Emotional Text Mining: Customer profiling in brand management

The widespread use of the Internet and the constant increase in users of social media platforms has made a large amount of textual data available. This represents a valuable source of information about the changes in people’s opinions and feelings. This paper presents the application of Emotional Text Mining (ETM) in the field of brand management. ETM is an unsupervised procedure aiming to profile social media users. It is based on a bottom-up approach to classify unstructured data for the identification of social media users’ representations and sentiments about a topic. It is a fast and simple procedure to extract meaningful information from a large collection of texts. As customer profiling is relevant for brand management, we illustrate a business application of ETM on Twitter messages concerning a well-known sportswear brand in order to show the potential of this procedure, highlighting the characteristics of Twitter user communities in terms of product preferences, representations, and sentiments.     

Discovering socially similar users in social media datasets based on their socially important locations

Socially similar social media users can be defined as users whose frequently visited locations in their social media histories are similar. Discovering socially similar social media users is important for several applications, such as, community detection, friendship analysis, location recommendation, urban planning, and anomaly user and behavior detection. Discovering socially similar users is challenging due to dataset size and dimensions, spam behaviors of social media users, spatial and temporal aspects of social media datasets, and location sparseness in social media datasets. In the literature, several studies are conducted to discover similar social media users out of social media datasets using spatial and temporal information. However, most of these studies rely on trajectory pattern mining methods or take into account semantic information of social media datasets. Limited number of studies focus on discovering similar users based on their social media location histories. In this study, to discover socially similar users, frequently visited or socially important locations of social media users are taken into account instead of all locations that users visited. A new interest measure, which is based on Levenshtein distance, was proposed to quantify user similarity based on their socially important locations and two algorithms were developed using the proposed method and interest measure. The algorithms were experimentally evaluated on a real-life Twitter dataset. The results show that the proposed algorithms could successfully discover similar social media users based on their socially important locations.     

User community detection via embedding of social network structure and temporal content

Identifying and extracting user communities is an important step towards understanding social network dynamics from a macro perspective. For this reason, the work in this paper explores various aspects related to the identification of user communities. To date, user community detection methods employ either explicit links between users (link analysis), or users’ topics of interest in posted content (content analysis), or in tandem. Little work has considered temporal evolution when identifying user communities in a way to group together those users who share not only similar topical interests but also similar temporal behavior towards their topics of interest. In this paper, we identify user communities through multimodal feature learning (embeddings). Our core contributions can be enumerated as (a) we propose a new method for learning neural embeddings for users based on their temporal content similarity; (b) we learn user embeddings based on their social network connections (links) through neural graph embeddings; (c) we systematically interpolate temporal content-based embeddings and social link-based embeddings to capture both social network connections and temporal content evolution for representing users, and (d) we systematically evaluate the quality of each embedding type in isolation and also when interpolated together and demonstrate their performance on a Twitter dataset under two different application scenarios, namely news recommendation and user prediction. We find that (1) content-based methods produce higher quality communities compared to link-based methods; (2) methods that consider temporal evolution of content, our proposed method in particular, show better performance compared to their non-temporal counter-parts; (3) communities that are produced when time is explicitly incorporated in user vector representations have higher quality than the ones produced when time is incorporated into a generative process, and finally (4) while link-based methods are weaker than content-based methods, their interpolation with content-based methods leads to improved quality of the identified communities.     

Investigating the COVID-19 vaccine discussions on Twitter through a multilayer network-based approach

Modeling discussions on social networks is a challenging task, especially if we consider sensitive topics, such as politics or healthcare. However, the knowledge hidden in these debates helps to investigate trends and opinions and to identify the cohesion of users when they deal with a specific topic. To this end, we propose a general multilayer network approach to investigate discussions on a social network. In order to prove the validity of our model, we apply it on a Twitter dataset containing tweets concerning opinions on COVID-19 vaccines. We extract a set of relevant hashtags (i.e., gold-standard hashtags) for each line of thought (i.e., pro-vaxxer, neutral, and anti-vaxxer). Then, thanks to our multilayer network model, we figure out that the anti-vaxxers tend to have ego networks denser (+14.39%) and more cohesive (+64.2%) than the ones of pro-vaxxer, which leads to a higher number of interactions among anti-vaxxers than pro-vaxxers (+393.89%). Finally, we report a comparison between our approach and one based on single networks analysis. We prove the effectiveness of our model to extract influencers having ego networks with more nodes (+40.46%), edges (+39.36%), and interactions with their neighbors (+28.56%) with respect to the other approach. As a result, these influential users are much more important to analyze and can provide more valuable information.     

