Mining network-level properties of Twitter altmetrics data

Towards understanding the role of content-based and contextualized features in detecting abuse on Twitter

This paper presents a novel approach for detecting abuse on Twitter. Abusive posts have become a major problem for social media platforms like Twitter. It is important to identify abuse to mitigate its potential harm. Many researchers have proposed methods to detect abuse on Twitter. However, most of the existing approaches for detecting abuse look only at the content of the abusive tweet in isolation and do not consider its contextual information, particularly the tweets posted before the abusive tweet. In this paper, we propose a new method for detecting abuse that uses contextual information from the tweets that precede and follow the abusive tweet. We hypothesize that this contextual information can be used to better understand the intent of the abusive tweet and to identify abuse that content-based methods would otherwise miss. We performed extensive experiments to identify the best combination of features and machine learning algorithms to detect abuse on Twitter. We test eight different machine learning classifiers on content- and context-based features for the experiments. The proposed method is compared with existing abuse detection methods and achieves an absolute improvement of around 7%. The best results are obtained by combining the content and context-based features. The highest accuracy of the proposed method is 86%, whereas the existing methods used for comparison have highest accuracy of 79.2%.

User engagement with scholarly tweets of scientific papers: a large-scale and cross-disciplinary analysis

This study investigates the extent to which scholarly tweets of scientific papers are engaged with by Twitter users through four types of user engagement behaviors, i.e., liking, retweeting, quoting, and replying. Based on a sample consisting of 7 million scholarly tweets of Web of Science papers, our results show that likes is the most prevalent engagement metric, covering 44% of scholarly tweets, followed by retweets (36%), whereas quotes and replies are only present for 9% and 7% of all scholarly tweets, respectively. From a disciplinary point of view, scholarly tweets in the field of Social Sciences and Humanities are more likely to trigger user engagement over other subject fields. The presence of user engagement is more associated with other Twitter-based factors (e.g., number of mentioned users in tweets and number of followers of users) than with science-based factors (e.g., citations and Mendeley readers of tweeted papers). Building on these findings, this study sheds light on the possibility to apply user engagement metrics in measuring deeper levels of Twitter reception of scholarly information.

Exploring the User Interaction Network in an Anxiety Disorder Online Community: An Exponential Random Graph Model with Topical and Emotional Effects

The increasing number of people with anxiety disorders presents challenges when gathering health information. Users in anxiety disorder online communities (ADOCs) share and obtain a variety of health information, such as treatment experience, drug efficacy, and emotional support. This interaction alleviates the difficulties involved in obtaining health information. Users engage in community interaction via posts, comments, and replies, which promotes the development of an online community as well as the wellbeing of community users, and research concerning the formation mechanism of the user interaction network in ADOCs could be beneficial to users. Taking the Anxiety Disorder Post Bar as the research object, this study constructed an ADOC user interaction network based on users' posts, comments, and personal information data. With the help of exponential random graph models (ERGMs), we studied the effects of the network structure, user attributes, topics, and emotional intensity on user interaction networks. We found that there was significant reciprocity in the user interaction network in ADOCs. In terms of user attributes, gender homogeneity had no impact on the formation of the user interaction network. Experienced users in the community had obvious advantages, and experienced users could obtain replies more easily from other members. In terms of topics, pathology popularization showed obvious homogeneity, and symptoms of generalized anxiety disorder showed obvious heterogeneity. In terms of emotional intensity, users with polarized emotions were more likely to receive replies from users with positive emotions. The probability of interaction between two users with negative emotions was small, and users with opposite emotional polarity tended to interact, especially when the interaction was initiated by users with positive emotions.

