 Research
 Open access
 Published:
Local 2connected bowtie structure of the Web and of social networks
Applied Network Science volume 9, Article number: 34 (2024)
Abstract
The explosive growth of the Web and of social networks motivates the need for analyzing the macroscopic structure of their underlying graphs. Although the characterization of the structure of a graph with respect to its pairwise connectivity has been known for over 15 years, just one subsequent study analyzed the world inside the giant strongly connected component, where it has been shown that the largest strongly connected component has its own microscopic bowtie structure defined with respect to pairwise 2connectivity among its vertices. In this paper, we introduce the local microscopic bowtie structure of the largest strongly connected component, demonstrating its selfsimilarity property. Our experiments, conducted on the several Web graphs and social networks demonstrate clear structural differences between considered Web and social networks.
Introduction
The explosive growth of the World Wide Web during the past two decades has created a vast infrastructure that serves daily online needs of people all over the world. In parallel with the growth of the World Wide Web, in the past decade several social networks emerged and were adopted quickly by a large portion of the world population. The most popular social networks nowadays consist of hundreds of millions up to a billion of active users and they naturally capture important social activities. Both the Web and social networks can be represented by directed graphs. In Web graphs, the vertices represent static HTML pages and the edges the hyperlinks among the pages. In social networks, the vertices correspond to users of a social network and the edges represent the whofollowswhom information. The analysis of the structure of the underlying graphs in both Web and social networks is undoubtedly important in many scenarios. For instance, the structure of the Web can be beneficial for improving the browsing experience (Carrière and Kazman 1997). Moreover, the knowledge of the macroscopic structure of the Web has been used for designing efficient algorithms for computing PageRank values (Kamvar et al. 2003).
The existing literature provides a good understanding of the connectivity structure in Web graphs and social networks. The bowtie structure, that was observed for the first time more than 15 years ago, has been applied on different graphs revealing the macroscopic structure of several graphs, and also helped to study the evolution of the Web and several social networks. All these studies observed a giant strongly connected component containing a substantial portion of the vertices, which is considered the core of the directed graph for both in Web graphs and social networks, and which is continuously growing as the graphs evolve. Although the characterization of the structure of a graph with respect to its pairwise connectivity has been known for over 15 years, just one subsequent study further analyzed the structure inside the giant strongly connected component of Web graphs and of social networks. In this scenario, it is necessary to analyze further the core of Web graphs and social networks in order to understand the structure of the core of the Web and of social networks. In this paper, we applied new theoretical notions and algorithms in order to conduct a novel analysis of the core of the Web and of social networks. We considered a natural extension of the methodology used in the pioneering study by Broder et al, by analyzing pairwise 2connectivity of Web and social networks on the local level. Our findings shed light on the structural properties of the core of the Web and of social networks.
Background and definitions
Let G = (V, E) be a directed graph (digraph), with m edges and n vertices. Digraph G is strongly connected if there is a directed path from each vertex to every other vertex. The strongly connected components of G are its maximal strongly connected subgraphs. Two vertices \(u, v \in V\) are strongly connected if they belong to the same strongly connected component of G. The size of a strongly connected component is given by its number of vertices. An edge (resp., a vertex) of G is a strong bridge (resp., a strong articulation point) if its removal increases the number of strongly connected components. Two vertices \(u, v \in V\) are said to be 2edgeconnected (resp., 2vertexconnected), and we denote this relation by \(u \longleftrightarrow _{2e} v\) (resp., \(u \longleftrightarrow _{2v} v\)), if there are two edgedisjoint (resp., two internally vertexdisjoint) directed paths from u to v and two edgedisjoint (resp., two internally vertexdisjoint) directed paths from v to u (note that a path from u to v and a path from v to u need not be edgedisjoint or vertexdisjoint). A 2edgeconnected component (resp., 2vertexconnected component) of a digraph G = (V, E) is defined as a maximal subset \(B \subseteq V\) such that \(v \longleftrightarrow _{2e} w\) (resp., \(v \longleftrightarrow _{2v} w\)) for all \(v, w \in B\)Georgiadis et al. (2018). A visualisation of mentioned definitions could be seen in Fig. 1.
Related works
