Research Internship - Turn-taking modeling for Spoken Dialogue Systems F/M

The internship will take place at Naver Labs Europe (France), within the Interactive Systems group, and will be co-supervised with the Speech team (Naver Cloud).

In the Interactive Systems group at Naver Labs Europe, we develop AI capabilities that enable robots to interact safely with humans, other robots, and systems. For a robot to be truly useful, it must represent its knowledge of the world, share what it learns, and interact with other agents, particularly humans. Our research integrates expertise in human-robot interaction, natural language processing, speech, information retrieval, data management, and low-code/no-code programming to create AI components that empower next-generation robots to perform complex real-world tasks.

We are looking for a PhD student in Computer Science, Machine Learning or Deep Learning, or an outstanding Master's student, for a 5-6 month research internship starting in the second semester of 2026. The internship topic is turn-taking modeling for spoken dialogue systems, and it will be hosted at Naver Labs Europe.

Turn-taking is a fundamental mechanism in human conversation that governs when speakers alternate between listening and speaking. In spoken dialogue systems, poor turn-taking often results in unnatural interruptions, long silences, or premature responses, which significantly degrade user experience. Creating fluid, human-like conversational flow remains a major challenge for voice assistants and real-time spoken interaction systems.

Current approaches to turn-taking often rely on acoustic cues such as silence duration, pitch, energy, or prosodic patterns.  Some methods integrate linguistic information, for example lexical content, syntactic completion, or dialogue-act prediction, and more recent systems also explore multimodal signals, including gaze, facial expressions, gestures, and head movements. However, combining these various signals effectively remains challenging, especially when real-time response and computational efficiency are required.

The goal of this internship is to explore and develop novel turn-taking models that integrate acoustic, linguistic, and potentially, non-verbal signals, to achieve more accurate and natural turn predictions. A particular emphasis will be placed on adaptability to different scenarios and on maintaining computational efficiency to enable real-time performance. The intern will review relevant literature and datasets, implement baseline and advanced models, conduct experiments on suitable dialogue corpora, and evaluate improvements in objective metrics (e.g., turn prediction accuracy, false interruption rate) as well as subjective naturalness.