Introducing the EMPIRE Index: A novel, value-based metric framework to measure the impact of medical publications

Article-level measures of publication impact (alternative metrics or altmetrics) can help authors and other stakeholders assess engagement with their research and the success of their communication efforts. The wide variety of altmetrics can make interpretation and comparative assessment difficult; available summary tools are either narrowly focused or do not reflect the differing values of metrics from a stakeholder perspective. We created the EMPIRE (EMpirical Publication Impact and Reach Evaluation) Index, a value-based, multi-component metric framework for medical publications. Metric weighting and grouping were informed by a statistical analysis of 2891 Phase III clinical trial publications and by a panel of stakeholders who provided value assessments. The EMPIRE Index comprises three component scores (social, scholarly, and societal impact), each incorporating related altmetrics indicating a different aspect of engagement with the publication. These are averaged to provide a total impact score and benchmarked so that a score of 100 equals the mean scores of Phase III clinical trial publications in the New England Journal of Medicine (NEJM) in 2016. Predictor metrics are defined to estimate likely long-term impact. The social impact component correlated strongly with the Altmetric Attention Score and the scholarly impact component correlated modestly with CiteScore, with the societal impact component providing unique insights. Analysis of fresh metrics collected 1 year after the initial dataset, including an independent sample, showed that scholarly and societal impact scores continued to increase, whereas social impact scores did not. Analysis of NEJM ‘notable articles’ showed that observational studies had the highest total impact and component scores, except for societal impact, for which surgical studies had the highest score. The EMPIRE Index provides a richer assessment of publication value than standalone traditional and alternative metrics and may enable medical researchers to assess the impact of publications easily and to understand what characterizes impactful research.

Are automated accounts driving scholarly communication on Twitter? a case study of dissemination of COVID-19 publications

From a network perspective, this study analyzes 659 users mentioning sampled COVID-19 articles 10 or more times on Twitter with a focus on their roles in facilitating the process of scholarly communication. Different from existing studies, we consider both the user types and the automation of accounts to profile influential users in the network of research dissemination. Our study found that similar to academic users, non-academic users can also be active players in communicating scientific publications. The results highlight the intensive interactions between human users and automated accounts, including bots and cyborgs, which accounted for 45% of connections among the top users. This study also demonstrates the important role of automated accounts in initiating and facilitating research dissemination. Specifically, (1) bot-assisted academic publishers showed the highest amplifier scores, which measures a user’s tendency of being the first to share information and reach out to others within their trusted networks, (2) 5.28% of the selected articles was first tweeted by automated research feeds, ranking the fourth among the 22 classified user groups, and (3) bot-assisted publishers and automated feeds of generic topics and news alerts were highly ranked in authority, a network measure to quantify the degree to which a user consumes important resources of relevant topics. In the conclusion section, we discuss future directions to improve the validity of Twitter metrics in assessing research impacts.

A death, infection, and recovery (DIR) model to forecast the COVID-19 spread

BACKGROUND

The SARS-Cov-2 virus (commonly known as COVID-19) has resulted in substantial casualties in many countries. The first case of COVID-19 was reported in China towards the end of 2019. Cases started to appear in several other countries (including Pakistan) by February 2020. To analyze the spreading pattern of the disease, several researchers used the Susceptible-Infectious-Recovered (SIR) model. However, the classical SIR model cannot predict the death rate.

OBJECTIVE

In this article, we present a Death-Infection-Recovery (DIR) model to forecast the virus spread over a window of one (minimum) to fourteen (maximum) days. Our model captures the dynamic behavior of the virus and can assist authorities in making decisions on non-pharmaceutical interventions (NPI), like travel restrictions, lockdowns, etc.

METHOD

The size of training dataset used was 134 days. The Auto Regressive Integrated Moving Average (ARIMA) model was implemented using XLSTAT (add-in for Microsoft Excel), whereas the SIR and the proposed DIR model was implemented using python programming language. We compared the performance of DIR model with the SIR model and the ARIMA model by computing the Percentage Error and Mean Absolute Percentage Error (MAPE).

RESULTS

Experimental results demonstrate that the maximum% error in predicting the number of deaths, infections, and recoveries for a period of fourteen days using the DIR model is only 2.33%, using ARIMA model is 10.03% and using SIR model is 53.07%.

CONCLUSION

This percentage of error obtained in forecasting using DIR model is significantly less than the% error of the compared models. Moreover, the MAPE of the DIR model is sufficiently below the two compared models that indicates its effectiveness.

