Negative Sentiment Research Articles

Strategi Social CRM Dalam Menangani Keluhan Pelanggan Pada Instagram Pos Indonesia

Every day, many customers submit complaints through the comment section on Pos Indonesia's main Instagram account (@posindonesia.ig). This study aims to understand how Pos Indonesia designs its Social CRM strategy to address customer complaints on the Instagram social media platform @posindonesia.ig. The theory employed in this research is the Social Media SOME model (Share, Optimize, Manage, Engage) developed by Luttrell (2018). A qualitative approach with descriptive research is utilized in this study. Data collection methods involve in-depth interviews, observation, and document analysis. The findings of this research reveal that the Social CRM Strategy implemented by Pos Indonesia's internal communication and contact center includes four main activities: (1) Share, where internal communication strives to reduce negative sentiment by paying attention to the content liked by followers, (2) Optimize, where internal communication focuses on educational content presented in a socializing format, (3) Manage, where internal communication monitors social media content to assess the level of impressions and sentiments from followers, and (4) Engage, where internal communication involves internal buzzers to provide positive comments with the hope that customers planning to express negative complaints will change their minds.

Analysis of Indonesian Public Sentiment towards the Cities of Jakarta, Surabaya and Medan Using the Naïve Bayes Algorithm

The theory of sentiment analysis is interesting to research, this theory aims to group the polarity of a text in a sentence, feature/aspect, or document and then determine whether the opinion contained in the sentence, feature/aspect, or document has positive or negative sentiment. . Sentiment analysis helps in understanding how people respond to certain issues related to these cities (Jakarta, Surabaya, Medan). This research uses the Naïve Bayes algorithm by conducting sentiment analysis on cities such as Jakarta, Surabaya and Medan, we can gain valuable insights to support better decision making in various aspects of city development and management. The results of this research provide an accuracy value of 77% for the city of Jakarta, 76% for the city of Surabaya, 75% for the city of Medan. These results also show that the Naïve Bayes model has quite good performance in carrying out sentiment analysis. Keywords: Indonesia, Sentimen, Naïve Bayes, Twitter

Sentiment Analysis of Alfagift Application User Reviews Using Long Short-Term Memory (LSTM) and Support Vector Machine (SVM) Methods

The rapid advancement of mobile apps has emerged as an important aspect of the routine of internet-connected users. In Indonesia, many companies are introducing their apps to improve the quality of service for users, and Alfamart is one of them. However, users have identified many shortcomings in these apps. This feedback is provided by users on the review feature of the Alfagift app on the Google Play Store. This research aims to apply a sentiment analysis approach to identify the application's shortcomings so that developers can understand the aspects that need to be improved to improve the quality of application services. The research stages include data collection, preprocessing, labeling, weighting, classification of LSTM and SVM methods, and performance evaluation using a confusion matrix. The dataset consists of 1000 reviews obtained through web scraping techniques. This research uses the Lexicon-based method to classify the dataset into positive, negative, and neutral categories. The analysis results show that 801 data are classified as positive sentiment, 77 as negative sentiment, and 122 as neutral sentiment. Based on testing, both SVM and LSTM methods show good performance. The best accuracy results were obtained using the SVM method, which amounted to 83.5%. Meanwhile, the LSTM method achieved an accuracy of 82%.

