An Empirical Study on Artificial Intelligence for Sentiment Analysis Techniques

2. INTRODUCTION Since the inception of online social networks, the amount of social data that is being published on a daily basis has been increasing at an unprecedented rate. Smart, GPS-enabled, always-connected personal devices have taken the data generation to a new level by making it tremendously easy to generate and share social content like check-in information, likes, microblogs (e.g., Twitter), multimedia data, and so on. There is an enormous value in reasoning about such streaming data and deriving meaningful insights from it in real-time. Examples of potential applications include advertising, sentiment analysis, detecting natural disasters, social recommendations, personalized trends, spam detection, to name a few. There is thus an increasing need to build scalable systems to support such applications. Complex nature of social networks and their rapid evolution, coupled with the huge volume of streaming social data and the need for real-time processing, raise many computational challenges that have not been addressed in prior work. Social network data comprises of two major components. First, there is a network (linkage) component that captures the underlying interconnection structure among the entities in the social network. Second, there is content data that is typically associated with the nodes and the edges in the social network. The social network data stream contains updates to both these components. The structure of the network may itself change rapidly in many cases, especially when things like webpages and user tags (e.g., Twitter hashtags) are treated as nodes of the network. However, most of the social network data stream consists of updates to the data associated with the nodes and the edges, e.g., status updates and other content uploaded by the users, communication among the users, and so on. There is interest in performing a wide variety of queries and analytics over such data streams in real time. The queries can range from simple publish-subscribe queries, where a user is interested in being notified when something happens in his friend circle, to complex anomaly detection queries, where the goal is to identify anomalous behavior as early as possible. In this paper, we present an introduction to this new research area of stream querying and reasoning over social data. This area combines aspects from several well-studied research areas, chief among them, social network analysis, graph databases, and data streams. We provide a formal definition of the problem, survey the related prior work, and discuss some of the key research challenges that need to be addressed (and some of the solutions that have been proposed). We note that we use the term stream reasoning in this paper to encompass a broad range of tasks including various types of analytics, probabilistic reasoning, statistical inference, and logical reasoning. We contrast our use of this term with the recent work by Valle et al. [104, 105] who define this term more specifically to refer to integration of logical reasoning systems with data streams in the context of the Semantic Web. Given the vast amount of work on this and related topics, it is not our intention to be comprehensive in this brief overview. Rather we aim to cover some of the key ideas and representative work.

Background Since the COVID-19 pandemic, social media has become an important arena for the public to transmit and exchange messages, feelings, opinions, and information about the epidemic. In the era of social media, many UGC contents from self-media and various information about the epidemic on social media have strong emotional colors. These contents are not only rich in resources for text sentiment analysis but also reveal the laws and characteristics of the evolution of users' emotional tendencies in public health emergencies. It even ties together the interaction between media content and society. Objective As the Sina Weibo platform's characteristics of communication are real-time, open, and “many-to-many,” the objective of this study is to collect Weibo-blog contents tagged with the outbreak of COVID-19 in a certain metropolis in China and analyze the emotional evolution situation of Weibo-blogs of the unexpected public health emergency involved. This will provide a dynamic understanding of the mechanisms underlying the evolution of emotional conditions in the context of public health emergencies. Methods This paper uses a Python crawler, the SnowNLP sentiment analysis model, and correlation analysis to calculate the emotional tendencies of the event “Covid-19 outbreak in a Chinese metropolis” on the Sina Weibo platform. The study was carried out in terms of the evolutionary stages of the event and the factors which induce it. Results This paper revealed characteristics of time-varying laws and dynamic propagation of users' emotional evolution in public health emergencies. (1) This study refers to the life cycle model of COVID-19, combined with the statistics of the time series of the quantities of Weibo-blog posting, and divides the law of the quantities of Weibo-blog posting changing with the event into three stages: outbreak period, stalemate period, and resolution period. (2) Users' emotional tendencies are changeable and unstable which are easily induced by various factors. (3) There is a significant positive correlation between the reported confirmed cases and the quantities of Weibo-blog posts. (4) Individual emotional tendencies will have a positive changing trend with the public's average emotional tendencies after the event occurs. (5) There is no correlation between reposts, comments, and Weibo-blog emotional tendencies. Conclusion The research found that, given staged evolution and repeated fluctuations of emotional tendencies, relevant departments should effectively use this law and set up different response plans according to different stages. In addition, what is highly coupled with users' emotional tendencies is not only the information about the virus but more large-scale infected by different intensities of emotions.

