Similarity in Visual Analytics—A Visual Analytics Approach for Finding Similar Publications

Recent studies show that the search for relevant publications requires researchers to invest a significant amount of time and work; some studies even underline that this task requires the most time and work out of all the tasks in the entire research process. To reduce the search time in the research process, we propose a visual analytics approach that combines models and methods of natural language process, machine learning, similarity measures, and interactive visual representations. The proposed approach is based on our previous works and enhances those with additional automatic assistance during the entire search process. Our visual analytics approach facilitates the presentation of large amounts of relevant results similar to the identified topics of interest and tailored to the needs of researchers during their research process. The proposed method enables annotation in the context of exploration, allowing researchers to quickly find and bookmark relevant publications during exploration and, by doing so, improve the publication recommendations. We analyze the annotations researchers make throughout their research journey to identify the topics of interest and use them as input for our learning and measurement methods. By utilizing researchers’ commonly observed annotation and exploration behavior, our approach counters information overload by generating labeled vectors of interest and providing similar publications.