DADA 2016 - 1st International Workshop on Domain Adaptation for Dialog Agents (DADA)

1st International Workshop on Domain Adaptation for Dialog Agents (DADA)
Co-located with European Conference on Machine Learning and Principles and Practice of Knowledge Discovery (ECML-PKDD)
Riva del Garda, Italy,
Conference date: September 19?23, 2016 (workshop date to be published)
Website: https://sites.google.com/site/ecmldaworkshop/home
Research in dialog systems has focused on the semantics and pragmatics of dialog, hand-crafted designs, computational linguistics, evaluation of dialog systems, and standardization across research community. Dialog systems across different applications encounter huge variability in the nature of task/application (e.g. from personal assistants to customer helpdesks) and users' behaviour (e.g. cooperativeness, novice). Statistical machine learning based dialog systems require large amount of labelled data to train models. Consequently, expanding the scope of a dialog system from one domain to another require significant amount of human labelling and model building effort. This has been one of the major bottlenecks in expansion of dialog systems to different domain and/or tasks.
The focus of the proposed workshop is to provide an engaging venue to researchers for proposing novel solutions involving applications of transfer learning and domain adaptation for next generation dialog systems, discussing technical challenges, and new possibilities in the field. This workshop aims to bring together researchers engrossed in developing dialog systems and technologies for other related tasks that can consume data from multiple domains and/or tasks to efficiently build a system in the domain of interest without much effort. We hope that it will provoke new directions of research in dialog systems to enable better analysis of intent from interactions and provide relevant responses, and augmenting our lives with much like a human agent/ personal assistant.
We encourage submissions on novel domain adaptation and transfer learning (DA/TL) techniques that enable efficient learning of dialog systems across different domains and applications. Topics of interest include but not limited to:
1. DA/TL for natural language understanding
2. DA/TL for natural language generation
3. Adapting dialog manager from one domain to another
4. DA/TL for conversation analysis
5. Intent detection & entity resolution in dialogs across domains
6. Efficient summarization of dialogs across domains
7. DA/TL for multilingual dialogs
8. Information extraction (topic, opinion, event) from dialogs across domains
9. DA/TL for human-computer interactions
10. Case studies and real-life experimentation on adapting dialog systems
11. Learning from multiple domains
Submitted papers can be of the following types:
1. Research Papers: These papers should report original research contributions and should be at most 8
pages.
2. Application and Case Studies: These papers will present novel applications of dialog systems, real-life
experiences of developing/deploying dialog systems in different applications and/or challenges/issues faced. They
should be at most 6 pages.
3. Position Papers: The extended abstracts will present novel research directions or identify challenging
problems. They should be at most 6 pages.
Submission Guidelines:
All submissions must be prepared in the camera-ready format. Please follow the submission guidelines for the ECML-
PKDD 2016 conference (http://www.ecmlpkdd2016.org/submission.html)
Important Dates:
Paper submission due: Monday, July 8, 2016
Notification of acceptance: Monday, July 25, 2016
Camera-ready copy due: : Monday, August 8, 2016
Workshop Organizers:
Shourya Roy, Xerox Research Centre India
Himanshu Sharad Bhatt, Xerox Research Centre India
Sandipan Dandapat, Xerox Research Centre India
Björn Gambäck, Norwegian University of Science and Technology in Trondheim, Norway, University of Missouri-Kansas City