Terrorist attacks sharpen the binary perception of "Us" vs. "Them".

Terrorist attacks not only harm citizens but also shift their attention, which has long-lasting impacts on public opinion and government policies. Yet measuring the changes in public attention beyond media coverage has been methodologically challenging. Here we approach this problem by starting from Wikipedia's répertoire of 5.8 million articles and a sample of 15 recent terrorist attacks. We deploy a complex exclusion procedure to identify topics and themes that consistently received a significant increase in attention due to these incidents. Examining their contents reveals a clear picture: terrorist attacks foster establishing a sharp boundary between "Us" (the target society) and "Them" (the terrorist as the enemy). In the midst of this, one seeks to construct identities of both sides. This triggers curiosity to learn more about "Them" and soul-search for a clearer understanding of "Us". This systematic analysis of public reactions to disruptive events could help mitigate their societal consequences.

Large network community detection by fast label propagation.

Many networks exhibit some community structure. There exists a wide variety of approaches to detect communities in networks, each offering different interpretations and associated algorithms. For large networks, there is the additional requirement of speed. In this context, the so-called label propagation algorithm (LPA) was proposed, which runs in near-linear time. In partitions uncovered by LPA, each node is ensured to have most links to its assigned community. We here propose a fast variant of LPA (FLPA) that is based on processing a queue of nodes whose neighbourhood recently changed. We test FLPA exhaustively on benchmark networks and empirical networks, finding that it can run up to 700 times faster than LPA. In partitions found by FLPA, we prove that each node is again guaranteed to have most links to its assigned community. Our results show that FLPA is generally preferable to LPA.

Survey on graph embeddings and their applications to machine learning problems on graphs.

Dealing with relational data always required significant computational resources, domain expertise and task-dependent feature engineering to incorporate structural information into a predictive model. Nowadays, a family of automated graph feature engineering techniques has been proposed in different streams of literature. So-called graph embeddings provide a powerful tool to construct vectorized feature spaces for graphs and their components, such as nodes, edges and subgraphs under preserving inner graph properties. Using the constructed feature spaces, many machine learning problems on graphs can be solved via standard frameworks suitable for vectorized feature representation. Our survey aims to describe the core concepts of graph embeddings and provide several taxonomies for their description. First, we start with the methodological approach and extract three types of graph embedding models based on matrix factorization, random-walks and deep learning approaches. Next, we describe how different types of networks impact the ability of models to incorporate structural and attributed data into a unified embedding. Going further, we perform a thorough evaluation of graph embedding applications to machine learning problems on graphs, among which are node classification, link prediction, clustering, visualization, compression, and a family of the whole graph embedding algorithms suitable for graph classification, similarity and alignment problems. Finally, we overview the existing applications of graph embeddings to computer science domains, formulate open problems and provide experiment results, explaining how different networks properties result in graph embeddings quality in the four classic machine learning problems on graphs, such as node classification, link prediction, clustering and graph visualization. As a result, our survey covers a new rapidly growing field of network feature engineering, presents an in-depth analysis of models based on network types, and overviews a wide range of applications to machine learning problems on graphs.

Correction to 'Intermediacy of publications'.

[This corrects the article DOI: 10.1098/rsos.190207.].

Intermediacy of publications.

Citation networks of scientific publications offer fundamental insights into the structure and development of scientific knowledge. We propose a new measure, called intermediacy, for tracing the historical development of scientific knowledge. Given two publications, an older and a more recent one, intermediacy identifies publications that seem to play a major role in the historical development from the older to the more recent publication. The identified publications are important in connecting the older and the more recent publication in the citation network. After providing a formal definition of intermediacy, we study its mathematical properties. We then present two empirical case studies, one tracing historical developments at the interface between the community detection literature and the scientometric literature and one examining the development of the literature on peer review. We show both conceptually and empirically how intermediacy differs from main path analysis, which is the most popular approach for tracing historical developments in citation networks. Main path analysis tends to favour longer paths over shorter ones, whereas intermediacy has the opposite tendency. Compared to the main path analysis, we conclude that intermediacy offers a more principled approach for tracing the historical development of scientific knowledge.

War pact model of shrinking networks.

Many real systems can be described by a set of interacting entities forming a complex network. To some surprise, these have been shown to share a number of structural properties regardless of their type or origin. It is thus of vital importance to design simple and intuitive models that can explain their intrinsic structure and dynamics. These can, for instance, be used to study networks analytically or to construct networks not observed in real life. Most models proposed in the literature are of two types. A model can be either static, where edges are added between a fixed set of nodes according to some predefined rule, or evolving, where the number of nodes or edges increases over time. However, some real networks do not grow but rather shrink, meaning that the number of nodes or edges decreases over time. We here propose a simple model of shrinking networks called the war pact model. We show that networks generated in such a way exhibit common structural properties of real networks. Furthermore, compared to classical models, these resemble international trade, correlates of war, Bitcoin transactions and other networks more closely. Network shrinking may therefore represent a reasonable explanation of the evolution of some networks and greater emphasis should be put on such models in the future.

Convex skeletons of complex networks.