Previous studies provided a good understanding of so called macroscopic connectivity structure of the underlying graphs for both Web and social networks. For the Web graph the first such understanding was given by Broder et al. (2000) more than two decades ago. In this study Broder et al. observed an existence of a macroscopic bowtie structure that was defined as follows. Macroscopic bowtie structure is formed around the largest strongly connected component (in short LSCC). The rest of the vertices were partitioned into six distinct subsets according to their relation to the LSCC. Precisely, the vertices that have a directed path to a vertex in the LSCC form the IN subset; the vertices that have a directed path from a vertex in LSCC form the OUT subset; the vertices that have a directed path from a vertex that belongs to the IN subset form a subset called INTENDRILS; the vertices that have a directed path to a vertex that belongs to the OUT subset form a subset called OUTTENDRILS; the vertices that have both a path from a vertex in the IN subset and a path to a vertex in the OUT subset form the subset called TUBES. The rest of the vertices which do not have a path neither from the LSCC or to the LSCC and also do not have a path neither from IN subset or to OUT subset form DISCONNECTED subset. See Fig. 2 for an illustration of the mentioned relations. At that time Broder et al. reported the existence of largest strongly connected component in the Web that contained about \(28\%\) of all the vertices. Although later studies showed that several structural properties are crawlersensitive (Serrano et al. 2007), such as the degree distribution or the sizes of the different components of the graph, they all agreed on the existence of a giant strongly connected component.
Subsequent studies proposed different structures of the Web graphs, such as the “daisy” structure (Donato et al. 2005) and the “teapot” structure (Zhu et al. 2008), both defined with respect to the largest strongly connected component. The “daisy” structure of the Web discovered by Donato et. al. revealed that the Web has a dense core or the largest strongly connected component and fragmented IN and OUT components hanging from the largest strongly connected component. In their study Zhu et. al. analyzed the Chinese Web on the different aggregation levels and reported that the Web on the page level has the “teapot” structure, where the largest strongly connected component has the biggest size, IN component has a medium size and OUT component has a small size. Their findings also revealed that with the increment of aggregation levels, the structure of the Chinese Web becomes increasingly close to the “daisy” structure.
In a much recent work Meusel et al. (2014) revisited aforementioned study of Broder et al. (2000) by collecting and analyzing a 2012 crawl of the Web containing 3.5 billion vertices and 128 billion edges. Their study revealed that the largest strongly connected component of the Web consisted of more than \(50\%\) of the total number of vertices. Their findings demonstrated that the Web became much more connected as it evolved, which was witnessed by the fact that the fraction of the vertices contained in the core of the Web (i.e., its LSCC) substantially increased.
Gabielkov et al. (2014) conducted analysis of the macroscopic structure of the underlying graph of Twitter social network, consisting of 505 millions of vertices and 23 billions of edges. They reported existence of the bowtie structure of the underlying social network formed around the largest strongly connected component containing as well about half of the total number of vertices. In their study Gabielkov et al. (2014) characterized the largest strongly connected component as “the core of the regular Twitter activity”.
All aforementioned studies gave us an understanding of the macroscopic or global connectivity structure of the Web and of social networks, which is centered around the core of the networks  their largest strongly connected component. Fujita et al. (2019) proposed a local bowtie structure of the Web, where they discovered that each community in the Web graph has its own bowtie structure, which is formed in the same way as a global bowtie structure, but not just around the largest strongly connected component, but around each strongly connected component. Their findings demonstrated a selfsimilarity property of the Web, i.e that the structure of the Web as a whole repeats itself in its communities (see Fig. 3).
The first study that analyzed the core of the Web and of social networks (i.e. their LSCC) was made by Italiano et al. (2017), where it was shown that the largest strongly connected component has its own microscopic \(bowtie\) structure. In their study they extended the notions of pairwise strong connectivity used in the pioneering study by Broder et al. by analyzing the pairwise \(2connectivity\) inside the LSCC. Their experiments were conducted on eight Web graphs and seven social networks and revealed an inner bowtie structure of the LSCC that was remarkably consistent across different Web graphs and across different social networks. The microscopic bowtie structure with respect to the pairwise 2edge connectivity between the nodes of the LSCC was defined as follows. The core of the structure was defined its largest 2edgeconnected block (in short, L2ECB), where all the vertices have two edgedisjoint paths to each other. The sets 2EIN and 2EOUT (in short, IN and OUT, respectively) were defined as the sets which contain the vertices that have two edgedisjoint paths to L2ECB, and from L2ECB, respectively. The vertices that do not have two edgedisjoint paths to L2ECB and also do not