Online social and news media generate rich and timely information about real-world events of all kinds. However, the huge amount of data available, along with the breadth of the user base, requires a substantial effort of information filtering to successfully drill down to relevant topics and events. Trending topic detection is therefore a fundamental building block to monitor and summarize information originating from social sources. There are a wide variety of methods and variables and they greatly affect the quality of results. We compare six topic detection methods on three Twitter datasets related to major events, which differ in their time scale and topic churn rate. We observe how the nature of the event considered, the volume of activity over time, the sampling procedure and the pre-processing of the data all greatly affect the quality of detected topics, which also depends on the type of detection method used. We find that standard natural language processing techniques can perform well for social streams on very focused topics, but novel techniques designed to mine the temporal distribution of concepts are needed to handle more heterogeneous streams containing multiple stories evolving in parallel. One of the novel topic detection methods we propose, based on n-grams cooccurrence and df -idft topic ranking, consistently achieves the best performance across all these conditions, thus being more reliable than other state-of-the art techniques.

International audienceSocial media have attracted considerable attention because their open-ended nature allows users to create lightweight semantic scaffolding to organize and share content. To date, the interplay of the social and topical components of social media has been only partially explored. Here, we study the presence of homophily in three systems that combine tagging social media with online social networks. We find a substantial level of topical similarity among users who are close to each other in the social network. We introduce a null model that preserves user activity while removing local correlations, allowing us to disentangle the actual local similarity between users from statistical effects due to the assortative mixing of user activity and centrality in the social network. This analysis suggests that users with similar interests are more likely to be friends, and therefore topical similarity measures among users based solely on their annotation metadata should be predictive of social links. We test this hypothesis on several datasets, confirming that social networks constructed from topical similarity capture actual friendship accurately. When combined with topological features, topical similarity achieves a link prediction accuracy of about 92%

Non Fungible Tokens (NFTs) are digital assets that represent objects like art, collectible, and in-game items. They are traded online, often with cryptocurrency, and are generally encoded within smart contracts on a blockchain. Public attention towards NFTs has exploded in 2021, when their market has experienced record sales, but little is known about the overall structure and evolution of its market. Here, we analyse data concerning 6.1 million trades of 4.7 million NFTs between June 23, 2017 and April 27, 2021, obtained primarily from Ethereum and WAX blockchains. First, we characterize statistical properties of the market. Second, we build the network of interactions, show that traders typically specialize on NFTs associated with similar objects and form tight clusters with other traders that exchange the same kind of objects. Third, we cluster objects associated to NFTs according to their visual features and show that collections contain visually homogeneous objects. Finally, we investigate the predictability of NFT sales using simple machine learning algorithms and find that sale history and, secondarily, visual features are good predictors for price. We anticipate that these findings will stimulate further research on NFT production, adoption, and trading in different contexts.

Background and aims There is poor knowledge on characteristics, comorbidities and laboratory measures associated with risk for adverse outcomes and in-hospital mortality in European Countries. We aimed at identifying baseline characteristics predisposing COVID-19 patients to in-hospital death. Methods and results Retrospective observational study on 3894 patients with SARS-CoV-2 infection hospitalized from February 19th to May 23rd, 2020 and recruited in 30 clinical centres distributed throughout Italy. Machine learning (random forest)-based and Cox survival analysis. 61.7% of participants were men (median age 67 years), followed up for a median of 13 days. In-hospital mortality exhibited a geographical gradient, Northern Italian regions featuring more than twofold higher death rates as compared to Central/Southern areas (15.6% vs 6.4%, respectively). Machine learning analysis revealed that the most important features in death classification were impaired renal function, elevated C reactive protein and advanced age. These findings were confirmed by multivariable Cox survival analysis (hazard ratio (HR): 8.2; 95% confidence interval (CI) 4.6–14.7 for age ≥85 vs 18–44 y); HR = 4.7; 2.9–7.7 for estimated glomerular filtration rate levels <15 vs ≥ 90 mL/min/1.73 m 2 ; HR = 2.3; 1.5–3.6 for C-reactive protein levels ≥10 vs ≤ 3 mg/L). No relation was found with obesity, tobacco use, cardiovascular disease and related-comorbidities. The associations between these variables and mortality were substantially homogenous across all sub-groups analyses. Conclusions Impaired renal function, elevated C-reactive protein and advanced age were major predictors of in-hospital death in a large cohort of unselected patients with COVID-19, admitted to 30 different clinical centres all over Italy.

