Next Article in Journal
Evaluating Adversarial Robustness of No-Reference Image and Video Quality Assessment Models with Frequency-Masked Gradient Orthogonalization Adversarial Attack
Previous Article in Journal
Deep Learning-Based Draw-a-Person Intelligence Quotient Screening
Previous Article in Special Issue
Exploiting Rating Prediction Certainty for Recommendation Formulation in Collaborative Filtering
 
 
Font Type:
Arial Georgia Verdana
Font Size:
Aa Aa Aa
Line Spacing:
Column Width:
Background:
Article

Exploring Factors Impacting User Satisfaction with Electronic Payment Services in Taiwan: A Text-Mining Analysis of User Reviews

by
Shu-Fen Tu
1 and
Ching-Sheng Hsu
2,*
1
Department of Information Management, Chinese Culture University, Taipei City 111396, Taiwan
2
Department of Information Management, Ming Chuan University, Taoyuan City 333321, Taiwan
*
Author to whom correspondence should be addressed.
Big Data Cogn. Comput. 2025, 9(7), 165; https://doi.org/10.3390/bdcc9070165
Submission received: 27 April 2025 / Revised: 18 June 2025 / Accepted: 20 June 2025 / Published: 25 June 2025
(This article belongs to the Special Issue Business Intelligence and Big Data in E-commerce)

Abstract

Electronic payments are becoming increasingly popular in Taiwan; however, there is a lack of studies examining the factors affecting user satisfaction with electronic payments in Taiwan. This study focuses on Android phone users to identify key factors influencing their experiences based on user reviews of electronic payment mobile applications. It analyzes which factors contribute to positive satisfaction and which lead to negative experiences. The study employed BERTopic for topic modeling, which flexibly accommodates multiple languages, enabling effective examination of reviews written in Chinese. Additionally, we utilized the semantic understanding capabilities of large-scale language models to preliminarily name the generated topics with the help of ChatGPT Plus. These preliminary names were then manually refined to determine the final topic titles. The findings reveal that for Android phone users, electronic payment services that enhance user convenience and offer discounts tend to foster positive satisfaction. Conversely, the instability of electronic payment applications results in many user complaints. These research results can provide valuable insights for specialized electronic payment institutions in Taiwan to enhance their services.

