Workshop Motivation and Description
Verbal communication in the form of spoken dialogues is the most natural and prevalent medium of information exchange between humans. Meetings, prevalent in enterprise environments, are a common communication channel to reach the collective agreements on decision making. Another useful application area of analysis of conversational dialogues is the customer care, where a customer usually interacts with a human or an automated agent (a chatbot) to work towards a specific task. e.g. troubleshoot a Linux workstation etc. .
There are several open problems in this domain that require an in-depth research such as understanding the nature of conversations, their underlying intent and evaluating the quality of new conversational search tools/algorithms. The workshop aims to accomplish the following two goals.
- Gather researchers from the IR and NLP communities to present their recent work on the topic and
- Provide a platform for new PhD students or early researchers to build and evaluate conversational systems on labeled data.
The first objective of the proposed SUD workshop series is to bring together researchers from diverse fields – Information Retrieval, Data Mining and Machine Learning, Natural Language Processing, Human Computer Interaction etc. to further advance the state of research in applications and impact of information retrieval to digital two-party or multi-party text or multi-modal dialogues. Given the diverse audience, participants and program of WSDM, it would be an ideal venue to advance research on such topics.
The second objective is to invite PhD students and new researchers to conduct research on an open dataset and discuss their findings with experts at the workshop. We specifically encourage research on a recently released dataset containing conversations extracted from movies, labeled with entities of interest and wikipedia passages relevant to the corresponding entities that add more context. The dataset, released as part of RCD Track at FIRE 2020, aims to explore the utility of IR systems to fetch more information about entities discussed in these dialogues.
Research track call for papersThe proposed workshop will consist of two tracks a) Research track and b) Data challenge track. Research paper track invites submissions from themes and topics aligned with research in Conversational Dialogues. The following topics associated with Intelligent Assistants are particularly of interest for SUD 2021 workshop:
- Voice search
- conversational search
- dialog systems
- Personal assistants
- dialogue models
- conversational interaction
Data Challenge Track
The data challenge track is a continuation of our effort to investigate the research question of contextualization of conversational dialogues. A shared task was recently organized - Retrieval from Conversational Dialogues (RCD) at FIRE 2020.Data challenge track seeks submissions that investigate both understanding of sample dialogues and support such conversations with more contextual information. As part of the challenge, we are releasing conversational data for two tasks. The first task is to detect entities in these conversations. The second task is to retrieve informative content about these entities to facilitate user understanding. Although prior work has looked into contextualizing short documents, such as tweets, the proposed data challenge track requires contextualizing aspects of a conversation.
Papers submitted to this track must report at least one experiment with the dataset released as a part of the RCD track. Authors are encouraged to either experiment and report results for the two RCD tasks - that of detecting entities in the text spans from movie dialogues or of retrieving relevant information given a dialog context, or reporting experiments on novel tasks on this dataset.
Papers submitted for both the regular research and the data challenge tracks should be 4-6 pages of length of content (plus 1 additional page for references). Papers must be submitted in PDF according to the new ACM format published in ACM guidelines, selecting the generic sigconf template. The review process is single-blind and therefore submissions should contain author information. Paper should be uploaded via Easychair via this submission link. Accepted papers will be included in the SUD 2021 proceedings (companion volume of WSDM 2021) and at least one author of each accepted contribution must attend the workshop. After acceptance, no additional authors can be added.
- Submission deadline - 26th January, 2021
- Notifications- 23rd February, 2021
- Camera-Ready - 2nd March, 2021
- Workshop day- 12th March, 2021
Program (Tentative Plan)
The following paper has been accepted in the workshop.
- Explaining outcomes of multi-party dialogues using causal learning: Priyanka Sinha, Pabitra Mitra, Antonio Anastasio Bruto da Costa, and Nikolaos Kekatos.
Due to a small number (specifically, 1) of papers accepted, there's no separate proceedings for the workshop. An arxiv version of the accepted paper can be found here.
Following is the program for the workshop.
- 11.00 GMT: Introduction -- About 5-10 mins
- 11.10 GMT: Keynote-1: Jeff Dalton, University of Glasgow --- 1.30 hours.
- 12.40 GMT: Comfort-break 1 --- 10 mins.
- 12.50 GMT: Industry Keynote-2: Surbhi Rathore, CEO, Symbl.AI -- 50 mins.
- 13.40 GMT: Paper presentation -- Priyanka Sinha, IIT-KGP, India --- 20 mins.
- 14.00 GMT: Invited Talk: Emine Yilmiaz, University College London, Amazon Scholar --- 20 mins
- 14.20 GMT: Invited Talk: Pabitra Mitra, Indian Institute of Technology, Kharagpur, India --- 20 mins
- 14.40 GMT: Panel discussion --- 15 mins.
- 14.55 GMT: Closing remarks -- 5 mins.
Keynote TalkJeff Dalton (University of Glasgow) Dr. Jeff Dalton is an Assistant Professor (Lecturer) in the School of Computing Science at the University of Glasgow where he leads the Glasgow Representation and Information Learning Lab (GRILL). His research focuses on text understanding and conversational information seeking. He completed his PhD at the University of Massachusetts Amherst in the Center for Intelligent Information Retrieval. Later in Google Research, he worked on Information Extraction as part of the Knowledge Discovery Team (Knowledge Vault) and language understanding in the Assistant Response Ranking team. He is the lead organizer for the TREC Conversational Assistance Track (CAsT) (http://treccast.ai) and previously helped organize the Complex Answer Retrieval track. He is the recipient of a prestigious UKRI Turing AI Acceleration Fellowship and received research awards from Google, Amazon, and Bloomberg. He holds multiple patents in retrieval and question answering systems. Title: Conversation Search: Current Challenges and Future Directions Abstract: This talk examines some of the fundamental challenges in Conversational Information Seeking (CIS). It describes lessons learned developing recent conversational search benchmarks (TREC CAsT and ConvAI3) and open research challenges. It also presents work from the Glasgow Representation and Information Learning Lab (GRILL) on neural models for conversational rewriting, entity-based ranking, and the use of feedback and initiative in CIS.
Invited TalkSurbhi Rathore (Symbl.ai) Surbhi is the CEO and Co-founder of Symbl.ai. Symbl is bringing to life her vision for a programmable platform that empowers developers and businesses to build unique conversational experiences without the hassle of upfront training or building their in-house data science expertise. She co-founded Symbl almost 2 years ago and backed by Amazon and a Techstars alum. Symbl raised an early-stage venture round of $6.5M and deployed capital to grow a team of tech enthusiasts to 30 people primarily distributed between India and Seattle. She comes with experience from technical and customer-obsessed roles in both startups and enterprises such as Nevis Networks and Amdocs. Before co-founding Symbl, she worked in the Conversational AI space with a focus on delivering value to Telecommunication users. She is an advocate for Women in AI with a personal mission to inspire more women to work in Data Science. In her free time, she loves to travel and spend time with her remote and distributed family. Title: Conversation intelligence for the growing digital ecosystem Abstract: Now more than ever, when the whole world is communicating on digital channels, the dangers of disconnected data silos and lost knowledge are more than ever. In this session, we will talk about applying AI/ML to human conversations at scale so that builders can strategize product strategy early in the life cycle of their business and future-proof growth. The session also entails several dimensions of conversation intelligence that can amplify customer experiences, superpower productivity and drive growth opportunities for businesses for several mission critical use cases.
Please reach out to the organizers for any questions. For further questions and clarifications, please mail to firstname.lastname@example.org.