Financial Report Sentiment Analysis Using Loughran-mcdonald Dictionary and BERT

In the ever-changing world of financial markets, understanding investor behavior and making informed decisions relies heavily on sentiment analysis. This study delves into the integration of traditional techniques, such as the Loughran- McDonald dictionary, with advanced natural language processing (NLP) methods utilizing BERT (Bidirectional Encoder Representations from Transformers). The goal is to enhance the accuracy and depth of sentiment analysis in financial reports.To begin, we employ the specialized Loughran-McDonald dictionary designed for financial sentiment analysis. This lexicon includes domainspecific word lists for positive and negative sentiments, forming a solid foundation for sentiment scoring. Expanding on this foundation, we incorporate BERT, an advanced transformerbased NLP model. BERT’s contextual understanding of language and ability to capture intricate semantic relationships within financial texts aim to overcome the limitations of rule-based sentiment analysis. The methodology involves preprocessing financial reports, integrating Loughran-McDonald sentiment scores, and fine-tuning BERT for financial sentiment classification. This hybrid approach leverages both the domain expertise encoded in the dictionary and BERT’s contextual comprehension of financial jargon and nuances. We validate and evaluate our implementation using a diverse dataset comprising quarterly earnings releases, annual reports, and other relevant disclosures. Performance metrics such as precision, recall, and F1 score are analyzed to assess the effectiveness of our hybrid approach compared to individual methods. The findings have significant implications for financial analysts, investors, and policymakers by providing a more nuanced understanding of sentiment in financial reports. Our hybrid approach aims to offer improved accuracy in capturing sentiment polarity while facilitating more informed decision-making in today’s complex and dynamic realm of financial markets.

Situated usage of generative AI in policy education: implications for teaching, learning, and research

ABSTRACT This study explores the contrasting sentiments towards the use of generative AI technologies among research postgraduate students in public policy. 14 interviews about the usage of generative AI technologies in the students’ research, teaching, and learning practices were conducted and used as the empirical data source for this project. Through qualitative and sentiment analysis, the research identified domains where students applied generative AI and discovered both positive and negative sentiments within the same application domains. The divergence in sentiments was interpreted using the ‘plans and situated actions’ framework, suggesting that technological expectations constrained by contextual environments lead to varied experiences of ‘enchantment’ and ‘disenchantment’. The findings emphasize the imperative for adaptable academic policies delineating acceptable AI usage in research, the implementation of discipline-specific AI training in universities, and the development of discipline-specific AI systems to cater to unique academic field needs.

Analisis Sentimen Ibu Kota Nusantara menggunakan Algoritma Support Vector Machine dan Naïve Bayes

The Government's policy of moving the Indonesian capital city (IKN) is considered controversial, sparking various responses from the public, especially on the social media platform X. This research aims to analyze tweet sentiment related to IKN and compare two algorithms. In this experiment, we collected 5,128 tweets regarding IKN from the X application. The dataset was classified into 2,598 positive sentiments and 1,659 negative sentiments. To analyze these sentiments, we used Text Mining techniques, comparing the Support Vector Machine (SVM) and Naive Bayes algorithms. To improve the performance of these algorithms in analyzing the data, SMOTE optimization was employed to address data imbalance. Our findings show that the SVM algorithm achieves an accuracy of 84%, while the Naive Bayes algorithm achieves an accuracy of 77%. Thus, it can be concluded that the SVM algorithm is superior to the Naive Bayes algorithm. Furthermore, the use of SMOTE optimization proved to enhance the ability of both algorithms to identify positive sentiment, as evidenced by the precision, recall, and F1-Score values. The SVM algorithm achieved a precision of 82%, recall of 86%, and F1-Score of 84%, while the Naive Bayes algorithm achieved a precision of 71%, recall of 92%, and F1-Score of 80%.

Analisis Sentimen Ulasan Aplikasi PosPay untuk Meningkatkan Kepuasan Pengguna dengan Metode K-Nearest Neighbor (KNN)

PT Pos Indonesia has launched a digital pospay service. Users who have a positive experience are more likely to return to application. User perceptions analysis can be known from the Review sentiments. Review sentiments that are classified as positive and negative are really needed by developers to improve services (user satisfaction). The research aims to increase user satisfaction of the PosPay application based on the application's review data. The source of data is a review of the pospay application at Google play store. The method used quantitative study method that is K-Nearest Neighbor (K-NN) that classify objects based on learning data that are closest to the object. Research variable is the word from user commentary that associated with the pospay application services. Application review data in scrapping, preprocessing, splits data (train data and test data). Supervised learning (TF-IDF and K-NN) prepared with python programming provides data visualizing. The research results show that the sentiment of Pospay application users tends to be positive. K-NN classification model produces 91% accuracy, 90% precision and recall by 99%. The key word of positive sentiment is: easy, helpful, transaction. Keyword negative sentiment: balance, pay, login.

