang (Westlake University)
- Danish Contractor (IBM Research)
- Dian Yu (Tencent)
- Diane Litman (University of Pittsburgh)
- Ehud Reiter (University of Aberdeen)
- Elizabeth Clark (University of Washington)
- Fanghua Ye (University College London)
- Tao Feng (Monash University)
- Guanyi Chen (Utrecht University)
- Hanjie Chen (University of Virginia)
- Hao Zhou (Tencent)
- Haotian Cui (Toronto University)
- Haochen Liu (Michigan State University)
- Houyu Zhang (Amazon)
- Ioannis Konstas (Heriot-Watt University)
- Jingjing Xu (Peking University)
- Jia-Chen Gu (USTC)
- Jinfeng Xiao (UIUC)
- Jian Wang (The Hong Kong Polytechnic University)
- Jingyang Li (Alibaba DAMO Academy)
- Jiwei Li (SHANNON.AI)
- Jun Xu (Baidu)
- Kai Song (ByteDance)
- Ke Shi (Tencent)
- Kun Qian (Columbia University)
- Kaixuan Zhang (Northwestern University)
- Libo Qin (MLNLP)
- Michael Johnston (Interactions)
- Meng Qu (MILA)
- Minjoon Seo (KAIST)
- Pei Ke (Tsinghua University)
- Peng Qi (Stanford University)
- Ravneet Singh (University of Pittsburgh)
- Ryuichi Takanobu (Tsinghua University)
- Rongxing Zhu (The University of Melbourne)
- Seokhwan Kim (Amazon Alexa AI)
- Shehzaad Dhuliawala (Microsoft Research Montreal)
- Srinivas Bangalore (Interactions)
- Zejiang Shen (AllenAI)
- Vaibhav Adlakha (MCGill and MILA)
- Wanjun Zhong (MSRA)
- Xi Chen (Tencent)
- Yifan Gao (The Chinese University of Hong Kong)
- Yekun Chai (Baidu)
- Yinhe Zheng (Alibaba DAMO Academy)
- Yiwei Jiang (Ghent University)
- Yajing Sun (Chinese Academy of Sciences)
- Yunqi Qiu (Chinese Academy of Sciences)
- Yosi Mass (IBM Research)
- Yutao Zhu (University of Montreal)
- Zheng Zhang (Tsinghua University)
- Zhenyu Zhang (Chinese Academy of Sciences)
- Zhenzhong Lan (Westlake University)
- Zhixing Tian (JD)
Sponsors
Gold