Neural opinion dynamics model for the prediction of user-level stance dynamics

Social media platforms allow users to express their opinions towards various topics online. Oftentimes, users’ opinions are not static, but might be changed over time due to the influences from their neighbors in social networks or updated based on arguments encountered that undermine their beliefs. In this paper, we propose to use a Recurrent Neural Network (RNN) to model each user’s posting behaviors on Twitter and incorporate their neighbors’ topic-associated context as attention signals using an attention mechanism for user-level stance prediction. Moreover, our proposed model operates in an online setting in that its parameters are continuously updated with the Twitter stream data and can be used to predict user’s topic-dependent stance. Detailed evaluation on two Twitter datasets, related to Brexit and US General Election, justifies the superior performance of our neural opinion dynamics model over both static and dynamic alternatives for user-level stance prediction.     

Retweets of officials’ alarming vs reassuring messages during the COVID-19 pandemic: Implications for crisis management

Coronavirus related discussions have spiraled at an exponential rate since its initial outbreak. By the end of May, more than 6 million people were diagnosed with this infection. Twitter witnessed an outpouring of anxious tweets through messages associated with the spread of the virus. Government and health officials replied to the troubling tweets, reassuring the public with regular alerts on the virus's progress and information to defend against the virus. We observe that social media users are worried about Covid 19-related crisis and we identify three separate conversations on virus contagion, prevention, and the economy. We analyze the tone of officials’ tweet text as alarming and reassuring and capture the response of Twitter users to official communications. Such studies can provide insights to health officials and government agencies for crisis management, specifically regarding communicating emergency information to the public via social media for establishing reassurance.     

Optimism and pessimism analysis using deep learning on COVID-19 related twitter conversations

This paper proposes a new deep learning approach to better understand how optimistic and pessimistic feelings are conveyed in Twitter conversations about COVID-19. A pre-trained transformer embedding is used to extract the semantic features and several network architectures are compared. Model performance is evaluated on two new, publicly available Twitter corpora of crisis-related posts. The best performing pessimism and optimism detection models are based on bidirectional long- and short-term memory networks.Experimental results on four periods of the COVID-19 pandemic show how the proposed approach can model optimism and pessimism in the context of a health crisis. There is a total of 150,503 tweets and 51,319 unique users. Conversations are characterised in terms of emotional signals and shifts to unravel empathy and support mechanisms. Conversations with stronger pessimistic signals denoted little emotional shift (i.e. 62.21% of these conversations experienced almost no change in emotion). In turn, only 10.42% of the conversations laying more on the optimistic side maintained the mood. User emotional volatility is further linked with social influence.     

Polarity Classification of Twitter Messages using Audio Processing

Polarity classification is one of the most fundamental problems in sentiment analysis. In this paper, we propose a novel method, Sound Cosine Similaritye Matching, for polarity classification of Twitter messages which incorporates features based on audio data rather than on grammar or other text properties, i.e., eliminates the dependency on external dictionaries. It is useful especially for correctly identifying misspelled or shortened words that are frequently encountered in text from online social media. Method performance is evaluated in two levels: i) capture rate of the misspelled and shortened words, ii) classification performance of the feature set. Our results show that classification accuracy is improved, compared to two other models in the literature, when the proposed features are used.     

