A Unified Process for Visual-Interactive Labeling

Authors. Jürgen Bernard, Matthias Zeppelzauer, Michael Sedlmair, Wolfgang Aigner

Venue. EuroVA (2017)

Materials. DOI PDF

Abstract. Assigning labels to data instances is a prerequisite for many machine learning tasks. Similarly, labeling is applied in visual-interactive analysis approaches. However, the strategies for creating labels often differ in the two fields. In this paper, we study the process of labeling data instances with the user in the loop, from both the machine learning and visual-interactive perspective. Based on a review of differences and commonalities, we propose the 'Visual-Interactive Labeling' (VIAL) process, conflating the strengths of both. We describe the six major steps of the process and highlight their related challenges.

@inproceedings{10.2312/eurova.20171123,
    title        = {A Unified Process for Visual-Interactive Labeling},
    author       = {Bernard, J\"{u}rgen and Zeppelzauer, Matthias and Sedlmair, Michael and Aigner, Wolfgang},
    year         = {2017},
    booktitle    = {Proceedings of the EuroVis Workshop on Visual Analytics},
    publisher    = {Eurographics Association},
    series       = {EuroVA '17},
    pages        = {73–77},
    doi          = {10.2312/eurova.20171123},
    url          = {https://doi.org/10.2312/eurova.20171123},
    abstract     = {Assigning labels to data instances is a prerequisite for many machine learning tasks. Similarly, labeling is applied in visual-interactive analysis approaches. However, the strategies for creating labels often differ in the two fields. In this paper, we study the process of labeling data instances with the user in the loop, from both the machine learning and visual-interactive perspective. Based on a review of differences and commonalities, we propose the 'Visual-Interactive Labeling' (VIAL) process, conflating the strengths of both. We describe the six major steps of the process and highlight their related challenges.},
    numpages     = {5}
}