Deciphering Negative Feedback through Handwritten Text

In the contemporary realm of digital communication, this study delves into a novel approach to analyzing sentiments, with a specific focus on handwritten negative reviews. The research aims to improve the precision of sentiment analysis by exploring the inclusion of handwritten text, providing distinctive insights into consumer feedback, and refining business strategies. Taking a historical perspective, the investigation traces the evolution of sentiment analysis within the domain of natural language processing, leading to the central question: "How can the accuracy of sentiment analysis for negative reviews be enhanced through the incorporation of handwritten text?" This question guides an exploration of the challenges and potentials associated with merging handwriting analysis with conventional sentiment analysis methods. The study puts forth hypotheses that address the expected advantages of this integration, seeking to develop an innovative framework capable of not only accurately detecting negative sentiments but also considering the individuality inherent in handwritten expressions. The research methodology encompasses a review of existing literature and empirical analysis, resulting in the creation of a unique hybrid sentiment analysis algorithm that assesses both textual and handwriting features. This work contributes to the advancement of sentiment analysis and has implications for businesses aiming to better understand and respond to negative consumer sentiments.

Sentiment Analysis of E-Grocery Application Reviews Using Lexicon-Based and Support Vector Machine

This research aims to conduct sentiment analysis of e-grocery application reviews using the Support Vector Machine (SVM) algorithm. Sentiment analysis is used to distinguish between positive and negative reviews by users who have provided reviews so that an evaluation of the services offered can be made. This research uses scraping techniques to obtain all the needed review data, focusing only on reviews of the Segari and Sayurbox applications. Datasets were collected from reviews using a library in Python, namely, google-play-scraper, obtained by the sayurbox application 4235 reviews and the segari application 5575. The dataset collected does not yet have a label, and the labeling process is impossible to perform manually by looking at the reviews one by one because it takes a long time and requires an expert in the field of language who can interpret the reviews and group them into positive and negative sentiments. Therefore, the sentiment-labeling process applies a lexicon-based method that works based on the inset lexicon dictionary by calculating each review's polarity value. The analysis process of this research uses the SVM algorithm because the SVM method has been proven to provide consistent and accurate results in various classification tasks, including sentiment analysis. The results show that the lexicon-based method and SVM produce good accuracy in determining the sentiment of e-grocery reviews, with a vegetable box application accuracy rate of 94%. In comparison, the segari application accuracy rate reached 97%.

Dynamics of Indonesian Public Opinion on the Rohingya Crisis in Time Perspective Using Traditional Machine Learning and Deep Learning

The Rohingya are an ethnic minority who currently still face persecution and discrimination in Myanmar, so they have to flee to neighboring countries, such as Indonesia. However, the polemic regarding the issue of the existence of Rohingya refugees in Indonesia still shows that there are differences of opinion between groups who support and oppose it. For this reason, this research aims to determine the dynamics of Indonesian public opinion regarding the Rohingya from 2015-2023 via Twitter, as well as find out the topics that are often discussed each year using LDA. This research compares classification methods using traditional machine learning algorithms (NB, SVM, LR, and DT) and deep learning algorithms (LSTM, GRU, LSTM-GRU, and GRU-LSTM). The research results show that the traditional machine learning algorithm, LR, has the highest accuracy. There has been a change in sentiment from initially being dominated by positive sentiment to negative sentiment which is more dominant in the last five years. The topics that are often discussed for positive sentiment are the support of the Indonesian people for the Rohingya in providing assistance and shelter, while the negative topics are related to concerns about the social, economic, and security impacts that may be caused by the presence of Rohingya refugees.

