Highlights from Ctrip Technology Center Deep Learning Meetup in Shanghai

Today, the Deep Learning Meetup organized by Ctrip Technology Center was held at Lingkong SOHO. As Shanghai's premier deep learning conference, it attracted engineers and professors from BI, machine learning, big data and other fields, with a fully packed venue and many international participants.

The speakers were highly distinguished, including academic representatives from University College London, Fudan University, Southeast University, Nanjing University, as well as technical experts from leading companies such as Ctrip, Huawei, Sogou and JianShu, who shared deep learning developments and practical applications across tourism, computational advertising, natural language processing, computer vision, and knowledge graphs.

The event fostered a strong technical community atmosphere, with professional presentations on stage, attentive note‑taking by the audience, interactive discussions, and enthusiastic applause.

A Ctrip Technology Center leader noted that, compared with Beijing, Shanghai has fewer technical sharing events, and Ctrip, as a leading internet company in Shanghai, is eager to build platforms that promote exchange and create a vibrant sharing atmosphere.

The meetup covered a wide range of topics:

Li Jian, Director of Ctrip Guide Community Development introduced deep learning applications in the guide community, covering NLP tasks such as sentiment analysis and address extraction, image description with facial expressions, video, and data content, and highlighted future focuses like knowledge graphs and virtual reality.

Liu Xiaohua, Researcher at Huawei Noah’s Ark Lab (Voice & Semantic) presented Huawei’s deep learning advances for voice and semantic tasks, discussing machine translation, natural language reasoning, and future trends in deep learning.

Wei Xiucan, PhD student, Computer Science Department, Nanjing University (LAMDA group) described research on fine‑grained image tasks, detailing deep convolutional feature selection and fusion to improve recognition accuracy for objects such as dogs, kangaroos, birds, and flowers.

Qi Guilin, Professor at Southeast University explored reasoning techniques in knowledge graphs and their application to a high‑school exam robot, explaining the concept of knowledge graphs and the goal of building intelligent tutoring robots.

Zhang Weinan, PhD candidate at University College London gave a detailed comparison of FM and FNN algorithms for multi‑value classification, contrasting them with neural networks, and demonstrated their performance on an online advertising click‑through prediction task alongside LR, CCPM, and PNN‑I.

Wu Zhonghuo, Senior Data Analyst, Ctrip Basic Business R&D introduced the architecture and mechanisms of Ctrip’s recommendation algorithms, showcasing how different business lines benefit from these algorithms and sharing cutting‑edge academic research in deep learning.

Shu Peng, Senior Researcher at Sogou explained deep learning applications in search advertising, detailing the ad business logic, the architecture of Sogou’s search ad system, multi‑model fusion for CTR prediction, evaluation methods, and discussed future integration of deep learning with CTR estimation.

Zhu Xiaohu, Data Scientist at JianShu presented an overview of leading experts and frontier projects in deep learning, covering RL, DRL, DQN, DDPG, their suitable scenarios, algorithm optimization, as well as AI safety and the impact on humanity.

Cui Wanyun, PhD candidate at Fudan University described QA systems based on knowledge graphs, outlining a three‑layer architecture (semantic entity layer, semantic short‑text layer, application system layer) and the techniques for entity recognition, template extraction, index building, and answer retrieval, concluding with a demo of their QA system.

The second half of the year will see more technical sharing sessions organized by Ctrip Technology Center across various professional fields, inviting interested peers to follow the official WeChat account "ctriptech" for updates.