With the increase in the number of users on social networks, sentiment analysis has been gaining attention. Sentiment analysis establishes the aggregation of these opinions to inform researchers about attitudes towards products or topics. Social network data commonly contain authors’ opinions about specific subjects, such as people’s opinions towards steps taken to manage the COVID-19 pandemic. Usually, people use dialectal language in their posts on social networks. Dialectal language has obstacles that make opinion analysis a challenging process compared to working with standard language. For the Arabic language, Modern Standard Arabic tools (MSA) cannot be employed with social network data that contain dialectal language. Another challenge of the dialectal Arabic language is the polarity of opinionated words affected by inverters, such as negation, that tend to change the word’s polarity from positive to negative and vice versa. This work analyzes the effect of inverters on sentiment analysis of social network dialectal Arabic posts. It discusses the different reasons that hinder the trivial resolution of inverters. An experiment is conducted on a corpus of data collected from Facebook. However, the same work can be applied to other social network posts. The results show the impact that resolution of negation may have on the classification accuracy. The results show that the F1 score increases by 20% if negation is treated in the text.

PurposeDespite the extensive academic interest in social media sentiment for financial fields, multimodal data in the stock market has been neglected. The purpose of this paper is to explore the influence of multimodal social media data on stock performance, and investigate the underlying mechanism of two forms of social media data, i.e. text and pictures.Design/methodology/approachThis research employs panel vector autoregressive models to quantify the effect of the sentiment derived from two modalities in social media, i.e. text information and picture information. Through the models, the authors examine the short-term and long-term associations between social media sentiment and stock performance, measured by three metrics. Specifically, the authors design an enhanced sentiment analysis method, integrating random walk and word embeddings through Global Vectors for Word Representation (GloVe), to construct a domain-specific lexicon and apply it to textual sentiment analysis. Secondly, the authors exploit a deep learning framework based on convolutional neural networks to analyze the sentiment in picture data.FindingsThe empirical results derived from vector autoregressive models reveal that both measures of the sentiment extracted from textual information and pictorial information in social media are significant leading indicators of stock performance. Moreover, pictorial information and textual information have similar relationships with stock performance.Originality/valueTo the best of the authors’ knowledge, this is the first study that incorporates multimodal social media data for sentiment analysis, which is valuable in understanding pictures of social media data. The study offers significant implications for researchers and practitioners. This research informs researchers on the attention of multimodal social media data. The study’s findings provide some managerial recommendations, e.g. watching not only words but also pictures in social media.

Online social networking (SN) data is a context and time rich data stream that has showed potential for predicting suicidal ideation and behaviour. Despite the obvious benefits of this digital media, predictive modelling of acute suicidal ideation (SI) remains underdeveloped at now. In combined with robust machine learning algorithms, social networking data may provide a potential path ahead. Researchers applied a machine learning models to a previously published Instagram dataset of youths. Using predictors that reflect language use and activity inside this social networking, researchers compared the performance of the out-of-sample, cross-validated model to that of earlier efforts and used a model explanation to further investigate relative predictor relevance and subject-level phenomenology. The application of ensemble learning approaches to SN data for the prediction of acute SI may reduce the complications and modelling issues associated with acute SI at these time scales. Future research is required on bigger, more diversified populations to refine digital biomarkers and assess their external validity with more rigor

Every day, vast amounts of social networking data is being produced and consumed at a constantly increasing rate. A user’s digital footprint coming from social networks or mobile devices, such as comments, check-ins and GPS traces contains valuable information about her behavior under normal as well as emergency conditions. The collection and analysis of mobile and social networking data before, during and after a disaster opens new perspectives in areas such as real-time event detection, crisis management and personalization and provides valuable insights about the extent of the disaster, its impact on the affected population and the rate of disaster recovery. Traditional storage and processing systems are unable to cope with the size of the collected data and the complexity of the applied analysis, thus distributed approaches are usually employed. In this work, we propose an open-source distributed platform that can serve as a backend for applications and services related to crisis detection and management by combining spatio-textual user generated data. The system focuses on scalability and relies on a combination of state-of-the art Big Data frameworks. It currently supports the most popular social networks, being easily extensible to any social platform. The experimental evaluation of our prototype attests its performance and scalability even under heavy load, using different query types over various cluster sizes.

