Main purpose of current research is to present evolutions in previous presented approaches of the author for manipulating social media content for disaster management of natural events. Those innovations suggest the adoption of machine learning for classifying both photos and text posted in social networks along with hybrid geo-referencing. As case study the author chose the Ianos cyclone, occurred between Italy and Greece, during September 2020. The geographic focus of the research was in Greece where the cyclone caused 4 human losses and damages in the urban environment. A dataset consisted of 4655 photos, with their corresponding captions, timestamps and location information was crawled from Instagram. The main hashtag used was #Ianos. Two data samples, one for each type, were classified manually for calibrating the classification models. The classes regarding photos were initially: (i) related and (ii) not related to Ianos, while the general classification schema for photos and text was: (i) Ianos event identification, (ii) consequences, scaled according to the impact of each report, (iii) precaution, (iv) disaster management: announcements, measures, volunteered actions. Author’s approach regarding classification suggests the use of convolutional neural networks and support vector machine algorithms for image and text classification respectively. The classified dataset, was geo-referenced by using commercial geocoding API and list-based geoparsing. The results of the research in current status are at an initial level, a subset of data though of automatically or manually processed information is presented in four related maps.