Reliability Aware Data Fusion

Due to ubiquitous sensors (GPS, Accelerometer), easy of use apps (Facebook, Twitter etc), presence of audio & video recording devices and higher internet connectivity, the key characteristics of raw data is changing. This new data can be characterized by 4Vs Volume, Velocity, Variety and Veracity. Moreover, due to popular trend of crowd sourcing or citizen sensors, it is reasonable to assume that people will provide multiple evidence of same event using different data types. For example during a Football match, some people will Tweet about Goals, Penalties etc while others will take a picture and upload it. Although the underlying modalities are different (text and image), the data describes the same event. Such multi modal evidences should be used to strengthen the belief in underlying physical event. Finally, each of the data point will have inherent uncertainty. The uncertainty can arise from inconsistent, incomplete, and ambiguous data as well as the trust worthiness of the user. Similarly, some sources are more reliable than others which will also play a part in overall reliability. The volume, velocity and variety are measurable/observable, however, there is no measure of truthfulness. Traditionally, CS research has focused on Volume and Velocity. However, multimodal data fusion and reliability are less explored. Through this tutorial will wish to draw the attention of researchers towards these dimensions by presenting some real life use cases, highlighting the key technical challenges, existing techniques and need for new .