Embodied visual exploration is critical for building intelligent visual agents. This paper presents the neural exploration with feature-based visual odometry and tracking-failure-reduction policy(Ne OR), a framework f...Embodied visual exploration is critical for building intelligent visual agents. This paper presents the neural exploration with feature-based visual odometry and tracking-failure-reduction policy(Ne OR), a framework for embodied visual exploration that possesses the efficient exploration capabilities of deep reinforcement learning(DRL)-based exploration policies and leverages feature-based visual odometry(VO) for more accurate mapping and positioning results. An improved local policy is also proposed to reduce tracking failures of feature-based VO in weakly textured scenes through a refined multi-discrete action space, keyframe fusion, and an auxiliary task. The experimental results demonstrate that Ne OR has better mapping and positioning accuracy compared to other entirely learning-based exploration frameworks and improves the robustness of feature-based VO by significantly reducing tracking failures in weakly textured scenes.展开更多
Online job advertisements on various job portals or websites have become the most popular way for people to find potential career opportunities nowadays.However,the majority of these job sites are limited to offering ...Online job advertisements on various job portals or websites have become the most popular way for people to find potential career opportunities nowadays.However,the majority of these job sites are limited to offering fundamental filters such as job titles,keywords,and compensation ranges.This often poses a challenge for job seekers in efficiently identifying relevant job advertisements that align with their unique skill sets amidst a vast sea of listings.Thus,we propose well-coordinated visualizations to provide job seekers with three levels of details of job information:a skill-job overview visualizes skill sets,employment posts as well as relationships between them with a hierarchical visualization design;a post exploration view leverages an augmented radar-chart glyph to represent job posts and further facilitates users’swift comprehension of the pertinent skills necessitated by respective positions;a post detail view lists the specifics of selected job posts for profound analysis and comparison.By using a real-world recruitment advertisement dataset collected from 51Job,one of the largest job websites in China,we conducted two case studies and user interviews to evaluate JobViz.The results demonstrated the usefulness and effectiveness of our approach.展开更多
With the rapid development of Web2.0 technology, more and more social annotation systems are emerging, such as Del.icio.us, Flickr, YouTube, and CiteULike. These systems help users to manage and share their digital re...With the rapid development of Web2.0 technology, more and more social annotation systems are emerging, such as Del.icio.us, Flickr, YouTube, and CiteULike. These systems help users to manage and share their digital resources, and have attracted a lot of users to annotate the resources with tags and bookmarks, which result in a large scale of tag data. Due to the exponential increase of social annotations, all the users are facing the same problem: How can we explore the desired resources efficiently in such a large tag dataset? Since the traditional methods such as tag cloud view and annotation match work well only in small annotation dataset, this paper studies the relationships of tag-tag, tag-resource and resource-resource through the co-occurrences and proposes a new efficient way for users to organize and explore the literature resources. Our research mainly focuses on two aspects:1) The hidden semantic relationships of popular tags and their relevant literature resources;2) the computing of literature resources similarity given a specific literature. A prototype system named PKUSpace is implemented and shows promising results.展开更多
Drug taxonomy could be described as an inherent structure of different pharmaceutical componential drugs. Unfortunately, the literature does not always provide a clear path to define and classify adverse drug events. ...Drug taxonomy could be described as an inherent structure of different pharmaceutical componential drugs. Unfortunately, the literature does not always provide a clear path to define and classify adverse drug events. While not a systematic review, this paper uses examples from the literature to illustrate problems that investigators will confront as they develop a conceptual framework for their research. It also proposes a targeted taxonomy that can facilitate a clear and consistent approach to understanding different drugs and could aid in the comparison to results of past and future studies. In terms of building the drugs taxonomy, symptoms information were selected, clustered and adapted for this purpose. Finally, although national or international agreement on taxonomy for different drugs is a distant or unachievable goal, individual investigations and the literature as a whole will be improved by prospective, explicit classification of different drugs using this new pharmacy information system （PIS） and inclusion of the study＇s approach to classification in publications. The PIS allows user to find information quickly by following semantic connections that surround every drug linked to the subject. It provides quicker search, faster and more intuitive understanding of the focus. This research work can pretend to become a leading provider of encyclopedia service for scientists and educators, as well as attract the scientific community-universities, research and development groups.展开更多
Visual analysis of social networks is usually based on graph drawing algorithms and tools. However, social networks are a special kind of graph in the sense that interpretation of displayed relationships is heavily de...Visual analysis of social networks is usually based on graph drawing algorithms and tools. However, social networks are a special kind of graph in the sense that interpretation of displayed relationships is heavily dependent on context. Context, in its turn, is given by attributes associated with graph elements, such as individual nodes, edges, and groups of edges, as well as by the nature of the connections between individuals. In most systems, attributes of individuals and communities are not taken into consideration during graph layout, except to derive weights for force-based placement strategies. This paper proposes a set of novel tools for displaying and exploring social networks based on attribute and connectivity mappings. These properties are employed to layout nodes on the plane via multidimensional projection techniques. For the attribute mapping, we show that node proximity in the layout corresponds to similarity in atgribute, leading to easiness in locating similar groups of nodes. The projection based on connectivity yields an initial placement that forgoes force-based or graph analysis algorithm, reaching a meaningful layout in one pass. When a force algorithm is then applied to this initial mapping, the final layout presents better properties than conventional force-based approaches. Numerical evaluations show a number of advantages of pre-mapping points via projections. User evaluation demonstrates that these tools promote ease of manipulation as well as fast identification of concepts and associations which cannot be easily expressed by conventional graph visualization alone. In order to allow better space usage for complex networks, a graph mapping on the surface of a sphere is also implemented.展开更多
基金supported by the National Natural Science Foundation of China (No.62202137)the China Postdoctoral Science Foundation (No.2023M730599)the Zhejiang Provincial Natural Science Foundation of China (No.LMS25F020009)。
文摘Embodied visual exploration is critical for building intelligent visual agents. This paper presents the neural exploration with feature-based visual odometry and tracking-failure-reduction policy(Ne OR), a framework for embodied visual exploration that possesses the efficient exploration capabilities of deep reinforcement learning(DRL)-based exploration policies and leverages feature-based visual odometry(VO) for more accurate mapping and positioning results. An improved local policy is also proposed to reduce tracking failures of feature-based VO in weakly textured scenes through a refined multi-discrete action space, keyframe fusion, and an auxiliary task. The experimental results demonstrate that Ne OR has better mapping and positioning accuracy compared to other entirely learning-based exploration frameworks and improves the robustness of feature-based VO by significantly reducing tracking failures in weakly textured scenes.
