VAMD: Visual Analytics for Multimodal Data
John Fallon, R. Jordan Crouse · 2019
From social media to smart traffic signals, digital records of the state of the world are being produced at unprecedented rates. These data may contain critical information regarding patterns of life or unfolding events, but the necessary pieces may be dispersed across multiple disparate sources. To make matters worse, the data produced is by nature massive, heterogeneous, and imperfect. To deal with this, current solutions typically analyze each data source within its own class. This reduces the problem's complexity but forfeits the potential benefits of discovering nuanced connections and details encompassing an entire situation. In order to exploit these potentially rich (but frequently one-sided) data sources, it is imperative that there exists an environment and supporting tooling that allows for the analysis of heterogeneous data sources within a combined view. This paper presents VAMD, a visual analytics system focused on dealing with large heterogeneous datasets. The design is supported with a use case focused around a disaster response scenario using a curated social media dataset featuring both text and accompanying images.