1. Introduction

The emergence of mobile devices and advancements in wireless communication have greatly changed our lifestyles. Accessing information and obtaining services has become routine, overcoming the restrictions of time and space [1,2]. The extensive use of mobile devices has lowered consumers’ search costs and facilitated more straightforward access to product information [3]. As a result, an increasing number of users are shopping through their mobile devices, which has led to a greater demand for electronic payment services [4]. To become an electronic payment institution in Taiwan, an organization must obtain approval from the relevant authorities, and its business items are regulated by law. We will provide a detailed explanation of this process in Section 2.4. Consumers access electronic payment services through mobile applications offered by these institutions. Mobile applications can influence user satisfaction in various ways, subsequently affecting how users perceive the service provider [5,6,7]. This is also true for electronic payment systems. To enhance the quality of electronic payments, it is crucial to understand users’ satisfaction levels with payment apps. In recent years, the impact of the pandemic has led many Taiwanese consumers to favor electronic payments for their contactless and non-cash features. Recent data indicates that the annual growth rate for electronic payments has reached 46%, rising from over TWD 111.4 billion in 2022 to TWD 163.5 billion in 2023 [8]. As of the end of November 2024, the total number of electronic payment account users in Taiwan had reached 30.38 million, which averages approximately 1.3 electronic payment accounts per person [9]. These statistics indicate significant recent growth in Taiwan’s electronic payment market. However, there is still limited research available on the topic of electronic payments in Taiwan. Liu and Tsai [10] conducted a study on the preparations for mobile payments in Taiwan by interviewing experts in the field. Kuo [11] examined the factors influencing Taiwanese users’ willingness to switch mobile payment platforms. Meanwhile, Tsai et al. [12] utilized the Technology Acceptance Model (TAM) [13,14,15,16] and the Fuzzy Analytic Hierarchy Process (FAHP) [17] to explore Taiwanese consumers’ willingness to use mobile payments in e-commerce. Additionally, Ng et al. [18] analyzed the strategies employed by Taiwanese companies to adapt to the growing trend of mobile payments and the associated innovation ecosystem. To our knowledge, there are limited studies that examine the factors influencing user satisfaction with electronic payment apps. While Chen et al. [19] investigated the factors affecting user behavior in bank mobile payment apps in Taiwan, it is important to note that banks operate as part-time electronic payment institutions. Taiwan has nine specialized electronic payment institutions, each with its own exclusive app. However, no research has focused on user satisfaction related to these nine electronic payment apps.
Some studies have designed questionnaires based on specific theoretical models to explore the factors affecting app user satisfaction. For instance, Chen et al. [19] developed a questionnaire grounded in the Technology Acceptance Model (TAM) and consumer values [20] to investigate user behavior related to bank mobile payment apps. Additionally, some studies have analyzed user-generated content (UGC), such as app user reviews, to understand the factors influencing app user satisfaction. Analyzing user-generated content (UGC) offers several advantages over traditional questionnaire survey methods, as it is less prone to sample selection bias and offers greater data collection efficiency. As a result, many recent studies have investigated user satisfaction by examining app user reviews [21]. The analysis process usually begins with preprocessing the user reviews, followed by topic modeling to identify the most frequently discussed subjects. Additionally, sentiment analysis is employed to gain insights into the users’ experiences with the app. Many studies use Latent Dirichlet Allocation (LDA) for topic modeling. However, LDA requires users to specify the number of topics to extract before the extraction process. In contrast, BERTopic identifies all potential topics and allows users to switch the clustering algorithm to determine the appropriate number of topics needed. Moreover, BERTopic is more adept at understanding and working with multiple languages than LDA. Regardless of the topic modeling method, the outcome is a collection of high-frequency words reflecting the main ideas. Topic modeling does not assign interpretable names to each set of keywords; instead, it is up to the researcher to label them. This process of manually naming topics involves extensively reading large volumes of text to grasp the contextual meanings of words, which can be both time-consuming and tedious, particularly when dealing with substantial amounts of text. Additionally, the differences among some topics may be quite subtle, making it challenging to identify them based solely on high-frequency words. Researchers need a deep understanding of the text content to make informed judgments, further complicating the task of naming topics.
Given the current research gap regarding electronic payment apps in Taiwan, particularly those offered by specialized electronic payment institutions, this study aims to propose solutions to the following research questions:
  • RQ1: What factors influence user satisfaction in app reviews for specialized electronic payment institutions in Taiwan?
  • RQ2: How do the factors influence user satisfaction and dissatisfaction, and how do they differ?
This study analyzes user reviews of electronic payment apps, as user-generated content (UGC) is highly valuable. This study primarily targets Android phone users, so user app reviews are mainly sourced from Google Play. We employ topic modeling to identify the factors influencing this satisfaction since it effectively extracts insights from user reviews. Given its flexibility in identifying and learning latent topics and its ability to understand multiple languages, we adopt BERTopic for topic modeling. Additionally, the study tackles the challenges of manually naming these topics using a large language model powered by generative artificial intelligence to assist in assigning names to the keyword sets identified through our analysis. In summary, this study makes several contributions to the knowledge in the field of capturing user experience and satisfaction. Firstly, by utilizing the strengths of BERTopic in processing short text, we can extract more detailed and insightful topics from users’ feedback comments. This is combined with sentiment analysis to accurately pinpoint users’ satisfaction and dissatisfaction concerning specific functions or aspects of the service. By merging these approaches, we can move beyond simple emotional assessments and gain a deeper understanding of the underlying reasons and specific needs that drive users’ emotions. Secondly, this research offers data-driven and precise identification of pain points and highlights within electronic payment services. We can not only identify the primary issues that users are most dissatisfied with or satisfied with, but we can also quantify the intensity of these emotions and their connection to specific topics. This information allows application developers to prioritize which functions need improvement, such as enhancing the convenience of account access and minimizing disruptions from system updates. Finally, by analyzing actual user review data, this study addresses the limitations of traditional survey methods, which often rely on small sample sizes and lack extensive user feedback. This approach enables us to acquire more comprehensive insights that reflect real usage scenarios, providing both academic and practical foundations for the ongoing enhancement of electronic payment applications, ultimately leading to improved user satisfaction and loyalty.
In Section 2, we introduce the techniques of topic modeling and sentiment analysis, while also reviewing relevant studies that focus on analyzing user reviews of mobile apps. Additionally, we provide background information on electronic payments in Taiwan that is essential to this study. Section 3 describes the proposed research method and process in detail. In Section 4, we present our findings and discuss their theoretical and practical contributions, managerial implications, and notable research limitations. Finally, the Conclusion summarizes the findings and suggests possible directions for future research.