A Novel Deep Neural Network-Based Approach to Measure Scholarly Research Dissemination Using Citations Network

We investigated the scientific research dissemination by analyzing the publications and citation data, implying that not all citations are significantly important. Therefore, as alluded to existing state-of-the-art models that employ feature-based techniques to measure the scholarly research dissemination between multiple entities, our model implements the convolutional neural network (CNN) with fastText-based pre-trained embedding vectors, utilizes only the citation context as its input to distinguish between important and non-important citations. Moreover, we speculate using focal-loss and class weight methods to address the inherited class imbalance problems in citation classification datasets. Using a dataset of 10 K annotated citation contexts, we achieved an accuracy of 90.7% along with a 90.6% f1-score, in the case of binary classification. Finally, we present a case study to measure the comprehensiveness of our deployed model on a dataset of 3100 K citations taken from the ACL Anthology Reference Corpus. We employed state-of-the-art graph visualization open-source tool Gephi to analyze the various aspects of citation network graphs, for each respective citation behavior.

Implications of Twitter in Health-Related Research: A Landscape Analysis of the Scientific Literature

Background: Twitter, representing a big social media network, is broadly used for the communication of health-related information. In this work, we aimed to identify and analyze the scientific literature on Twitter use in context of health by utilizing a bibliometric approach, in order to obtain quantitative information on dominant research topics, trending themes, key publications, scientific institutions, and prolific researchers who contributed to this scientific area. Methods: Web of Science electronic database was searched to identify relevant papers on Twitter and health. Basic bibliographic data was obtained utilizing the "Analyze" function of the database. Full records and cited references were exported to VOSviewer, a dedicated bibliometric software, for further analysis. A term map and a keyword map were synthesized to visualize recurring words within titles, abstracts and keywords. Results: The analysis was based on the data from 2,582 papers. The first papers were published in 2009, and the publication count increased rapidly since 2015. Original articles and reviews were published in a ratio of 10.6:1. The Journal of Medical Internet Research was the top journal, and the United States had contributions to over half (52%) of these publications, being the home-country of eight of the top ten most productive institutions. Keyword analysis identified six topically defined clusters, with professional education in healthcare being the top theme cluster (consisting of 66 keywords). The identified papers often investigated Twitter together with other social media, such as YouTube and Facebook. Conclusions: A great diversity of themes was found in the identified papers, including: professional education in healthcare, big data and sentiment analysis, social marketing and substance use, physical and emotional well-being of young adults, and public health and health communication. Our quantitative analysis outlines Twitter as both, an increasingly popular data source, and a highly versatile tool for health-related research.

The nature of rapid response to COVID-19 in Latin America: an examination of Argentina, Brazil, Chile, Colombia and Mexico

Purpose

The coronavirus-19 (COVID-19) pandemic mobilized the international scientific community in the search for its cure and containment. The purpose of this paper is to examine the nature of the rapid response to the COVID-19 of the scientific community in selected Latin American countries (Argentina, Brazil, Chile, Colombia and Mexico) in the period running from January to August 2020. Rapid response is reconceptualized from its original meaning in health policy, as the swift mobilization of existing scientific resources to address an emergency (DeVita et al., 2017).

Design/methodology/approach

The paper explores the rapid response of the Argentinian, Brazilian, Chilean, Colombian and Mexican scientific communities from the perspective of bibliometric and altmetric data. The authors will examine scientific publications indexed to the Web of Science (WoS) dealing with COVID-19. Besides patterns of scientific output and impact as measured by citations, the authors complement the analysis with altmetric analysis. The aim is to verify whether or not factors that explain the extent of scientific impact can also be identified with respect to the wider impact made evident by altmetric indicators (Haustein, 2016).

Findings

The authors identified a somewhat limited response of the Argentinian, Brazilian, Chilean, Colombian and Mexican scientific communities to COVID-19 in terms of quantity of publications. The authorship of publications in the topic of COVID-19 was associated with authorship of publications dealing with locally relevant diseases. Some factors appear to contribute to visibility of scientific outputs. Papers that involved wider international collaborations and authors with previous publications in arboviruses were associated with higher levels of citations. Previous work on arbovirus was also associated with higher altmetric attention. The country of origin of authors exerted a positive effect on altmetric indicators.