The reflection of offline activities on users’ online social behavior: An observational study

The ever increasing presence of online social networks in users’ daily lives has led to the interplay between users’ online and offline activities. There have already been several works that have studied the impact of users’ online activities on their offline behavior, e.g., the impact of interaction with friends on an exercise social network on the number of daily steps. In this paper, we consider the inverse to what has already been studied and report on our extensive study that explores the potential causal effects of users’ offline activities on their online social behavior. The objective of our work is to understand whether the activities that users are involved with in their real daily life, which place them within or away from social situations, have any direct causal impact on their behavior in online social networks. Our work is motivated by the theory of normative social influence, which argues that individuals may show behaviors or express opinions that conform to those of the community for the sake of being accepted or from fear of rejection or isolation. We have collected data from two online social networks, namely Twitter and Foursquare, and systematically aligned user content on both social networks. On this basis, we have performed a natural experiment that took the form of an interrupted time series with a comparison group design to study whether users’ socially situated offline activities exhibited through their Foursquare check-ins impact their online behavior captured through the content they share on Twitter. Our main findings can be summarised as follows (1) a change in users’ offline behavior that affects the level of users’ exposure to social situations, e.g., starting to go to the gym or discontinuing frequenting bars, can have a causal impact on users’ online topical interests and sentiment; and (2) the causal relations between users’ socially situated offline activities and their online social behavior can be used to build effective predictive models of users’ online topical interests and sentiments.     

Voluntary sharing and mandatory provision: Private information disclosure on social networking sites

Private information disclosure on social networking sites (SNS) is one of the most important and active issues in the information management arena. The growing phenomenon of platforms requiring users to disclose personal information exposes the limitations of previous studies that only focus on users’ voluntary disclosure. In this study, we define two modes of users’ private information disclosure behavior: voluntary sharing and mandatory provision. Using the Communication Privacy Management theory, we built a framework to explain the impact of individual characteristics, context, motivation, and benefit–risk ratio on the user's willingness to disclose voluntarily or mandatorily. Our research shows that voluntary sharing is more likely to be driven by positive factors, such as perceived benefits, social network size, and personalization, while mandatory provision is affected by individual characteristics such as age, privacy policy, and perceived risks. One of our interesting findings is that perceived risk has less impact on voluntary sharing than previous studies suggested. When encouraging users to share information voluntarily, platforms do not need to pay as much attention to reducing perceived risk as in the mandatory providing mode, but should focus on improving perceived benefits. Being the first to classify and compare the private information disclosure modes of SNS users, our research enriches the existing literature and opens up new avenues for researchers and social networking platforms.     

Marketing analysis of wineries using social collective behavior from users’ temporal activity on Twitter

Marketing professionals face challenges of increasing complexity to adapt classic marketing strategies to the phenomenon of social networks. Companies are currently trying to take advantage of the useful collective knowledge available on social networks to support different types of marketing decisions. The appropriate analysis of this information can offer marketing professionals with important competitive advantages. This work proposes a new methodology to extract the social collective behavior of Twitter users concerning a group of brands based on the users’ temporal activity. Time series of mentions made by individual users to each company’s Twitter account are aggregated to obtain collective activity data for the companies, which is a consequence of both the company’s and other users’ actions. These data are processed using classical unsupervised machine learning techniques, such as temporal clustering and hidden Markov models, to extract collective temporal behavior patterns and models of the dynamics of customers over time for a single brand and groups of brands. The derived knowledge can be used for different tasks, such as identifying the impact of a marketing campaign on Twitter and comparatively assessing the social behaviors of different brands and groups of brands to assist in making marketing decisions. Our methodology is validated in a case study from the wine market. Twitter data were gathered from four regions of different countries around the world with important wineries (Italy: Veneto, Portugal: Porto and Douro Valley, Spain: La Rioja, and United States: Napa Valley), and comparative behavior analysis was carried out from the perspective of the use of Twitter as a communication channel for marketing campaigns.     