基金founded by Huazhong University of Science and Technology Teaching Research Project number(s):2023100.
文摘Online job advertisements on various job portals or websites have become the most popular way for people to find potential career opportunities nowadays.However,the majority of these job sites are limited to offering fundamental filters such as job titles,keywords,and compensation ranges.This often poses a challenge for job seekers in efficiently identifying relevant job advertisements that align with their unique skill sets amidst a vast sea of listings.Thus,we propose well-coordinated visualizations to provide job seekers with three levels of details of job information:a skill-job overview visualizes skill sets,employment posts as well as relationships between them with a hierarchical visualization design;a post exploration view leverages an augmented radar-chart glyph to represent job posts and further facilitates users’swift comprehension of the pertinent skills necessitated by respective positions;a post detail view lists the specifics of selected job posts for profound analysis and comparison.By using a real-world recruitment advertisement dataset collected from 51Job,one of the largest job websites in China,we conducted two case studies and user interviews to evaluate JobViz.The results demonstrated the usefulness and effectiveness of our approach.
基金supported by the Specialized Research Fund for the Doctoral Program of Higher Education(Grant No.20070001073)the National Natural Science Foundation of China(Grant Nos.90412010 and60773162)
文摘With the rapid development of Web2.0 technology, more and more social annotation systems are emerging, such as Del.icio.us, Flickr, YouTube, and CiteULike. These systems help users to manage and share their digital resources, and have attracted a lot of users to annotate the resources with tags and bookmarks, which result in a large scale of tag data. Due to the exponential increase of social annotations, all the users are facing the same problem: How can we explore the desired resources efficiently in such a large tag dataset? Since the traditional methods such as tag cloud view and annotation match work well only in small annotation dataset, this paper studies the relationships of tag-tag, tag-resource and resource-resource through the co-occurrences and proposes a new efficient way for users to organize and explore the literature resources. Our research mainly focuses on two aspects:1) The hidden semantic relationships of popular tags and their relevant literature resources;2) the computing of literature resources similarity given a specific literature. A prototype system named PKUSpace is implemented and shows promising results.
文摘Drug taxonomy could be described as an inherent structure of different pharmaceutical componential drugs. Unfortunately, the literature does not always provide a clear path to define and classify adverse drug events. While not a systematic review, this paper uses examples from the literature to illustrate problems that investigators will confront as they develop a conceptual framework for their research. It also proposes a targeted taxonomy that can facilitate a clear and consistent approach to understanding different drugs and could aid in the comparison to results of past and future studies. In terms of building the drugs taxonomy, symptoms information were selected, clustered and adapted for this purpose. Finally, although national or international agreement on taxonomy for different drugs is a distant or unachievable goal, individual investigations and the literature as a whole will be improved by prospective, explicit classification of different drugs using this new pharmacy information system （PIS） and inclusion of the study＇s approach to classification in publications. The PIS allows user to find information quickly by following semantic connections that surround every drug linked to the subject. It provides quicker search, faster and more intuitive understanding of the focus. This research work can pretend to become a leading provider of encyclopedia service for scientists and educators, as well as attract the scientific community-universities, research and development groups.
基金FAPESP, CNPq and CAPES for their financial support
文摘Visual analysis of social networks is usually based on graph drawing algorithms and tools. However, social networks are a special kind of graph in the sense that interpretation of displayed relationships is heavily dependent on context. Context, in its turn, is given by attributes associated with graph elements, such as individual nodes, edges, and groups of edges, as well as by the nature of the connections between individuals. In most systems, attributes of individuals and communities are not taken into consideration during graph layout, except to derive weights for force-based placement strategies. This paper proposes a set of novel tools for displaying and exploring social networks based on attribute and connectivity mappings. These properties are employed to layout nodes on the plane via multidimensional projection techniques. For the attribute mapping, we show that node proximity in the layout corresponds to similarity in atgribute, leading to easiness in locating similar groups of nodes. The projection based on connectivity yields an initial placement that forgoes force-based or graph analysis algorithm, reaching a meaningful layout in one pass. When a force algorithm is then applied to this initial mapping, the final layout presents better properties than conventional force-based approaches. Numerical evaluations show a number of advantages of pre-mapping points via projections. User evaluation demonstrates that these tools promote ease of manipulation as well as fast identification of concepts and associations which cannot be easily expressed by conventional graph visualization alone. In order to allow better space usage for complex networks, a graph mapping on the surface of a sphere is also implemented.