have two edgedisjoint paths from L2ECB, form the set 2EOTHER (in short, OTHER), which also could be considered as an “analogy” of the DISCONNECTED set of the macroscopic bowtie structure. The sets L2ECB, IN, OUT and OTHER were considered to be the components of proposed microscopic bowtie structure. The visualization of microscopic bowtie structure of the LSCC could be seen in Fig. 4. Italiano et al. (2017) demonstrated that the L2ECB component occupies in total about 40% of vertices in Web graphs and around 60% in social networks, while the component 2EIN occupies about 43% and 8%, and the component 2EOUT occupies about 6% and 10% of total amount of vertices, respectively. Their findings showed that the structure of the largest strongly connected component is remarkably consistent across different Web graphs and across social networks. Although the analysis in this study demonstrated noticeable differences in the structures of the LSCC of Web and of social networks, it undoubtedly needed further investigation of their structural properties.
In this paper we make a step further in the structural analysis made by Italiano et al. (2017) and extend the notion introduced by Fujita et al. of local bowtie structure with respect to strong connectivity to the notion of local bowtie structure with respect to pairwise \(2connectivity\) among the vertices inside the giant strongly connected component. We demonstrate existence of the local microscopic \(bowtie\) structure of the largest strongly connected component, showing in this way \(selfsimilarity\) property of the largest strongly connected component itself. Next, we analyze structural differences of the Web and of social networks by considering the distributions of average PageRank, Outdegree and Indegree values over the components of their local microscopic bowtie structures. Our findings demonstrate differences in the structure of the LSCC for Web and for social networks.
The structure of this paper is organized as follows. In Sect. 2 we describe the datasets on which we performed our analysis. Section 3 contains methodology behind it. In Sect. 4 we present distributions of centrality measures among components of the microscopic local bowtie structure for considered Web and social networks. We also present distributions of 2connectivity characteristics among the components of the local structure. Based on local microscopic \(bowtie\) structure we demonstrate selfsimilarity property of the largest strongly connected component of considered Web and of social networks and show structural differences between two types of networks.
Data
We conducted our analysis on the collection of 3 Web graphs and 3 social networks. This amount of networks were processed due to computational challenges to process bigger graphs. Among Web graphs we considered the following datasets. Stanford and Berkstan are the Web graphs of the Stanford and UCBerkley University domains. The Google graph represents the Web pages under the google.com domain, as they were released as part of the Google programming contest in 2002. The datasets Stanford, Berkstan, and Google were taken from the SNAP data repository (Leskovec and Sosič 2014). We also considered the following social networks. Epinions (Richardson et al. 2003) represents a user oriented product review website. Academia (Fire et al. 2013) is the underlying graph of the network of researchers registered on academia.edu. \(Google+\) (Fire et al. 2013) contains a small subgraph of Google’s famous social network. Table 1 summarizes the characteristics of the datasets that we considered with their largest strongly connected components.
Methodology
Our methodology behind our microscopic local structural network analysis is the following. We consider microscopic bowtie structure introduced in Italiano et al. (2017) and extend it to the local level, that is, we consider bowtie structures, each of which is formed not around the largest 2edgeconnected block (L2ECB), but around of the each of 2edgeconnected blocks (2ECBs) of the largest strongly connected component of the digraph G. In other words local microscopic bowtie structure consists of 2edgeconnected blocks, IN, OUT and OTHER components inside the largest strongly connected component. Note, that IN, OUT and OTHER components are defined in the same way as for global microscopic bowtie structure. The visualization of proposed structure could be seen in Fig. 5. It is worth mentioning that methodology for extracting local microscopic bowtie structure could be applied to any strongly connected component of any directed graph G.
Microscopic local bowtie structure
In this section we present local microscopic bowtie structure of the largest strongly connected component for considered Web and social networks. We also analyze components of the microscopic local bowtie structure for considered Web and social networks. We compute several centrality measures for 2ECB, OUT, IN and OTHER components of local structure and show how they differ for Web and social networks. More specifically, we compute average Pagerank, average Outdegree and average Indegree for 2ECB, OUT, IN and OTHER components of the local structure. We also compute the portion of the vertices of each set of OUT, IN and OTHER that are not 2edgeconnected with any other vertex, i.e. they form singleton 2edgeconnected blocks, as well as the portion of the pairs that are 2edgeconnected.
