Recap of the First Global Task‑Oriented Multi‑Turn Dialogue System Competition (JDDC)

The first worldwide Task‑Oriented Multi‑Turn Dialogue System Challenge, JDDC, has concluded, and this report provides a comprehensive recap of the event, including opening remarks, competition details, award ceremonies, and forward‑looking discussions.

JD.com Chief Technology Officer Zhang Chen emphasized that the next shift in human‑computer interaction is natural voice interaction, stressing the importance of multi‑turn and multi‑person dialogue as the core of future conversational AI.

Vice President of JD.com’s AI Platform and Research Department, Dr. Zhou Bowen, introduced the concept of moving from "Conversation as a Service" to "Service as a Conversation," highlighting how AI‑driven dialogue can proactively understand and fulfill user needs across various service industries.

The competition, launched on April 15, 2018, released a million‑scale, anonymized real‑customer dialogue dataset and provided GPU resources for six months. Participants were required to build multi‑turn dialogue systems, with task‑oriented dialogue identified as the most complex scenario.

Key speakers such as Hu Yichuan (Co‑founder & CTO of Laiye Network Technology) and Liu Dan (Technical Director of JD.com’s Intelligent Dialogue R&D) discussed the challenges of combining algorithms, compute resources, and data, and the importance of integrating emotion detection and personalized assistance in intelligent customer service.

The preliminary round relied solely on machine evaluation, while the final round combined machine scores, JD.com’s gold‑level human客服, and expert judges. The champion team from Fudan University and Tongji University used a semantic‑matching model and excelled in human‑centric evaluation, securing the overall victory.

Other top teams from Peking University, the Chinese Academy of Sciences, Tsinghua University, and several other institutions received runner‑up and special innovation awards, recognizing achievements in technical progress, algorithmic innovation, and system architecture.

Judges highlighted that the best solutions balance algorithmic sophistication with thorough data and scenario analysis, noting that even non‑state‑of‑the‑art models can win when combined with effective data usage and user‑experience design.

The post‑award round‑table forum explored the future of intelligent dialogue, predicting a shift toward multimodal interaction, self‑learning, and personalization, with contributions from academia and industry leaders emphasizing explainability, logical behavior, and broader application beyond chatbots to routing, assistance, and management.

JD.com’s AI initiatives now span computer vision and natural language processing, with collaborations across leading universities and research labs, aiming to translate competition breakthroughs into real‑world product improvements that enhance user experience and drive industry advancement.