Media framing of artificial intelligence: an analysis of framing in Romanian online publications

Artificial intelligence (AI) is increasingly at the center of public discourse. Considering the current importance of AI technologies, but also the influence that the mass media has on users' adoption of new technologies and their regulation, the article aims to analyze the attitudes of online publications in Romania towards AI. By using content analysis, the attitudes and themes of 100 articles selected from the most visited Romanian news websites were examined. The results reveal a relatively balanced distribution of attitudes, with 37% of articles expressing negative sentiments, 35% positive and 28% neutral. The most frequently addressed topics include the impact on jobs, especially the negative one, regulations in the field and new functionalities. Compared to the publications in English, as can be seen from the specialized literature, the Romanian articles present a more balanced attitude towards technology, with less emphasis on the dangers. However, the analysis shows that current events can strongly influence public discourse, and approaches vary by publication. These findings suggest a diverse debate, subject to rapid changes, with important implications for AI communication, education and policy in the Romanian space.

Analyzing an Interest in GPT 4o through Sentiment Analysis using CRISP-DM

This study investigates the sentiment of viewers towards GPT-4o technology videos by analyzing 1538 English language posts using two sentiment analysis tools, VADER and TextBlob. The analysis reveals a fair level of agreement between the two tools, with 929 posts (60.40%) classified consistently, yielding a Cohen’s kappa statistic of 0.388. The sentiment distribution among the posts is as follows: 182 posts (19.59%) exhibit negative sentiments, 390 posts (41.98%) are neutral, and 357 posts (38.43%) show positive sentiments. These findings highlight the importance of utilizing multiple tools for comprehensive sentiment analysis and underscore the complexity of interpreting public reactions to AI advancements. The study provides valuable insights into the nuanced responses of viewers, emphasizing the diverse perspectives towards the GPT-4o technology.

Semi-supervised topic representation through sentiment analysis and semantic networks

This paper proposes a novel approach to topic detection aimed at improving the semi-supervised clustering of customer reviews in the context of customers' services. The proposed methodology, named SeMi-supervised clustering for Assessment of Reviews using Topic and Sentiment (SMARTS) for Topic-Community Representation with Semantic Networks, combines semantic and sentiment analysis of words to derive topics related to positive and negative reviews of specific services. To achieve this, a semantic network of words is constructed based on word embedding semantic similarity to identify relationships between words used in the reviews. The resulting network is then used to derive the topics present in users' reviews, which are grouped by positive and negative sentiment based on words related to specific services. Clusters of words, obtained from the network's communities, are used to extract topics related to particular services and to improve the interpretation of users' assessments of those services. The proposed methodology is applied to tourism review data from Booking.com, and the results demonstrate the efficacy of the approach in enhancing the interpretability of the topics obtained by semi-supervised clustering. The methodology has the potential to provide valuable insights into the sentiment of customers toward tourism services, which could be utilized by service providers and decision-makers to enhance the quality of their services.

Sexual assault and the matrix of harm: Sexual assault survivors narrate their whole lives in more negative ways.

This paper uses data from the American Voices Project, an interview study based on a random population sample, to explore the relationship between assault experiences and how people narrate their lives. Using quantitative sentiment analysis, we find that survivors of assault express significantly greater negative sentiment when asked to tell their life stories. These negative sentiments are observable throughout the entire interview, including before questions of assault are asked. Survivors of assault narrate their experiences with more anger, disgust, fear, and sadness, and less anticipation, joy, and trust than those who do not report assault experiences. We provide evidence that the negative sentiment associated with sexual violence is greater than having lost a parent and less than having experienced a significant financial hardship within the last month. We contextualize these findings with a qualitative evaluation of the interview transcripts, further substantiating our finding. Overall, this paper suggests that sexual violence is part of what we have called, drawing inspiration from the work of Beth Richie, a "matrix of harm" that structures people's lives. While our finding is consistent with what we might expect given the negative life experiences and sequalae associated with sexual assault, it has important implications. Sentimental differences in narrating life experience are an important yet relatively understudied phenomenon, and experiences of assault are rarely asked about yet may be consequential to both quantitative and qualitative accounts of social processes.