2. Related Works

2.1. Topic Modeling

Topic modeling is a natural language processing (NLP) technology that utilizes pattern recognition and machine learning to identify topics within documents. Its main principle is to analyze word patterns and frequencies across a collection of texts to create a list of topics or topic clusters. This process enables the grouping of documents that share similar themes. Topic modeling is a valuable method for rapidly analyzing large volumes of documents. One common method used for topic modeling is Latent Dirichlet Allocation (LDA) [22], a probabilistic model based on the assumption that each word in a document can be associated with a specific topic. LDA treats a document as a collection of words and assumes that it contains a random mixture of latent topics. To identify the topics within a document, LDA calculates the probability of specific words being associated with each topic and analyzes how these words are distributed throughout the document. While LDA is intuitive and computationally efficient, determining the appropriate number of topics can be challenging. The choice of the number of topics has a significant impact on the analysis results. Currently, two main methods can assist researchers in finding the ideal number of topics: perplexity-based measures and coherence-based measures [23].
Some studies have indicated that the topics generated by Latent Dirichlet Allocation (LDA) may be too general or even irrelevant to the documents [24]. This shows that LDA’s ability to understand semantics is still insufficient. In response, Grootendorst [25] proposed a topic modeling method based on the Transformer model called BERTopic. This method leverages a powerful language model, making BERTopic superior to LDA in terms of semantic understanding. The core idea behind BERTopic is to utilize a pre-trained BERT model to encode text and extract topics through unsupervised learning. The process of topic extraction in BERTopic consists of several steps. First, a pre-trained model embeds each document in a corpus into high-dimensional vectors. Next, the UMAP algorithm is applied to reduce the dimensionality of these vectors. Following that, HDBSCAN is used for clustering the reduced-dimensional vectors. Finally, c-TF-IDF is employed to extract key terms and condense the number of topics, resulting in the final topic representation.
Compared to traditional bag-of-words methods like LDA, BERTopic does not require pre-defining the number of topics. Additionally, it captures semantic relationships among words in the corpus more effectively, which means it typically does not need extensive text preprocessing. BERTopic also demonstrates improved scalability when dealing with larger corpora [24]. Studies have shown that when applied to short texts, BERTopic can yield clearer topics and provide more novel insights than LDA [26]. Given that user reviews for apps are typically brief, along with the various advantages of BERTopic outlined above, this study will adopt BERTopic as the topic modeling method.

2.2. Sentiment Analysis

Numerous studies on users’ online reviews have highlighted the significance of sentiment analysis [27]. The emotions users express in these reviews reveal their perspectives on the product and can significantly influence the purchasing intentions of other potential customers. Additionally, understanding the factors that lead to positive or negative emotions can help companies grasp what matters most to their customers. This knowledge enables companies to focus on improving critical aspects of their service, ultimately enhancing service quality and increasing customer satisfaction [28,29]. Sentiment analysis can yield either discrete emotions or sentiment polarity [28]. The concept of discrete emotions was first introduced by psychologist Paul Ekman, who identified six basic emotional states: happiness, anger, sadness, disgust, surprise, and fear. These emotions are commonly used to classify facial expressions [30,31,32]. Sentiment polarity, on the other hand, refers to the direction or tendency of emotions and is typically categorized into three types: positive, neutral, and negative. Sentiment polarity is often combined with sentiment intensity to create a sentiment score, which serves as a quantitative indicator of human opinions and emotions [33]. Research investigating user satisfaction and behavioral intentions through online user reviews typically employs topic modeling to identify influencing factors and utilizes sentiment scores or user ratings to measure user satisfaction and behavioral intentions [21,27,28,29,34,35,36].
Many sentiment analysis tools available in the market primarily focus on English text, but the user reviews analyzed in this study are mainly written in traditional Chinese. Therefore, this study employs KeyMoji as the sentiment analysis tool, as it is capable of processing Chinese comments. In addition to its ability to handle Chinese, KeyMoji offers several unique features. First, when calculating emotions, KeyMoji considers the cognitive processes involved in human understanding and language use, along with the concept of semantic scope in formal semantics. This enables its emotional calculations to align closely with human cognition. Additionally, KeyMoji provides an option to adjust the sensitivity of emotion calculations to context [37].

