SOCIAL media have become pervasive and ubiquitous and represent a source of valuable information. The literature on social media makes a distinction between influencers and influence. The former are social media users with a broad audience. For example, influencers can have a high number of followers on Twitter, or a multitude of friends on Facebook, or a broad array of connections on LinkedIn. The term influence is instead used to refer to the social impact of the content shared by social media users. The majority of these studies has focused on the role of influencers. Our claim is that while the information shared by influencers has a broader reach, the content of messages plays a critical role and can be a determinant of the social influence of the message irrespective of the centrality of the message’s author. This thesis starts from the observation that social networks of influence follow a power-law distribution function, with a few hub nodes and a long tail of peripheral nodes, consistent with the so-called small-world phenomenon. In social media, hub nodes represent social influencers, but influential content can be generated by peripheral nodes and spread along possibly multi-hop paths originated in peripheral network layers. This thesis provides a conceptual framework and related software tool to assess influence and identification of influencers. The assessment of influence and influencers is performed in two steps. First, an empirical analysis is conducted in order to verify the assumption that content can have an impact on influence. We propose a visual approach to the graphical representation and exploration of peripheral layers and clusters by exploiting the theory of k-shell decomposition analysis and power-law based modified force-directed method to clearly display local multi-layered neighborhood clusters around hub nodes. We put forward few hypotheses that tie specificity, frequency of tweets and frequency of retweets and are tested on data samples of roughly one million tweets. Overall, results highlight the effectiveness of our approach, providing interesting visual insights on how unveiling the structure of the periphery of the network can visually show the potential of peripheral nodes in determining influence and content relationship. Secondly, this thesis aims to provide a novel visual framework to analyze, explore and interact with Twitter ‘Who Follows Who’ relationships, by visually browsing the friends’ network to identify the key influencers based upon the actual influence of the content they share. As part of this research, we have developed NavigTweet, a novel visualization tool for the influence-based exploration of Twitter network. The core concept of the proposed approach is to identify influencers by browsing through a user’s friends’ network. Then, a power-law based modified force-directed method is applied to clearly display the network graph in a multi-layered and multi-clustered way. To gather some insight into the user experience with the pilot release of NavigTweet, we have conducted a qualitative pilot user study. We report on the study and its results, with initial pilot release.
I SOCIAL MEDIA sono diventati uno strumento pervasivo e ampiamente diffuso, rappresentando cosi una fonte di informazioni preziose. La letteratura sui social media evidenzia una distinzione tra influencer e influence. I primi sono utenti dei social media con un vasto pubblico. Ad esempio, gli influencer possono avere un alto numero di follower su Twitter, un elevato numero di amici su Facebook, o una vasta gamma di connessioni su LinkedIn. Il termine influence viene invece usato per indicare l'impatto sociale dei contenuti condivisi dagli utenti dei social media. La maggior parte di questi studi si e concentrata sul ruolo degli influenzatori, le cui informazioni condivise hanno una portata molto ampia. La nostra tesi invece, si concentra sul contenuto dei messaggi, che gioca un ruolo critico e puo essere un fattore determinante dell'influenza sociale del messaggio indipendentemente dalla centralita dell'autore. Questa tesi parte dalla constatazione che le reti sociali di influenza seguono una funzione di distribuzione power-law, con pochi nodi hub e con una lunga coda di nodi periferici, coerenti con il cosiddetto fenomeno small-world. Nel contesto dei social media, i nodi hub rappresentano influenzatori sociali, tuttavia il contenuto influente puo essere generato da nodi periferici e diffondersi coi lungo possibili percorsi multi-hop nati in livelli della rete periferica. Questa tesi fornisce un quadro concettuale e un relativo strumento software al fine di valutare l'influenza e di identificare gli influenzatori. La valutazione di influenza e influenzatori viene eseguita in due fasi. In primo luogo, viene condotta un'analisi empirica per verificare l'ipotesi che il contenuto possa avere un impatto sull'influence. Proponiamo dunque un approccio visivo per la rappresentazione grafica e l'esplorazione di layer periferici e cluster, sfruttando la teoria dell'analisi k-shell decomposition, mentre per quanto riguarda la visualizzazione di local multi-layered neighbourhood cluster intorno ai hub-nodes viene applicato