Special Sessions & Workshops:

IWSDS 2023 will host two special sessions and one workshop:

Special Session: Multi-Party Conversational AI

Organizing Committee: 

Oliver Lemon, Heriot-Watt University, Edinburgh, UK

Christian Dondrup, Heriot-Watt University, Edinburgh, UK

Daniel Hernández García, Heriot-Watt University, Edinburgh, UK

Nancie Gunson, Heriot-Watt University, Edinburgh, UK

Angus Addlesee, Heriot-Watt University, Edinburgh, UK

Weronika Sieińska, Heriot-Watt University, Edinburgh, UK

The program of IWSDS 2023 will include a special session on multi-party conversational AI, where more than two agents are involved in a conversational interaction. The objectives of this session are to review work done in the past in the field of multi-party conversational AI, to showcase recent and ongoing work, and to identify paths forward. Topics of interest are (but are not limited to): designing social state representations and models for multi-party interactions, addressee identification, speaker diarization, common ground detection, transformers for multi-party dialogue, datasets and simulators for multi-party dialogue, reinforcement learning for managing multi-party dialogue, handling split utterances (user utterances split between dialogue turns and utterances of other speakers), anaphora and ellipsis resolution in multi-party dialogue, multimodal input and output in multi-party dialogue systems, monitoring conversation status (sentiment analysis, detection of (dis)agreements and misunderstandings, etc.), methods and metrics for evaluating multi-party dialogue systems.

Special Session: Dialogue Systems for Multilingual and Under-resourced Language Speakers

Organizing Committee: 

A.Seza Doğruöz, Universiteit Gent, Belgium

Sunayana Sitaram, Microsoft Research, India

Current dialogue systems target mostly monolingual and high resource languages and their speakers. However, millions of speakers around the world (e.g., India, Africa, Europe as well as indigenous and immigrant communities in the US) are multilingual and it is normal for these speakers and communities to switch within or across languages in daily lives (Doğruöz et al., 2021; Sitaram et al., 2019). In addition, most languages of the world are still under-resourced. Therefore, there is a need for dialogue systems to be more inclusive and target both the multilingual and under-resourced languages and their speakers. The aim of this special session is to bring together researchers from the SDS community and encourage research and discussion around the unique challenges (e.g., data collection, model building, sociolinguistic aspects and system evaluation) for multilingual and under-resourced languages.

Co-located Workshop: Language-Based AI Agent Interaction with Children

Organizing Committee: 

Douglas Fidaleo, Disney Research

Eda Okur, Intel Labs

James Kennedy, Disney Research

Lenitra Durham, Intel Labs

Maike Paetzel-Prüsmann, Disney Research

Naveen Kumar, Disney Research

Ramesh Manuvinakurike, Intel Labs

Saurav Sahay, Intel Labs

Sinem Aslan, Intel Labs

Workshop's website: https://aichildinteraction.github.io/

This workshop aims to bring together researchers looking into multimodal interactions between children and artificial agents to discuss research problems that center around interactivity and go beyond just processing child speech. There is interest in discussing approaches to intent classification in child speech, designing dialogue flow with characters that primarily interact with children, as well as repair strategies, active listening behavior, and other aspects of dialogue modelling. Moreover, multi-party conversations involving several children, children and their adult caregivers or several artificial characters are of particular interest to this workshop. Even though the research community around child-agent and child-robot interaction has grown over the past years, research in many of the areas stated above is still in early stages given the difficulty in collecting and annotating datasets involving child speech. Hence, this will also be a topic of interest.