In the following four subsections we analyze each set of the 2ECB, OUT, IN and OTHER components of local bowtie structure for the Web and social networks that we considered. For each set of the 2ECB, OUT, IN and OTHER components of local bowties we present distributions of the vertices, edges, average PageRank, Outdegree and Indegree centralities, as well as distributions of \(2edgeconnected\) pairs and singleton \(2edgeconnected\) blocks among all of the components of the local structure. In this way we can see the differences in the distributions of the mentioned metrics among the components of the local structures of the Web and of social networks. We ask the interested reader to proceed with the following subsections.
Analysis of 2edgeconnected blocks of the local bowtie structure
In the following in order to show differences in the structure of the Web and of social networks we placed the histograms for the Web graphs on a right side of the figures and the histograms for the social networks on the left side of the figures. In Figs. 6 and 7 we can see the distributions of the vertices and edges among all of the 2ECBs of local bowtie structures. We can observe that for both Web and social networks almost 100% the 2ECBs occupy considerably small amount of both vertices and edges in our structures. In these plots there is no difference in the core structure between the Web and the social networks.
In Fig. 8 we can see the distributions of the average PageRank values among 2ECBs of the local bowtie structures for considered Web and social networks. As we can see the distributions have clear and consistent patterns for both Web and social networks, which differ significantly between them. More precisely, for more than 50% of the local bowties the average PageRank value is close to 0 and only a few 2ECBs have higher values of the average PageRank. The distribution of the average PageRank values among 2ECBs of the local bowtie structures for considered social networks is much different. More specifically, only a few blocks have the maximum average PageRank value, and the rest of the blocks are distributed in decreasing or increasing order.
In Figs. 9 and 10 we can see the distributions of the average Outdegree and Indegree values among 2ECBs of the local bowtie structures for our Web and social networks. As we can observe the average Oudegree has a decreasing order for the Web and for social networks. The pick of the average Outdegree for the Web networks lies between 5 and 10 values, while for the social networks, it lies between 2 and 4 values. Similar observation could be made for the average Indegree of the Web and of social networks.
Analysis of OUT components of the local bowtie structure
As we can see in Fig. 11 all our Web graphs have very similar distribution of the vertices among OUT components of our local structures. The picture is slightly different for the social networks, where for the Epinions social network the majority of the OUT components occupy about one third of the vertices in the largest strongly connected component of the network. The similar observation could be made for the edges among OUT components, shown in Figure 12.
In Fig. 13 we show the distribution of the PageRank values among OUT components of the networks that we considered. As we can see the distributions are quite uniform for all our Web graphs. The distribution is more skewed comparing to the distribution of the average PageRank for the 2ECBs. The distribution is different for our social networks, where we can observe that for the less amount of OUT components average PageRank values are close to 0.
In Fig. 14 we present the distribution of Outdegree values for the Web and social networks that we considered. As we can see the Web graphs again have a quite uniform distribution among our components. For the social networks we can observe different pattern of distribution, which remains consistent among our networks. Similar observation could be actually made for the distribution of Indegree values for the Web and social networks that we considered (Fig. 15).
In Fig. 16 we show the distribution of \(2edgeconnected\) pairs of our Web and social networks. As we can see for the Web graphs about 30% of 2edgeconnected pairs have the value close to the 0 and the rest 70% have in average the value of 2e\({10}\). Different observation could be made for our social networks. As we can see about 80% of \(2edgeconnected\) pairs have the value close to 0 for academia and \(Google+\) social networks, while for Epinions social network this percentage drops to 45%.
Similar differences between distributions of the singletons appear in Fig. 17.
Analysis of IN components of the local bowtie structure
In Fig. 18 we show distribution of the vertices among IN components in our local bowtie structures. As we can see for our considered Web graphs about 80% of our IN components occupy very low amount of vertices. For social networks corresponding value is 60, 30 and 50% for Epinions, Academia and \(Google+\) networks, respectively. Similar picture could be observed for the distributions of edges in our IN components as shown in Fig. 19.
In Fig. 20 we can observe distributions of the average PageRank values among IN components of our local bowtie structures. As it could be seen in Fig. 20 both our Web and social networks have quite distinguishable distributions.
