2026 FinVolution Global Data Science Competition: Enhancing Voice AI with Turn-Taking Modeling in Conversations
FinVolution Group has recently launched the 2026 FinVolution Global Data Science Competition, focusing on turn-taking modeling in conversations to enhance voice AI capabilities. The challenge aims to improve AI's ability to determine when to speak during interactions, mimicking human conversational instincts. Participants will analyze dual-channel dialogue data to predict speech events within 800 milliseconds, enabling AI to respond appropriately to user intent.
The dataset for the competition consists of real dual-channel telephone conversations from 35 regions in China, featuring various dialects and speaking styles. Participants can utilize audio recordings with ASR transcripts and word-level timestamps to develop audio-based or multimodal systems. The competition serves as a bridge between academic research and practical applications, addressing the challenge of turn-taking in voice interactions to enhance the naturalness of conversations for users.
Supported by the China Computer Federation (CCF) Technical Committee on Natural Language Processing and Fudan University's Natural Language Processing Lab, the 2026 challenge is an official partner competition of the 15th CCF International Conference on Natural Language Processing and Chinese Computing (NLPCC 2026). Top-performing teams will have the opportunity to present their work at NLPCC 2026, engaging with the global NLP research community to advance voice AI technology.
The competition timeline includes three stages: Preliminary Round (May 13 – June 19), where participants train locally and submit prediction results for real-time scoring; Semifinals (June 20 – July 16), requiring semifinalists to submit Docker images for evaluation; and Finals (July 17 – August 13), where top teams will compete for a prize pool of RMB 308,000 (approximately USD 42,900). The competition aims to drive innovation in voice AI technology and enhance user experiences through more intuitive and human-like interactions.