Mining user interests over active topics on social networks

Inferring users’ interests from their activities on social networks has been an emerging research topic in the recent years. Most existing approaches heavily rely on the explicit contributions (posts) of a user and overlook users’ implicit interests, i.e., those potential user interests that the user did not explicitly mention but might have interest in. Given a set of active topics present in a social network in a specified time interval, our goal is to build an interest profile for a user over these topics by considering both explicit and implicit interests of the user. The reason for this is that the interests of free-riders and cold start users who constitute a large majority of social network users, cannot be directly identified from their explicit contributions to the social network. Specifically, to infer users’ implicit interests, we propose a graph-based link prediction schema that operates over a representation model consisting of three types of information: user explicit contributions to topics, relationships between users, and the relatedness between topics. Through extensive experiments on different variants of our representation model and considering both homogeneous and heterogeneous link prediction, we investigate how topic relatedness and users’ homophily relation impact the quality of inferring users’ implicit interests. Comparison with state-of-the-art baselines on a real-world Twitter dataset demonstrates the effectiveness of our model in inferring users’ interests in terms of perplexity and in the context of retweet prediction application. Moreover, we further show that the impact of our work is especially meaningful when considered in case of free-riders and cold start users.     

How Many Bots in Russian Troll Tweets?

Increased usage of bots through the Internet in general, and social networks in particular, has many implications related to influencing public opinion. Mechanisms to distinguish humans from machines span a broad spectrum of applications and hence vary in their nature and complexity. Here we use several public Twitter datasets to build a model that can predict whether or not an account is a bot account based on features extracted at the tweet or the account level. We then apply the model to Twitter's Russian Troll Tweets dataset. At the account level, we evaluate features related to how often Twitter accounts are tweeting, as previous research has shown that bots are very active at some account levels and very low at others. At the tweet level, we noticed that bot accounts tend to sound more formal or structured, whereas real user accounts tend to be more informal in that they contain more slang, slurs, cursing, and the like. We also noted that bots can be created for a range of different goals (e.g., marketing and politics) and that their behaviors vary based on those distinct goals. Ultimately, for high bot-prediction accuracy, models should consider and distinguish among the different goals for which bots are created.     

Altmetrics – a complement to conventional metrics

Emerging metrics based on article-level does not exclude traditional metrics based on citations to the journal, but complements them. Both can be employed in conjunction to offer a richer picture of an article use from immediate to long terms. Article-level metrics (ALM) is the result of the aggregation of different data sources and the collection of content from multiple social network services. Sources used for the aggregation can be broken down into five categories: usage, captures, mentions, social media and citations. Data sources depend on the tool, but they include classic metrics indicators based on citations, academic social networks (Mendeley, CiteULike, Delicious) and social media (Facebook, Twitter, blogs, or Youtube, among others). Altmetrics is not synonymous with alternative metrics. Altmetrics are normally early available and allow to assess the social impact of scholarly outputs, almost at the real time. This paper overviews briefly the meaning of altmetrics and describes some of the existing tools used to apply this new metrics: Public Library of Science - Article-Level Metrics, Altmetric, Impactstory and Plum.Key words: altmetrics, article-level metrics, citation, social media     

A novel recommendation method based on social network using matrix factorization technique

The rapid development of information technology and the fast growth of Internet have facilitated an explosion of information which has accentuated the information overload problem. Recommender systems have emerged in response to this problem and helped users to find their interesting contents. With increasingly complicated social context, how to fulfill personalized needs better has become a new trend in personalized recommendation service studies. In order to alleviate the sparsity problem of recommender systems meanwhile increase their accuracy and diversity in complex contexts, we propose a novel recommendation method based on social network using matrix factorization technique. In this method, we cluster users and consider a variety of complex factors. The simulation results on two benchmark data sets and a real data set show that our method achieves superior performance to existing methods.     

How CEOs use Twitter: A comparative analysis of Global and Latin American companies

Social networks are becoming a key communication tool for organizations, but also for top managers like CEOs. Among the different available platforms, Twitter is one of the greatest and it is considered one of the most suitable to share information and engage in dialogue with stakeholders. In this way, this paper analyzes the presence of CEOs on the most active social network sites, and assess the activity and interaction of these top managers on Twitter. CEOs from Global and Latin American companies were selected, to compare their performance. The results of the study show that the presence of CEOs in social networks is very low, and the majority of those that are present on them are not adequately using their Twitter accounts. Although the general presence and performance on are low, LatAm CEOs have a better presence on social networks and they are more active on Twitter, but Global CEOs have better interaction results on their accounts. So, this area of strategic communication should be improved by communication practitioners, since the CEO communication is nowadays a key communication issue for any organization.     