In Fig. 21 we can see distributions of average Outdegree values for our Web and social networks. We can observe a clear pattern for the social networks. About 30% of IN components have very low average outdegree. For the social networks the pattern of distribution is quite different, about 50% of all IN components have outdegree values equal to 15 on average. Similar distinguished distributions could be observed for the Indegree average values among IN components, as shown in Fig. 22.
In Fig. 23 we can observe the distribution of the \(2edgeconnected\) pairs among IN components of our local bowtie structures. As we can see for the Web graphs about 80% of IN components have the amount of \(2edgeconnected\) pairs close to zero, while for the social networks corresponding value represented by about 50% of the components on average. Similar observation could be made for the distribution of the singletons among the same components. As we could observe also with the previous metrics we can see consistent pictures of the distributions among Web and social networks (Fig. 24).
Analysis of OTHER components of the local bowtie structure
In Figs. 25 and 26 we show the distributions of the vertices and edges over OTHER components for our Web and social networks. As in the cases of previous components we see clear difference between the same distributions for Web and for social networks. At the same time we could observe a surprisingly consistent picture among the distributions of the same type of the network.
In Fig. 27 we can see the distributions of the average Pagerank values among OTHER components in our local bowtie structures. As we can observe about 70% of the components have the value of average Pagerank equal to \(1\textrm{e}{6}\). Different picture could be seen for the social networks where the majority of the components have \(0.5\textrm{e}{6}\), \(4\textrm{e}{6}\) and \(9\textrm{e}{6}\) average Pagerank values.
In Fig. 28 we show the distributions of the average Outdegree values among OTHER components of our structures. We can observe very distinguishable distributions for the Web and for social networks that we considered. As we can see about 70% of our components have the average values 8.5, 5.5 and 11.5 for Stanford, Google and Berkstan Web graphs, respectively. That is different for the social networks that we considered, where about 50% of the components have the average Outdegree values have the value of 1.5 on average.
As we can see in Fig. 29 about 70% of our components have the average Indegree values about 3 for social networks. For Web graphs the corresponding value drops to 50–60% of the amount of all components.
As we can see in Figs. 30 and 31 the distributions of \(2edgeconnected\) pairs and singletons among OTHER components are quite similar for considered Web and social networks.
Conclusion
In this paper we extended the notion of local bowtie structure with respect to strong connectivity to the local bowtie structure with respect to pairwise 2connectivity among the vertices inside the largest strongly connected component. We have conducted our experiments on several Web and social networks and unveiled local microscopic \(bowtie\) structure inside the largest strongly connected component with the microscopic bowtie formed around each 2connected block. Revealed local microscopic bowtie structure demonstrates \(selfsimilarity\) property of the largest strongly connected component. In order to investigate structural differences between Web and social networks that we considered, we analyzed distributions of the vertices, edges, average PageRank, Outdegree, and Indegree centralities, as well as \(2edgeconnected\) pairs and singletons among all the components of their microscopic local bowtie structures. Our results show how average PageRank, Outdegree, and Indegree centralities change in the structure of the Web and of social networks and demonstrate the dependency of the centrality measures on the 2connected structure of the Web and of social networks. Our results also show the difference in the structures of Web and of social networks, proving the structure being one of the fundamental characteristics of the network.
Our study is the first to apply the notions of 2connectivity to analyze microscopic structure of Web graphs and of social networks on the local level. We believe that carrying out this analysis on more types of networks such as biological networks will shed more light on their structural properties and will help to see structural differences between different types of networks.
Finally, in the global bowtie structure with respect to the strongly connected components of a digraph, there exist three additional sets of vertices referred as OUTTENDRILS, INTENDRILS, and TUBES. These are the vertices that have a path to a vertex in OUT, a path from a vertex in IN, and both a path from a vertex in OUT and to a vertex in IN, respectively. Although these sets can be naturally extended to fit the structure of 2connectivity, it is not yet known whether these sets can be computed efficiently (i.e., in linear time), which is a crucial bottleneck when dealing with large scale graphs. We leave as a last open problem whether these sets can be computed in linear time.
Analysis conducted in our study is valid for any strongly connected component of any directed graph.
Availability of data and materials
The data that support the findings will be available in repository ’StructuralNetworkAnalysis’ at https://github.com/EugeniyaP/StructuralNetworkAnalysis.git from the date of publication.
Abbreviations
 G :