Formal long-term care, individualisation and filial responsibility: a multi-level analysis of 21 European countries

PurposeDue to the rapid ageing of the continent's population, a significant surge in long-term care expenses for the elderly is expected across Europe in the coming years. Could a potential solution to this challenge involve the increased informal care provided by adult children? In this context, we examined a general view or moral duty of European citizens regarding whether adult children should assume responsibility for providing long-term care to the elderly.Design/methodology/approachOur multilevel analysis draws on individual-level data from the 2017 European Value Study and country-level data from various European sources encompassing participants from 21 member countries of the European Union.FindingsThe findings reveal that in nations where public long-term care services are sufficiently available and of superior quality, there is a negative sentiment towards the notion that adult children should bear the responsibility for elderly care. In total, 71% of the country-level variance in our dependent variable was explained by the availability and quality of formal long-term care in a country. Furthermore, various individual-level attributes contribute to shaping attitudes towards care-giving responsibility. We observed that women, middle-aged individuals, those without religious affiliations, those with modern gender role attitudes and non-immigrants tended to hold unfavourable attitudes towards the responsibility of adult children in long-term care provision.Originality/valueThere are relatively many studies on the general attitude of the population towards filial obligation. However, so far there have been very few studies available that examine the population's attitude towards the obligation of adult children to commit to their parents' long-term care. Our research explains the variation of the phenomenon in Europe with both country-level and individual-level factors.

Analisis Sentimen Terhadap Kinerja Kepolisian Indonesia Menggunakan Metode Multinomial Naive Bayes, Long Short-Term Memory, dan Lexicon-Based

According to the Indonesian Survey Institute, the public trust index in the Indonesian Police decreased to 53% as of October 2022. The level of public satisfaction and trust in police can be measured by examining how the police handle each case that occurs. This research focuses on analyzing and measuring the public’s sentiment toward the police’s performance among Twitter users in Indonesia. After scraping (collecting) data from Twitter, out of the 14,000 collected tweets, only 12,222 tweets were for analysis after the preprocessing stage. Then, the data was labeled into two categories, negative sentiment and positive sentiment using InSet Lexicon. The results showed that 53% (6,478) tweets have a negative sentiment, while 47% (5,744) tweets have a positive sentiment. The topics frequently appear and discussed related to such as the shooting case of Brigadier J, the Kanjuruhan tragedy, the security of the G20 Bali Summit, securing Christmas 2022 and securing New Year 2023. Subsequently, classification was performed on the test data using the Multinomial Naïve Bayes and Long Short-Term Memory (LSTM) methods. In the testing phase, a confusion matrix used to calculate accuracy, recall, precision, and f1-score. The results showed that the Multinomial Naïve Bayes method achieved an accuracy score of 78%, precision of 78.5%, recall of 77%, and an f1-score of 77%, with an average score of 77.6%. Similarly, the LSTM method resulted in an equal accuracy, precision, recall, and f1-score of 87%.

Sentiment Analysis of Google Play Store User Reviews on Digital Population Identity App Using K-Nearest Neighbors

