On April 10th, 2024, the 65thZhi-Xin forum was held in 117 Lecture hall at Zhixin building. This forum was jointly organized by the College of Electronic and Information Engineering, Tongji University and Shanghai Research Institute for Intelligent Autonomous Systems. Prof. Shi ZHOU from University College London, UK, was invited to give a talk on “Improving Opinion Measurement using Neural-Embedding based on Social Connectivity Data”, which was moderated by Prof. Hao Zhang.
Firstly, Prof. Shi ZHOU introduced the topic of opinion dynamics in social networks, gave a formal definition and practical meaning of "opinion", and pointed out that the current research in this field mainly based on unreliable or erroneous opinion polls and text analysis. Then, he presented in detail a viewpoint measurement method based on neural network embedding of social relationship data, with input data as a dichotomous network that ignores in-group social relationships. The method is divided into two main stages, including word2vec-based neural embedding, projection vector definition and relative relationship measurement. After that, Prof. Zhou discussed about projection vectors, including supervised learning methods based on prior knowledge to select vector clusters and unsupervised learning methods based on K-means clustering for class cluster center determination. Finally, Prof. Zhou experimentally verified the performance of the cluster-based embedding method and demonstrated the advantages of the method in terms of computational complexity.
After the talk, Prof. Zhou had a cordial exchange and discussion with teachers and students, and he also encouraged them to actively broaden their horizons, explore, discover, and solve new scientific problems based on his experience. This talk further expanded the vision of our teachers and students and enhanced their knowledge and understanding of the field of complex networks.
Written by Hao Zhang
Photographed by Yunjiao Wang/Zijian Jiang