Background Hydroxychloroquine (HCQ) was proposed as potential treatment for COVID-19. Objective We set-up a multicenter Italian collaboration to investigate the relationship between HCQ therapy and COVID-19 in-hospital mortality. Methods In a retrospective observational study, 3,451 unselected patients hospitalized in 33 clinical centers in Italy, from February 19, 2020 to May 23, 2020, with laboratory-confirmed SARS-CoV-2 infection, were analyzed. The primary end-point in a time-to event analysis was in-hospital death, comparing patients who received HCQ with patients who did not. We used multivariable Cox proportional-hazards regression models with inverse probability for treatment weighting by propensity scores, with the addition of subgroup analyses. Results Out of 3,451 COVID-19 patients, 76.3% received HCQ. Death rates (per 1,000 person-days) for patients receiving or not HCQ were 8.9 and 15.7, respectively. After adjustment for propensity scores, we found 30% lower risk of death in patients receiving HCQ (HR=0.70; 95%CI: 0.59 to 0.84; E-value=1.67). Secondary analyses yielded similar results. The inverse association of HCQ with inpatient mortality was particularly evident in patients having elevated C-reactive protein at entry. Conclusions HCQ use was associated with a 30% lower risk of death in COVID-19 hospitalized patients. Within the limits of an observational study and awaiting results from randomized controlled trials, these data do not discourage the use of HCQ in inpatients with COVID-19.

Introduction: A hypercoagulable condition was described in patients with COVID-19 and proposed as a possible pathogenic mechanism contributing to disease progression and lethality. Aim: We evaluated if in-hospital administration of heparin improved survival in a large cohort of Italian COVID-19 patients. Methods: In a retrospective observational study, 2,574 unselected patients hospitalised in 30 clinical centres in Italy from February 19, 2020 to May 23, 2020 with laboratory-confirmed SARS-CoV-2 infection, were analysed. The primary end-point in a time-to event analysis was in-hospital death, comparing patients who received heparin (low-molecular weight heparin (LMWH) or unfractionated heparin (UFH)) with patients who did not. We used multivariable Cox proportional-hazards regression models with inverse probability for treatment weighting by propensity scores. Results: Out of 2,574 COVID-19 patients, 70.1% received heparin. LMWH was largely the most used formulation (99.5%). Death rates for patients receiving heparin or not were 7.4 and 14.0 per 1,000 person-days, respectively. After adjustment for propensity scores, we found a 40% lower risk of death in patients receiving heparin (HR=0.60; 95%CI: 0.49 to 0.74; E-value=2.04). This association was particularly evident in patients with a higher severity of disease or strong coagulation activation. Conclusions: In-hospital heparin treatment was associated with lower mortality, particularly in severely ill COVID-19 patients and in those with strong coagulation activation. The results from randomised clinical trials are eagerly awaited to provide clear-cut recommendations.

Machine learning and AI-assisted trading have attracted growing interest for the past few years.Here, we use this approach to test the hypothesis that the inefficiency of the cryptocurrency market can be exploited to generate abnormal profits. We analyse daily data for 1, 681 cryptocurrencies for the period between Nov. 2015 and Apr. 2018. We show that simple trading strategies assisted by state-of-the-art machine learning algorithms outperform standard benchmarks. Our results show that non-trivial, but ultimately simple, algorithmic mechanisms can help anticipate the short-term evolution of the cryptocurrency market.

1 The e ects of ethyl alcohol and wine (red and white) on haemostatic parameters and experimental thrombosis were studied in rats; NO was evaluated as a possible mediator of these e ects. 2 We found that red wine (12% alcohol) supplementation (8.4+0.4 ml d 71 in drinking water, for 10 days) induced a marked prolongation of`template' bleeding time (BT) (258+13 vs 132+13 s in controls; P50.001), a decrease in platelet adhesion to ®brillar collagen (11.6+1.0 vs 32.2+1.3%; P50.01) and a reduction in thrombus weight (1.45+0.33 vs 3.27+0.39 mg; P50.01). 3 Alcohol-free red wine showed an e ect similar to red wine. In contrast, neither ethyl alcohol (12%) nor white wine (12% alcohol) a ected these systems. 4 All these e ects were also observed after red wine i.v. injection (1 ml kg 71 of 1 : 4 dilution) 15 min before the experiments.5 The e ects of red wine were prevented by the NO inhibitor, N o nitro-L-arginine-methyl ester (L-NAME). L-arginine, not D-arginine, reversed the e ect of L-NAME on red wine infusion. 6 Red wine injection induced a 3 fold increase in total radical-trapping antioxidant parameter values of rat plasma with respect to controls, while white wine and alcohol did not show any e ect. 7 Our study provides evidence that red wine modulates primary haemostasis and prevents experimental thrombosis in rats, independently of its alcohol content, by a NO-mediated mechanism.