Research limitations/implications

A limitation in the analysis is that, due to the nature of the data source (WoS), the authors were unable to verify the career status and the productivity of the authors in the sample. Nonetheless, the results appear to suggest that there is some overlapping in authors conducting research in Arboviruses and COVID-19. Career status and productivity should be the focus of future research.

Practical implications

In the context of countries with limited scientific resources, like the ones investigated in our Latin American sample, previous efforts in the study of locally relevant diseases may contribute to the creation of an expertise that can be applied when a health emergency brings about a novel disease.

Originality/value

The originality of the paper rests on the fact that the authors identified that previous work on arbovirus contributed to the scientific visibility of publications on COVID-19.

Detailed analysis of Ethereum network on transaction behavior, community structure and link prediction

Ethereum, the second-largest cryptocurrency after Bitcoin, has attracted wide attention in the last few years and accumulated significant transaction records. However, the underlying Ethereum network structure is still relatively unexplored. Also, very few attempts have been made to perform link predictability on the Ethereum transactions network. This paper presents a Detailed Analysis of the Ethereum Network on Transaction Behavior, Community Structure, and Link Prediction (DANET) framework to investigate various valuable aspects of the Ethereum network. Specifically, we explore the change in wealth distribution and accumulation on Ethereum Featured Transactional Network (EFTN) and further study its community structure. We further hunt for a suitable link predictability model on EFTN by employing state-of-the-art Variational Graph Auto-Encoders. The link prediction experimental results demonstrate the superiority of outstanding prediction accuracy on Ethereum networks. Moreover, the statistic usages of the Ethereum network are visualized and summarized through the experiments allowing us to formulate conjectures on the current use of this technology and future development.

Sentiment analysis of tweets through Altmetrics: A machine learning approach

The purpose of the study is to (a) contribute to annotating an Altmetrics dataset across five disciplines, (b) undertake sentiment analysis using various machine learning and natural language processing–based algorithms, (c) identify the best-performing model and (d) provide a Python library for sentiment analysis of an Altmetrics dataset. First, the researchers gave a set of guidelines to two human annotators familiar with the task of related tweet annotation of scientific literature. They duly labelled the sentiments, achieving an inter-annotator agreement (IAA) of 0.80 (Cohen’s Kappa). Then, the same experiments were run on two versions of the dataset: one with tweets in English and the other with tweets in 23 languages, including English. Using 6388 tweets about 300 papers indexed in Web of Science, the effectiveness of employed machine learning and natural language processing models was measured by comparing with well-known sentiment analysis models, that is, SentiStrength and Sentiment140, as the baseline. It was proved that Support Vector Machine with uni-gram outperformed all the other classifiers and baseline methods employed, with an accuracy of over 85%, followed by Logistic Regression at 83% accuracy and Naïve Bayes at 80%. The precision, recall and F1 scores for Support Vector Machine, Logistic Regression and Naïve Bayes were (0.89, 0.86, 0.86), (0.86, 0.83, 0.80) and (0.85, 0.81, 0.76), respectively.

Measuring similarity in co-occurrence data using ego-networks

The co-occurrence association is widely observed in many empirical data. Mining the information in co-occurrence data is essential for advancing our understanding of systems such as social networks, ecosystems, and brain networks. Measuring similarity of entities is one of the important tasks, which can usually be achieved using a network-based approach. Here, we show that traditional methods based on the aggregated network can bring unwanted indirect relationships. To cope with this issue, we propose a similarity measure based on the ego network of each entity, which effectively considers the change of an entity's centrality from one ego network to another. The index proposed is easy to calculate and has a clear physical meaning. Using two different data sets, we compare the new index with other existing ones. We find that the new index outperforms the traditional network-based similarity measures, and it can sometimes surpass the embedding method. In the meanwhile, the measure by the new index is weakly correlated with those by other methods, hence providing a different dimension to quantify similarities in co-occurrence data. Altogether, our work makes an extension in the network-based similarity measure and can be potentially applied in several related tasks.