The immense growth of the social Web, which has made a large amount of user data easily and publicly available, has opened a whole new spectrum for research in social behavioral sciences. However, as the volume of social media content increases at a very fast rate, it becomes extremely difficult to systematically obtain high-level information from this data. As a consequence, tasks related to the analysis of historical news events based on social media data have not been explored, which limits any type of comparative historical research, causality analysis, and discovery of knowledge from patterns extracted from aggregated social media event information.In this work, we target this issue by proposing a compact high-level representation of news events using social media information. This representation explicitly includes temporal information about the event and information about locations, in particular of geopolitical entities. We call this a spatio-temporal context-aware event representation. Our hypothesis is that by including social, temporal, and spatial information in the event representation, we are enabling the analysis of historical world news from a social and geopolitical perspective. This facilitates, new information retrieval tasks related to historical event information extraction and international relations analysis. We support our claims by presenting two applications of this idea: the first, a visual tool, named Galean, for retrieval and exploration of historical news events within their geopolitical and temporal context. The second, a quantitative analysis of a 2-year Twitter dataset of news events reported by U.S. and U.K. media, which we explore using data mining techniques on our event representations. We present two case studies of event exploration using Galean and user evaluation of this tool, as well as details of our data mining empirical results.

By modeling macro-economical indicators using digital traces of human activities on mobile or social networks, we can provide important insights to processes previously assessed via paper-based surveys or polls only. We collected aggregated workday activity timelines of US counties from the normalized number of messages sent in each hour on the online social network Twitter. In this paper, we show how county employment and unemployment statistics are encoded in the daily rhythm of people by decomposing the activity timelines into a linear combination of two dominant patterns. The mixing ratio of these patterns defines a measure for each county, that correlates significantly with employment (0.46pm0.02) and unemployment rates (-0.34pm0.02). Thus, the two dominant activity patterns can be linked to rhythms signaling presence or lack of regular working hours of individuals. The analysis could provide policy makers a better insight into the processes governing employment, where problems could not only be identified based on the number of officially registered unemployed, but also on the basis of the digital footprints people leave on different platforms.

Advances in computing power, natural language processing, and digitization of text now make it possible to study a culture's evolution through its texts using a "big data" lens. Our ability to communicate relies in part upon a shared emotional experience, with stories often following distinct emotional trajectories and forming patterns that are meaningful to us. Here, by classifying the emotional arcs for a filtered subset of 1,327 stories from Project Gutenberg's fiction collection, we find a set of six core emotional arcs which form the essential building blocks of complex emotional trajectories. We strengthen our findings by separately applying Matrix decomposition, supervised learning, and unsupervised learning. For each of these six core emotional arcs, we examine the closest characteristic stories in publication today and find that particular emotional arcs enjoy greater success, as measured by downloads.

Community detection techniques are widely used to infer hidden structures within interconnected systems. Despite demonstrating high accuracy on benchmarks, they reproduce the external classification for many real-world systems with a significant level of discrepancy. A widely accepted reason behind such outcome is the unavoidable loss of non-topological information (such as node attributes) encountered when the original complex system is represented as a network. In this article we emphasize that the observed discrepancies may also be caused by a different reason: the external classification itself. For this end we use scientific publication data which i) exhibit a well defined modular structure and ii) hold an expert-made classification of research articles. Having represented the articles and the extracted scientific concepts both as a bipartite network and as its unipartite projection, we applied modularity optimization to uncover the inner thematic structure. The resulting clusters are shown to partly reflect the author-made classification, although some significant discrepancies are observed. A detailed analysis of these discrepancies shows that they carry essential information about the system, mainly related to the use of similar techniques and methods across different (sub)disciplines, that is otherwise omitted when only the external classification is considered.

This paper introduces a data-driven methodology to study the historical evolution of mathematical thinking and its spatial spreading. To do so, we have collected and integrated data from different online academic datasets. In its final form, the database includes a large number ( $N\sim200\mbox{K}$ ) of advisor-student relationships, with affiliations and keywords on their research topic, over several centuries, from the 14th century until today. We focus on two different issues, the evolving importance of countries and of the research disciplines over time. Moreover we study the database at three levels, its global statistics, the mesoscale networks connecting countries and disciplines, and the genealogical level.

In the last few years thousands of scientific papers have investigated sentiment analysis, several startups that measure opinions on real data have emerged and a number of innovative products related to this theme have been developed. There are multiple methods for measuring sentiments, including lexical-based and supervised machine learning methods. Despite the vast interest on the theme and wide popularity of some methods, it is unclear which one is better for identifying the polarity (i.e., positive or negative) of a message. Accordingly, there is a strong need to conduct a thorough apple-to-apple comparison of sentiment analysis methods, as they are used in practice, across multiple datasets originated from different data sources. Such a comparison is key for understanding the potential limitations, advantages, and disadvantages of popular methods. This article aims at filling this gap by presenting a benchmark comparison of twenty-four popular sentiment analysis methods (which we call the state-of-the-practice methods). Our evaluation is based on a benchmark of eighteen labeled datasets, covering messages posted on social networks, movie and product reviews, as well as opinions and comments in news articles. Our results highlight the extent to which the prediction performance of these methods varies considerably across datasets. Aiming at boosting the development of this research area, we open the methods’ codes and datasets used in this article, deploying them in a benchmark system, which provides an open API for accessing and comparing sentence-level sentiment analysis methods.

Gender homophily, or the preference for interaction with individuals of the same gender, has been observed in many contexts, especially during childhood and adolescence. In this study we investigate such phenomenon by analyzing the interactions of the ∼10 million users of Tuenti, a Spanish social networking service popular among teenagers. In dyadic relationships we find evidence of higher gender homophily for women. We also observe a preference of users with more friends to connect to the opposite gender. A particularly marked gender difference emerges in signing up for the social networking service and adding the first friends, and in the interactions by means of wall messages. In these contexts we find evidence of a strong homophily for women, and little or no homophily for men. By examining the gender composition of triangle motifs, we observe a marked tendency of users to group into gender homogeneous clusters, with a particularly high number of male-only triangles. We show that age plays an important role in this context, with a tendency to higher homophily for young teenagers in both dyadic and triadic relationships. Our findings have implications for addressing gender gap issues, understanding adolescent online behavior and technology adoption, and modeling social networks.

Segregation is widespread in all realms of human society. Several influential studies have argued that intolerance is not a prerequisite for a segregated society, and that segregation can arise even when people generally prefer diversity. We investigated this paradox experimentally, by letting groups of high-school students play four different real-time interactive games. Incentives for neighbor similarity produced segregation, but incentives for neighbor dissimilarity and neighborhood diversity prevented it. The participants continued to move while their game scores were below optimal, but their individual moves did not consistently take them to the best alternative position. These small differences between human and simulated agents produced different segregation patterns than previously predicted, thus challenging conclusions about segregation arising from these models.