Figure 5
The figure shows seven circles. Each circle has nodes inside it, and all the nodes are marked with an inverted triangle. All the nodes are densely connected to each other with lines. The largest circle at the top is labeled “Linguistics Acoustics Computer Science, Interdisciplinary Applications.” Below this, the second largest circle is oriented and labeled “Mathematics, Applied Statistics and Probability Engineering, Multidisciplinary.” On the top right, the third largest circle is labeled “Business Health Care Sciences and Services Computer Science, Information Systems.” Below this, in the center, a medium-sized circle is labeled “Psychology, Multidisciplinary Communication Social Sciences, Interdisciplinary.” On the right center, another medium-sized circle is labeled “Imaging Science and Photographic Technology Radiology, Nuclear Medicine and Medical Imaging Transportation Science and Technology.” On the left bottom, a small circle is labeled “Social Sciences, Mathematical Methods Engineering, Aerospace Telecommunications.” At the bottom, the last circle is labeled “Thermodynamics, Computer Science, Hardware and Architecture, Engineering, Mechanical.” Some of the nodes are as follows: “adversarial examples,” “such logs,” “click models,” “recommendation,” “search engine queries,” “search,” “commercial search engine,” “search engine,” “machine learning,” “link prediction,” “mobile devices,” “data-driven,” “data sources,” “mobility data,” “service providers,” “gradient descent,” “big data area,” “speech recognition,” “object detection,” “translation model,” “short texts,” “novel algorithm,” “translation quality,” “neural machine translation,” “question representation,” “translation language model,” “word embeddings,” “3d print cloud,” “strong baselines,” “point cloud,” “ranking model,” “maximum likelihood,” “neural network,” “convolutional neural network,” “real-world data set,” “learning models,” “classification problem,” “method classification,” “network structure,” “matrix factorization,” “labeled data,” “query,” “auxiliary data,” “user preference learning methods,” “domain knowledge,” “input data,” “top global search engine,” “image recognition,” “truth discovery,” “data points,” “gradient method,” “data streams,” “convergence rate,” “deep convolutional neural networks,” “reinforcement learning,” “input image,” “computer vision,” “facial expression recognition,” “single image,” “depth estimation,” “memory accesses,” “data analysis,” “storage,” “system,” “data center,” “cloud computing,” “data center edge efficiency,” “high performance,” “energy efficiency,” “airflow,” “power consumption,” “total cost of ownership,” “real estate,” and so on.Baidu’s network map of the most frequent topics in scientific publications from 2013 to 2018. Source: Author’s analysis based on data extracted from Web of Science