A convex network can be defined as a network such that every connected induced subgraph includes all the shortest paths between its nodes. A fully convex network would therefore be a collection of cliques stitched together in a tree. In this paper, we study the largest high-convexity part of empirical networks obtained by removing the least number of edges, which we call a convex skeleton. A convex skeleton is a generalization of a network spanning tree in which each edge can be replaced by a clique of arbitrary size. We present different approaches for extracting convex skeletons and apply them to social collaboration and protein interactions networks, autonomous systems graphs and food webs. We show that the extracted convex skeletons retain the degree distribution, clustering, connectivity, distances, node position and also community structure, while making the shortest paths between the nodes largely unique. Moreover, in the Slovenian computer scientists coauthorship network, a convex skeleton retains the strongest ties between the authors, differently from a spanning tree or high-betweenness backbone and high-salience skeleton. A convex skeleton thus represents a simple definition of a network backbone with applications in coauthorship and other social collaboration networks.

Clustering Scientific Publications Based on Citation Relations: A Systematic Comparison of Different Methods.

Clustering methods are applied regularly in the bibliometric literature to identify research areas or scientific fields. These methods are for instance used to group publications into clusters based on their relations in a citation network. In the network science literature, many clustering methods, often referred to as graph partitioning or community detection techniques, have been developed. Focusing on the problem of clustering the publications in a citation network, we present a systematic comparison of the performance of a large number of these clustering methods. Using a number of different citation networks, some of them relatively small and others very large, we extensively study the statistical properties of the results provided by different methods. In addition, we also carry out an expert-based assessment of the results produced by different methods. The expert-based assessment focuses on publications in the field of scientometrics. Our findings seem to indicate that there is a trade-off between different properties that may be considered desirable for a good clustering of publications. Overall, map equation methods appear to perform best in our analysis, suggesting that these methods deserve more attention from the bibliometric community.

Quantifying the consistency of scientific databases.

Science is a social process with far-reaching impact on our modern society. In recent years, for the first time we are able to scientifically study the science itself. This is enabled by massive amounts of data on scientific publications that is increasingly becoming available. The data is contained in several databases such as Web of Science or PubMed, maintained by various public and private entities. Unfortunately, these databases are not always consistent, which considerably hinders this study. Relying on the powerful framework of complex networks, we conduct a systematic analysis of the consistency among six major scientific databases. We found that identifying a single "best" database is far from easy. Nevertheless, our results indicate appreciable differences in mutual consistency of different databases, which we interpret as recipes for future bibliometric studies.

Network-based statistical comparison of citation topology of bibliographic databases.

Modern bibliographic databases provide the basis for scientific research and its evaluation. While their content and structure differ substantially, there exist only informal notions on their reliability. Here we compare the topological consistency of citation networks extracted from six popular bibliographic databases including Web of Science, CiteSeer and arXiv.org. The networks are assessed through a rich set of local and global graph statistics. We first reveal statistically significant inconsistencies between some of the databases with respect to individual statistics. For example, the introduced field bow-tie decomposition of DBLP Computer Science Bibliography substantially differs from the rest due to the coverage of the database, while the citation information within arXiv.org is the most exhaustive. Finally, we compare the databases over multiple graph statistics using the critical difference diagram. The citation topology of DBLP Computer Science Bibliography is the least consistent with the rest, while, not surprisingly, Web of Science is significantly more reliable from the perspective of consistency. This work can serve either as a reference for scholars in bibliometrics and scientometrics or a scientific evaluation guideline for governments and research agencies.

SkipCor: skip-mention coreference resolution using linear-chain conditional random fields.

Coreference resolution tries to identify all expressions (called mentions) in observed text that refer to the same entity. Beside entity extraction and relation extraction, it represents one of the three complementary tasks in Information Extraction. In this paper we describe a novel coreference resolution system SkipCor that reformulates the problem as a sequence labeling task. None of the existing supervised, unsupervised, pairwise or sequence-based models are similar to our approach, which only uses linear-chain conditional random fields and supports high scalability with fast model training and inference, and a straightforward parallelization. We evaluate the proposed system against the ACE 2004, CoNLL 2012 and SemEval 2010 benchmark datasets. SkipCor clearly outperforms two baseline systems that detect coreferentiality using the same features as SkipCor. The obtained results are at least comparable to the current state-of-the-art in coreference resolution.

Unfolding communities in large complex networks: combining defensive and offensive label propagation for core extraction.

Label propagation has proven to be a fast method for detecting communities in large complex networks. Recent developments have also improved the accuracy of the approach; however, a general algorithm is still an open issue. We present an advanced label propagation algorithm that combines two unique strategies of community formation, namely, defensive preservation and offensive expansion of communities. The two strategies are combined in a hierarchical manner to recursively extract the core of the network and to identify whisker communities. The algorithm was evaluated on two classes of benchmark networks with planted partition and on 23 real-world networks ranging from networks with tens of nodes to networks with several tens of millions of edges. It is shown to be comparable to the current state-of-the-art community detection algorithms and superior to all previous label propagation algorithms, with comparable time complexity. In particular, analysis on real-world networks has proven that the algorithm has almost linear complexity, O(m¹·¹9), and scales even better than the basic label propagation algorithm (m is the number of edges in the network).