il metodo force-directed modificato e basato sulla distribuzione power-law. Vengono inoltre presentate alcune ipotesi che legano la specificita, la frequenza di tweet e la frequenza di retweet, testate su campioni di dati di circa un milione di tweet. Nel complesso, i risultati evidenziano l'efficacia del nostro approccio, fornendo interessanti spunti visivi su come comprendere la struttura della periferia della rete, mostrando il potenziale dei nodi periferici nel determinare l'inflence e il contenuto relazionale. In secondo luogo, questa tesi si propone di fornire un innovativo quadro visivo con lo scopo di analizzare, esplorare ed interagire con relazioni di Twitter di tipo ‘Who Follows Who’, navingando visivamente la rete di amici, per identificare gli influencer chiave basati sulla influence effettiva del contenuto che condividono. Come parte di questa ricerca, abbiamo sviluppato NavigTweet, uno nuovo strumento di visualizzazione per l'esplorazione dell'influence-based dei network di Twitter. Il concetto di base del metodo proposto e quello di identificare gli influencer navigando attraverso la rete di amici di un utente. Successivamete, viene applicato un metodo di force-directed modificato e basato sulla distribuzione power-law, con lo scopo di visualizzare in modo chiaro il grafico di rete tramite un approccio multi-layer e multi-cluster. A fine di ottenere conoscenza dall'esperienza degli utenti con il rilascio pilota NavigTweet, abbiamo condotto uno studio pilota qualitativo dell'utente. Diamo un report sullo studio e sui suoi risultati insieme al rilascio caso pilota iniziale.
A visual framework for the empirical analysis of social influencers and influence
HUSSAIN, AJAZ
Abstract
SOCIAL media have become pervasive and ubiquitous and represent a source of valuable information. The literature on social media makes a distinction between influencers and influence. The former are social media users with a broad audience. For example, influencers can have a high number of followers on Twitter, or a multitude of friends on Facebook, or a broad array of connections on LinkedIn. The term influence is instead used to refer to the social impact of the content shared by social media users. The majority of these studies has focused on the role of influencers. Our claim is that while the information shared by influencers has a broader reach, the content of messages plays a critical role and can be a determinant of the social influence of the message irrespective of the centrality of the message’s author. This thesis starts from the observation that social networks of influence follow a power-law distribution function, with a few hub nodes and a long tail of peripheral nodes, consistent with the so-called small-world phenomenon. In social media, hub nodes represent social influencers, but influential content can be generated by peripheral nodes and spread along possibly multi-hop paths originated in peripheral network layers. This thesis provides a conceptual framework and related software tool to assess influence and identification of influencers. The assessment of influence and influencers is performed in two steps. First, an empirical analysis is conducted in order to verify the assumption that content can have an impact on influence. We propose a visual approach to the graphical representation and exploration of peripheral layers and clusters by exploiting the theory of k-shell decomposition analysis and power-law based modified force-directed method to clearly display local multi-layered neighborhood clusters around hub nodes. We put forward few hypotheses that tie specificity, frequency of tweets and frequency of retweets and are tested on data samples of roughly one million tweets. Overall, results highlight the effectiveness of our approach, providing interesting visual insights on how unveiling the structure of the periphery of the network can visually show the potential of peripheral nodes in determining influence and content relationship. Secondly, this thesis aims to provide a novel visual framework to analyze, explore and interact with Twitter ‘Who Follows Who’ relationships, by visually browsing the friends’ network to identify the key influencers based upon the actual influence of the content they share. As part of this research, we have developed NavigTweet, a novel visualization tool for the influence-based exploration of Twitter network. The core concept of the proposed approach is to identify influencers by browsing through a user’s friends’ network. Then, a power-law based modified force-directed method is applied to clearly display the network graph in a multi-layered and multi-clustered way. To gather some insight into the user experience with the pilot release of NavigTweet, we have conducted a qualitative pilot user study. We report on the study and its results, with initial pilot release.File | Dimensione | Formato | |
---|---|---|---|
2015_11_PhD_Hussain.pdf
accessibile in internet per tutti
Descrizione: Full Thesis
Dimensione
19.5 MB
Formato
Adobe PDF
|
19.5 MB | Adobe PDF | Visualizza/Apri |
I documenti in POLITesi sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
https://hdl.handle.net/10589/113642