A human-oriented mutual assistive framework using collaborative filtering towards disasters

Originally, collaborative filtering was adopted in purchase recommendation systems (e.g., Amazon.com) based on purchased history. In this paper, we apply collaborative filtering on the basis of accumulated feedbacks of the data extracted from social media from a community of users to build up a knowledge-based framework that can match offers to needs in disaster and emergency situations. This framework is constructed by high-level data fusion, i.e., incorporating text-based natural language processing with image-based processing using long-term relevance feedback, and learns user's preferences and adjusts their needs and offers accordingly. It can be deemed as a fundamental trial for timely mutual assist in disasters.