The traditional data query algorithm based on clustering strategy library ignores the association features of social network data, characteristic data acquisition exist a large number of redundant features and frequent relationship among features is low, resulting in the social network data query efficiency and the accuracy is poor, so a fast query algorithm for social network data based on fuzzy degree function based on association features is proposed, it is based on Apriori algorithm for data association feature mining of social network to obtain the maximum frequent association feature set; for association feature preprocessing, it reduce the maximum frequent association feature set by feature dimension reduction and de redundancy algorithm, to obtain better social network maximal frequent associated feature set; when using fuzzy function to query social network data quickly, it uses data of a single gene ambiguity function to build a fast data query diagram, input the best frequent feature set of social network, and output the query results of social network data with the highest priority. The experimental results show that the proposed algorithm has the advantages of high efficiency and high accuracy in social network data query.

Social network is model in order to structuring social. In knowledge management, social structures need to be revealed to see social behavior and social change based on the interactions that have taken place. In knowledge technology, the semantic social structure is extracted from information sources to interpret social behavior and its changes, and the meaning is supported by social network analysis. However, social networks do not only reveal social structures and manage them, but social recognition by exploring the potential of social members (social actors) and the community so as to present competitiveness. Competence is defined by the existence of social actors, while the existence of social actors is expressed from the relationship between them whereby the relationship also proves the meaning the competence. By involving artificial intelligence (AI), the meaning can be affirmed in a variety of different ways depending on the treatment given to the source of information so that it is possible to predict and express on the characteristics and behavior of the data. It based on forensic data and information retrieval, for trusty information is generated. Moreover, social network mining is present by considering social network data and the resultant of extraction method so that social networks not only contain information but become knowledge. This serves to bridge the social network analysis gap based on primary and secondary data, so social engineering is possible by exploring all the potential presented by social network extraction.

Grouping users in social networks is an important process that improves matching and recommendation activities in social networks. The data mining methods of clustering can be used in grouping the users in social networks. However, the existing general purpose clustering algorithms perform poorly on the social network data due to the special nature of users' data in social networks. One main reason is the constraints that need to be considered in grouping users in social networks. Another reason is the need of capturing large amount of information about users which imposes computational complexity to an algorithm. In this paper, we propose a scalable and effective constraint-based clustering algorithm based on a global similarity measure that takes into consideration the users' constraints and their importance in social networks. Each constraint's importance is calculated based on the occurrence of this constraint in the dataset. Performance of the algorithm is demonstrated on a dataset obtained from an online dating website using internal and external evaluation measures. Results show that the proposed algorithm is able to increases the accuracy of matching users in social networks by 10% in comparison to other algorithms.

This study examines how tourists perceive food quality in Berlin using AI-driven sentiment analysis tools. The goal is to understand the factors shaping tourists’ perceptions and provide insights to improve the hospitality industry and customer satisfaction. By analyzing reviews from online platforms, this research identifies key themes and trends in tourists’ feedback. The use of AI, specifically for sentiment analysis, supports efficient and detailed evaluation of customer opinions. This study employed lexicon-based sentiment analysis to evaluate tourists’ feedback on online platforms and compared the sentiment scores of textual feedback with their direct rating scores. The results show that integrating sentiment scores derived from AI tools with tourists’ rating scores provides deeper insights into service quality within the tourism sector.

Social networks contain sensitive information, and direct press conferences disclose people’s privacy. The published social network data needs privacy protection. The currently released social network data has seriously damaged the network structure due to privacy protection. This paper proposes a social network graph publishing model that preserves the social network structure. This model builds an uncertainty graph based on privacy protection probability to generate social network data for publication. It is proved theoretically that the social network graph publishing model satisfies the definition of differential privacy. This paper uses three real social network datasets to evaluate the effectiveness of the proposed method, showing that the method effectively retains the network structure.