Directed graph
 V :

Set of the vertices of a directed graph G
 E :

Set of the edges of a directed graph G
 SCC :

Strongly connected component of a directed graph G
 2ECB(G):

Twoedgeconnected blocks of a directed graph G
 2VCB(G):

Twovertexconnected blocks of a directed graph G
 LSCC :

Largest strongly connected component of a directed graph G
 OUT :

OUT component of a macroscopic bowtie structure of a directed graph G
 IN :

IN component of a macroscopic bowtie structure of a directed graph G
 \(IN{}TENDRILS\) :

INTENDRILS component of a macroscopic bowtie structure of a directed graph G
 \(OUT{}TENDRILS\) :

OUT{}TENDRILS component of a macroscopic bowtie structure of a directed graph G
 TUBES :

TUBES component of a macroscopic bowtie structure of a directed graph G
 DISCONNECTED :

DISCONNECTED component of a macroscopic bowtie structure of a directed graph G
 DISC.:

DISCONNECTED component of a macroscopic bowtie structure of a directed graph G
 L2ECB :

Largest twoedgeconnected block of the largest strongly connected component of a directed graph G
 2ECBs :

Twoedgeconnected blocks of the largest strongly connected component of a directed graph G
 2ECB :

Twoedgeconnected block of the largest strongly connected component of a directed graph G
 \(2E{}IN\) :

Component IN of a microscopic bowtie structure of the largest strongly connected component of a directed graph G
 \(2E{}OUT\) :

Component OUT of a microscopic bowtie structure of the largest strongly connected component of a directed graph G
 \(2E{}OTHER\) :

Component OTHER of a microscopic bowtie structure of the largest strongly connected component of a directed graph G
 \(2E{}DISC.\) :

Component OTHER of a microscopic bowtie structure of the largest strongly connected component of a directed graph G
 IN :