How do scientific papers from different journal tiers gain attention on social media?

The emergence of social media has fundamentally changed the way of scholarly communication and allows for scientific research to be shared at an unprecedented speed and scale. While many studies have discussed what papers attract most online attention, how they prevail online is unclear. In this paper, we explore the diffusion patterns of ～170,000 papers with different journal tiers from 2012 to 2019 based on over 3 million Twitter mentions. We first categorize journals by the elite (the top of Q1) and non-elite (Q2～Q4) according to their journal impact factor quartiles, then use network analysis and time series analysis to characterize papers’ dynamic diffusion process, and finally discuss papers’ engaged users and disciplinary contexts. Results show that though elite journal papers spread significantly faster and more broadly than non-elite, some non-elite journal papers reached a sizable audience. For non-elite journal papers, a decent size of social media reach can be achieved through the aid of highly influential users or multiple waves of small spread that span a long period. As a result, popular non-elite journal papers tend to be more viral than the elite, focusing discussions around topics close to daily life. This study provides a new perspective to characterize the diffusion process of scientific papers and helps to further enhance the understanding of such a process.     

SimilCatch: Enhanced social spammers detection on Twitter using Markov Random Fields

The problem of social spam detection has been traditionally modeled as a supervised classification problem. Despite the initial success of this detection approach, later analysis of proposed systems and detection features has shown that, like email spam, the dynamic and adversarial nature of social spam makes the performance achieved by supervised systems hard to maintain. In this paper, we investigate the possibility of using the output of previously proposed supervised classification systems as a tool for spammers discovery. The hypothesis is that these systems are still highly capable of detecting spammers reliably even when their recall is far from perfect. We then propose to use the output of these classifiers as prior beliefs in a probabilistic graphical model framework. This framework allows beliefs to be propagated to similar social accounts. Basing similarity on a who-connects-to-whom network has been empirically critiqued in recent literature and we propose here an alternative definition based on a bipartite users-content interaction graph. For evaluation, we build a Markov Random Field on a graph of similar users and compute prior beliefs using a selection of state-of-the-art classifiers. We apply Loopy Belief Propagation to obtain posterior predictions on users. The proposed system is evaluated on a recent Twitter dataset that we collected and manually labeled. Classification results show a significant increase in recall and a maintained precision. This validates that formulating the detection problem with an undirected graphical model framework permits to restore the deteriorated performances of previously proposed statistical classifiers and to effectively mitigate the effect of spam evolution.     

Achieving and maintaining important roles in social media

A significant problem common for different domains of applications is an issue of obtaining and keeping an influential position in the respective social media society. In this paper a new approach is proposed based on the analysis of roles of users in groups identified within society. Three different dimensions of user behavior are considered as key elements of these roles: Their activity, influence and cooperativeness/competition. Taking into account measures describing these dimensions, a set of roles characterizing behaviors of users in groups is formulated. We propose an original set of roles with their justification in sociological models, develop an easy extendable model of a social system and conduct experiments to allow us to define patterns describing stability and variability of given roles as well as statistics of transitions in time between these considered roles. To define the roles, we took into account different features of user interactions, both quantitative and qualitative. We propose an integrated approach to the analysis of role changes in the context of group evolution. We consider behavior of users in groups with different sizes and differences between them. We also analyse the stability of individual roles players by users in groups and often occurring transitions between individual roles. The obtained results, interpreted also from the sociological point of view, allow the formulation of general recommendations on which behaviors of users could ensure achieving and maintaining influential roles in social media. The most frequent patterns of transitions between roles are identified and significant similarities between them for two considered blog portals are described. The approaches and methods of analysis presented in the paper may be applied to support decisions leading to obtaining and maintaining influential positions in social media, which may be useful for the promotion of goods and services, leading business or political campaigns.