The impact of the World Wide Web goes beyond communicating between computers. The web changed the way we interact with organizations, people, documents, and data. Social media is one of the new buzzwords of the web. It is a comparatively new term that covers a wide range of online applications, platforms, and media that support online interactions, collaboration, and the sharing of content. It includes all that constitutes human interaction in online platforms. Social media has had grown explosively in a short time and has a profound impact on advertising, publicity, marketing, public opinion, entertainment, software services, and decision-making. The content offered in social networks is exciting to users. However, it is challenging, as users need to spend time and effort finding content that might be of their interest. Navigating such an extensive collection of items becomes difficult if there are no tools to help the users. To overcome this problem, social media relies on data mining solutions such as recommender systems and information retrieval systems. In particular, recommender systems have found applications in many domains including movies, music, videos, news, books, and products in general. Depending on the context they produce a list of recommended items using a variety of techniques. The goal of this dissertation is to research new approaches in recommender systems that help to satisfy the needs of users in social media. Specifically, we examine the role of recommender systems for users in two types of social media, Location-Based Social Networks and academic social networks. LBSNs enable their users to share the places they go to and with whom they are. Academic social networks help modern researchers organize their scientific libraries and discover relevant papers to their research. For both types of social media most of the existing work focuses on recommendations for individual users. The proposed approaches are distinguishable from others in that we focus on providing recommendations to a group of users, rather than to individuals. Moreover, we investigate the importance of item-to-item recommendations and propose a new method for recommending on infrequent items.

Social support networks have been become one of the mostly influential metaphors in many areas around the world, especially in terms of mitigating disaster consequences and revivifying disaster affected livelihoods. More importantly, reciprocal supports ties are more powerful in healing disaster wounds of communities. Especially social capital legacies enrich by reciprocal support networks by the ways in which making may hopes among disaster victims. In this context, the objective of this study is to investigate the social support network behaviors in supporting 16 flood-affected households in the mass flooding event occurred in 2017. Particularly, the study researched social support network behaviors at different flood inundation phases such as before, during, and after. This study collected primary data (mainly social network data) using household survey and filed observations. The study used the Social Network Analysis (SNA) method for the network data analysis. This article shows that flood affected households have received social network supports in different magnitudes at before, during, and after flood inundation phases. More importantly, provision of foods, water and basic needs, sheltering, clearing contaminated households, and emotional supports have mobilized and reciprocated among victims in reviving their livelihoods. The social support network legacies have evolved at different flood inundation phases. This study fills the gabs in the flood disaster discourse on Sri Lankan context.

The students' learning styles affect the understanding of the materials provided in university. In the learning process, emotional factors strongly affect students' motivation and learning style. Students post their activities, experiences, feelings in Twitter. Tweets sentiment analysis results represent the students' emotional tendency. The knowledge about learning styles and emotional tendency are employed to find suitable teaching techniques and approaches. The study employs learning style that involves visual, aural, reading/writing, kinaesthetic (VARK) to identify the students learning styles. The results are that students' learning styles are classified in 13.33% visual, 36.67% aural, 20% read/write and 30% kinaesthetic. The emotional tendencies of the students are 33.33% positive, 53.33% neutral, 13.33% negative. The mapping concept based on the learning styles and emotional tendencies finds that positive emotional tendency is related to all learning styles. The neutral emotional tendency is related with aural, read/write, kinaesthetic. The negative emotional tendency is related to visual.

Social media is a valuable source of information that allows to study people opinions of many events that happen every day. Nowadays social networks are one of the most important communication methods that people use. Feelings towards nations can be measured today thanks to the advances in machine learning and big data. In this paper we present a method to identify if there are negative sentiments towards China as a result of the COVID-19 virus. The method is based on sentiment analysis and extracts information from the Twitter social network. This analysis was done with the VADER library, a rule-based tool that provides classification algorithms. A dataset of 30,000 tweets was built for three time windows: December 2019 (before the pandemic), March 2020 (month in which the pandemic was confirmed by the World Health Organization), and May 2020 (when some countries started the de-escalation phase). Results show that sentiments became negative towards China and social network data allows to confirm this situation.KeywordsSentiment towards ChinaText miningVADERMachine learningCOVID-19