Component IN of a microscopic bowtie structure of the largest strongly connected component of a directed graph G
 OUT :

Component OUT of a microscopic bowtie structure of the largest strongly connected component of a directed graph G
 OTHER :

Component OTHER of a microscopic bowtie structure of the largest strongly connected component of a directed graph G
 SNAP :

Stanford Network Analysis Project
References
Broder A, Kumar R, Maghoul F, Raghavan P, Rajagopalan S, Stata R, Tomkins A, Wiener J (2000) Graph structure in the web. Comput Netw 33(1):309–320
Carrière SJ, Kazman R (1997) Webquery: searching and visualizing the web through connectivity. Comput Netw ISDN Syst 29(8):1257–1267
Donato D, Leonardi S, Millozzi S, Tsaparas P (2005) Mining the inner structure of the web graph. In: WebDB, pp 145–150
Fire M, TenenboimChekina L, Puzis R, Lesser O, Rokach L, Elovici Y (2013) Computationally efficient link prediction in a variety of social networks. ACM Trans Intell Syst Technol 5(1):10
Fujita Y, Kichikawa Y, Fujiwara Y, Souma W, Iyetomi H (2019) Local bowtie structure of the web. Appl Netw Sci 4(1):15
Gabielkov M, Rao A, Legout A (2014) Studying social networks at scale: macroscopic anatomy of the twitter social graph. In: ACM SIGMETRICS performance evaluation review, vol 42, pp 277–288. ACM
Georgiadis L, Italiano GF, Karanasiou A, Parotsidis N, Paudel N (2018) Computing 2connected components and maximal 2connected subgraphs in directed graphs: an experimental study. In: 2018 Proceedings of the twentieth workshop on algorithm engineering and experiments (ALENEX), pp 169–183. SIAM
Italiano GF, Parotsidis N, Perekhodko E (2017) What’s inside a bowtie: analyzing the core of the web and of social networks. In: Proceedings of the 2017 international conference on information system and data mining, pp 39–43
Kamvar S, Haveliwala T, Manning C, Golub G (2003) Exploiting the block structure of the web for computing PageRank. Stanford University Technical Report
Leskovec J, Sosič R (2014) SNAP: a general purpose network analysis and graph mining library in C++. http://snap.stanford.edu/snap
Meusel R, Vigna S, Lehmberg O, Bizer C (2014) Graph structure in the web—revisited: a trick of the heavy tail. In: Proceedings of the companion publication of the 23rd international conference on World Wide Web Companion, pp 427–432. International World Wide Web Conferences Steering Committee
Richardson M, Agrawal R, Domingos P (2003) Trust management for the semantic web. In: The Semantic WebISWC 2003, pp 351–368. Springer
Serrano M, Maguitman A, Boguñá M, Fortunato S, Vespignani A (2007) Decoding the structure of the www: a comparative analysis of web crawls. ACM Trans Web (TWEB) 1(2):10
Zhu JJ, Meng T, Xie Z, Li G, Li X (2008) A teapot graph and its hierarchical structure of the Chinese web. In: Proceedings of the 17th international conference on world wide web, pp 1133–1134. ACM
Acknowledgements
I dedicate this work to God, Saint Nicholas, Saint Sergius, Saint Luka Voyno Yasenetsky, Saint Eugenia, Saint Ksenia and my parents.
Funding
Open access funding provided by Northeastern University Library This research received no specific grant from any funding agency, commercial or notforprofit sectors.
Author information
Authors and Affiliations
Contributions
Major contributor—EP.
Corresponding author
Ethics declarations
Competing interests
The authors declare that they have no Conflict of interest
Additional information
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Rights and permissions
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.
About this article
Cite this article
Perekhodko, E. Local 2connected bowtie structure of the Web and of social networks. Appl Netw Sci 9, 34 (2024). https://doi.org/10.1007/s4110902400638y
Received:
Accepted:
Published:
DOI: https://doi.org/10.1007/s4110902400638y