EMAssistant: A Learning Analytics System for Social and Web Data Filtering to Assist Trainees and Volunteers of Emergency Services

An increasing number of Machine Learning based systems are being designed to filter and visualize the relevant information from social media and web streams for disaster management. Given the dynamic disaster events, the notion of relevant information evolves, and thus, the active learning techniques are often considered to keep updating the predictive models for the relevant information filtering. However, the active relevant feedback provided by the human annotators to update the models are not validated. As a result, they can introduce unconscious biases in the learning process of humans and can result in an inaccurate or inefficient predictive system. Therefore, this paper describes the design and implementation of an open-source technology-based learning analytics system ' EMAssistant' for the emergency volunteers or practitioners - referred as the trainee, to enhance their experiential learning cycle with the cause-effect reasoning on providing relevant feedback to the machine learning model. This continuous integration between the cause (providing feedback) and the effect (observing predictions from the updated model) in a visual form will likely to improve the understanding of the trainees to provide more accurate feedback. We propose to present the system design as well as provide hands-on exercises for the conference session.