The Digital Population Identity Application provides convenience for users to access and manage their population data digitally. Based on the increasing usage of the Digital Population Identity Application on the Google Play Store, various user reviews of the application have emerged. Therefore, sentiment analysis is needed to provide a deeper understanding of user perceptions and to classify user reviews of the Digital Population Identity Application. Sentiment analysis is a computational study of opinions, feelings, and emotions expressed in text, using the K-Nearest Neighbors method, which is a classification method based on the closest distance or similarity to objects in the training data. Using 5000 relevant review data from September 2022 to December 2023, after labeling them into positive, negative, and neutral sentiment classes, the results show 3581 negative sentiments, 1031 positive sentiments, and 388 neutral sentiments. Testing was conducted by applying the K-Nearest Neighbors method in the classification stage, testing this method by varying K values from 1 to 10. The best results were obtained with a training data ratio of 90% to testing data ratio of 10%. The best results were achieved at K values of 8, 9, and 10, with an accuracy of 81%, precision of 82%, recall of 95%, and an F1-Score of 88%. With a training data ratio of 70% to testing data ratio of 30%, the best results were obtained at K values of 6, 7, 8, 9, and 10, with an accuracy of 80%, precision of 81%, recall of 95%, and an F1-Score of 88%. Based on the results of this research, the K-Nearest Neighbors method can be used for sentiment classification of user reviews with good results.

Comparative Analysis of SVM and NB Algorithms in Evaluating Public Sentiment on Supreme Court Rulings

The legal events that happened to Ferdy Sambo and the Supreme Court’s decision in the cassation triggered emotional reactions and various opinions among the public, especially on social media sites such as Xapps. Some comments reflect people’s concerns about fairness in the legal system. They doubted the integrity of legal institutions or believed that decisions were unfair or in line with vested interests. This research aims to analyze public perceptions of Supreme Court decisions. The research process includes data collection, preprocessing, labeling, weighting, classification using Support Vector Machine and Naïve Bayes, and performance evaluation using a confusion matrix. A dataset of 624 was taken from X apps using the Twitter scraping technique. The lexicon method is used for data labeling, dividing the data into positive, negative, and neutral classes. The analysis results show 46 tweets categorized as positive sentiment, 133 tweets categorized as negative sentiment, and 422 tweets categorized as neutral sentiment. Based on testing with a data ratio of 80:20, both SVM and NB methods show good performance. The SVM criteria showed an accuracy of 0.84, precision of 0.61, recall of 0.78, and f1-score of 0.66, while the NB criteria showed an accuracy of 0.73, precision of 0.37, recall of 0.57, and f1-score of 0.35.

Dissecting the impact of the three E, S, G pillars on credit risk

This study explores the impact of Environmental, Social, and Governance (ESG) scores on the credit risk of Chinese A-share listed companies. Based on the empirical study, we analyse both the individual and combined effects of E, S, and G scores. Our findings indicate that improvements in ESG scores significantly reduce credit risk, particularly among green companies and those with high transparency. This conclusion remains robust after accounting for potential endogeneity bias through the application of the instrumental variable method. Furthermore, our study delves into how ESG scores impact credit risk. An increase in ESG scores reduces a company’s debt default risk by decreasing the negative ESG-related news sentiment, enhancing stock liquidity, lowering financing costs, and improving the quality of internal controls. These insights contribute to a deeper understanding of ESG factors in risk management and offer practical guidance for companies seeking to enhance their ESG performance in relation to credit risk.

Analisis Sentimen Masyarakat Berdasarkan Komentar Kerja Sama Tiktok Shop dan Tokopedia di Instagram Menggunakan Metode Naïve Bayes Classifier

The reopening of the Tiktok Shop and collaboration with Tokopedia has caught the public's attention. That's proven by a post on Instagram about their collaboration that received various responses. This research aims to conduct sentiment analysis to ascertain whether the public approves of the two's partnership. The Naïve Bayes Classifier method was used to analyze 641 comment data from the period 11 December 2023 to 11 February 2024. The results show the composition of positive, negative, neutral sentiment and unclassified data, as well as accuracy, precision and recall. Of the 641 data recorded, there were 269 neutral sentiment data, 194 negative sentiment data, 176 positive sentiment data, and 437 unclassified data. The Naïve Bayes Classifier model with oversampling techniques obtained an accuracy of 83%, precision of 81%, and recall of 93%.