Sumit has a keen interest in personalized natural language generation to make human-machine conversations increasingly interactive and immersive. He has a background in Computer Science with further specialization into Artificial Intelligence. After having obtained an MTech degree (IIIT-Allahabad, India, 2012), he forayed into the corporate world as a Software Engineer for seven years (2012-2019) before heading back to academia starting with an MSc (AI) degree (KU Leuven, Belgium, 2020).

Expertise

  • Psychology

    • Humans
    • Conversation
    • Dialogue
  • Computer Science

    • Conversational Agent
    • Roles
    • Explainable Artificial Intelligence
  • Social Sciences

    • Influence
    • Research

Organisations

Embodied conversational agents (ECAs) have gained significant research interest. Their usage as well has surged with more immersive and interactive interfaces emerging. The project focuses to drive research into the generation of personalized explanations using natural language by ECAs. The primary mechanism to be explored in this regard is alignment, i.e., to model the virtual agent such that it reflects the syntactic structure, words of choice, and non-verbal behaviour of another participant(s) in a conversation. The research output is proposed to be evaluated against a real-life use case by building on top of the dialogue and argumentation framework of an existing project named Council of Coaches. A secondary domain (use case) is also expected to be identified at a later stage.

Following are the major expected milestones in this research:
1. Design of alignment based, personalized explanation strategies for embodied
conversational agents, based on studies of alignment in human communication.
2. Implementation of these strategies in one or more embodied conversational agents
functioning as virtual health coaches.
3. Evaluation of the effects of personalized explanation strategies on
comprehensibility and persuasiveness, based on human-agent interaction
experiments.
4. Small scale implementation and evaluation of the above in the context of a
secondary use case to be developed in the inter-sectoral secondment in
accordance with the stakeholder.

Publications

2024
Examining Lexical Alignment in Human-Agent Conversations with GPT-3.5 and GPT-4 ModelsIn Chatbot Research and Design: 7th International Workshop, CONVERSATIONS 2023 Oslo, Norway, November 22–23, 2023 Revised Selected Papers (pp. 94-114). Springer. Wang, B., Theune, M. & Srivastava, S.https://doi.org/10.1007/978-3-031-54975-5_6
2023

Research profiles

Address

Visiting address

University of Twente

Citadel (building no. 09), room H238
Hallenweg 15
7522 NH Enschede
Netherlands

Navigate to location

Organisations

Scan the QR code or
Download vCard