The global COVID-19 outbreak is worrisome both for its high rate of spread, and the high case fatality rate reported by early studies and now in Italy. We report a new methodology, the Patient Information Based Algorithm (PIBA), for estimating the death rate of a disease in real-time using publicly available data collected during an outbreak. PIBA estimated the death rate based on data of the patients in Wuhan and then in other cities throughout China. The estimated days from hospital admission to death was 13 (standard deviation (SD), 6 days). The death Science of the Total Environment 727 (2020) 138394 rates based on PIBA were used to predict the daily numbers of deaths since the week of February 25, 2020, in China overall, Hubei province, Wuhan city, and the rest of the country except Hubei province. The death rate of COVID-19 ranges from 0.75% to 3% and may decrease in the future. The results showed that the real death numbers had fallen into the predicted ranges. In addition, using the preliminary data from China, the PIBA method was successfully used to estimate the death rate and predict the death numbers of the Korean population. In conclusion, PIBA can be used to efficiently estimate the death rate of a new infectious disease in real-time and to predict future deaths. The spread of 2019-nCoV and its case fatality rate may vary in regions with different climates and temperatures from Hubei and Wuhan. PIBA model can be built based on known information of early patients in different countries.

Improving disaster management and recovery techniques is one of national priorities given the huge toll caused by man-made and nature calamities. Data-driven disaster management aims at applying advanced data collection and analysis technologies to achieve more effective and responsive disaster management, and has undergone considerable progress in the last decade. However, to the best of our knowledge, there is currently no work that both summarizes recent progress and suggests future directions for this emerging research area. To remedy this situation, we provide a systematic treatment of the recent developments in data-driven disaster management. Specifically, we first present a general overview of the requirements and system architectures of disaster management systems and then summarize state-of-the-art data-driven techniques that have been applied on improving situation awareness as well as in addressing users’ information needs in disaster management. We also discuss and categorize general data-mining and machine-learning techniques in disaster management. Finally, we recommend several research directions for further investigations.

Map search is a major vertical in all popular search engines. It also plays an important role in personal assistants on mobile, home or desktop devices. A significant fraction of map search traffic is comprised of “address queries” - queries where either the entire query or some terms in it refer to an address or part of an address (road segment, intersection etc.). Here we demonstrate that correctly understanding and tagging address queries are critical for map search engines to fulfill them. We describe several recurrent sequence architectures for tagging such queries. We compare their performance on two subcategories of address queries - single entity (aka single point) addresses and multi entity (aka multi point) addresses, and finish by providing guidance on the best practices when dealing with each of these subcategories.

No abstract

Abstract-Hurricane Sandy affected the east coast of U.S. in 2012 and posed immense threats to businesses, human lives and properties. In order to minimize the consequent loss of a catastrophe like this, a critical task in disaster management is to understand situation updates about the disaster from a large number of disaster-related documents, and obtain a big picture of the disaster's trends and how it affects different areas. In this paper, we present a two-layer storyline generation framework which generates an overall or a global storyline of the disaster events in the first layer, and provides condensed information about specific regions affected by the disaster (i.e., a locationspecific storyline) in the second layer. To generate the overall storyline of a disaster, we consider both temporal and spatial factors, which are encoded using integer linear programming. While for location-specific storylines, we employ a Steiner tree based method. Compared with the previous work of storyline generation, which generates flat storylines without considering spatial information, our framework is more suitable for largescale disaster events. We further demonstrate the efficacy of our proposed framework through the evaluation on the datasets of three major hurricane disasters.