2.3. Analysis of User Reviews of Mobile Payment Apps

Mobile payment is a service that allows consumers to make transactions using their mobile devices. This service can be enabled through various technologies, such as Near Field Communication (NFC) and QR codes [38]. Accordingly, electronic payment can also be regarded as a kind of mobile payment. With the increasing use of smartphones, consumers can easily transfer funds to purchase goods and services or pay bills directly from their devices. Several studies have investigated consumer behavioral intentions regarding mobile payments. These studies typically employed theoretical models to develop questionnaires distributed in surveys to identify factors influencing user satisfaction and the intention to adopt mobile payment systems. However, most of these studies focused on initial adoption and usage, with less attention given to post-adoption behaviors and ongoing use. While understanding users’ pre-adoption intentions is important for initial market expansion, it is crucial for mobile service providers to focus on retaining existing customers, as this directly impacts profitability. Therefore, analyzing customer satisfaction after adoption can help mobile service providers understand the possibility of converting existing customers into loyal users [39]. Recently, many mobile service providers have developed their own mobile applications to deliver services. Users can download these apps from either the Google Play Store or the Apple App Store, depending on their device’s operating system. Both stores allow users to rate and review apps, which has become a valuable resource for mobile service providers to gain insights into users’ opinions and experiences with their services [40]. As a result, several studies have performed text mining on app user reviews to assess user satisfaction and behavioral intentions after adoption. Some of these studies focus on voice assistant apps [21], some on Metaverse platforms [29], some on health care [27], and others on banking apps [35,41,42,43]. Some recent studies have also focused on mobile payment applications [34,44,45].
Verkijika and Neneh [45] argue that existing research primarily focuses on the intention to recommend mobile payment applications while overlooking the actual recommendation behavior. To address this gap, they examined the factors that influence users’ recommendations of mobile payment apps by analyzing user reviews. They collected reviews from 16 mobile payment apps available on the Google Play Store, specifically targeting those who explicitly recommend the app or express an intention to recommend it. Following the methodology of Lang et al. [46], they categorized the reviews into three groups: positive, negative, and neutral, based on user ratings. They then applied LDA for text analysis on the positive and negative reviews. The LDA technique extracted 10 sets of keywords from positive reviews and 5 sets from negative ones. Two independent researchers reviewed these keywords and manually generated interpretable topics to explain why users offer positive or negative recommendations for mobile payment systems. Yi et al. [34] analyzed user reviews using LDA for nine mobile payment applications on the Apple App Store. They identified four sets of keywords and assigned topic names to each group, which were used as key quality factors for the mobile payment apps. They then performed sentence-level sentiment analysis to evaluate users’ perceptions of each quality factor. Finally, they examined the download rankings of the apps as a measure of their market performance and employed panel regression models to investigate the relationship between user evaluations of these quality factors and the market performance of the apps. Perea-Khalifi et al. [45] analyzed three domestic P2P payment apps in Spain: Bizum, Twyp, and Verse. The payment service providers for these apps differ in nature. Bizum operates as a service integrated into the apps of several banks, meaning it is not a standalone application. Users must first download their bank’s app and then enable the Bizum service within that app. In contrast, Twyp and Verse are independent applications. Twyp is offered by a single bank, while Verse is provided by a private company. The researchers aimed to investigate whether the differing nature of these payment service providers influences user perceptions. To do this, they first analyzed variations in word frequency and relevance in user reviews for the three apps. They subsequently examined the emotions expressed in the reviews based on Plutchik’s seven-emotion model and assessed the frequency of these seven emotions within the user reviews. Finally, they applied LDA to extract six keyword groups from the reviews. They named the topics of these groups based on the existing literature and analyzed the distribution of document probabilities for each topic and their relevance to the seven emotions. In summary, studies that utilize text mining to analyze app user reviews employ topic modeling to identify the subjects that users are concerned about, sentiment analysis to gauge users’ emotions, and a regression model to examine the relationship among the topics and emotions expressed.

2.4. Electronic Payment Institutions in Taiwan

An electronic payment institution in Taiwan is defined as an organization that has obtained a license from the Financial Supervisory Commission (FSC) per the Act Governing Electronic Payment Institutions [47]. These institutions are authorized to conduct various licensed business activities overseen by the FSC. In other words, an institution that has not received approval from the government authority as an electronic payment institution is prohibited from conducting the business items specified in the regulations. It can only facilitate payments on behalf of others. Currently, nine specialized electronic payment institutions exist in Taiwan. Their names and the corresponding number of users are listed in Table 1 [9], ranked in descending order by the number of users. Each of these nine institutions has developed its own electronic payment app, which consumers need to download onto their mobile phones to access the electronic payment services offered. Chen et al. employed structural equation modeling to explore the factors that influence consumer behavior in the use of bank mobile payment apps in Taiwan. However, there has yet to be a study examining user satisfaction with electronic payments in Taiwan. This research focuses on specialized electronic payment institutions in Taiwan and utilizes text mining to analyze user reviews of their mobile applications. As indicated in Table 1, users of the top four companies account for 80% of all users. Therefore, this study will concentrate on the mobile apps of these four companies. Table 2 lists the names and features of the apps offered by the four institutions. The first three features related to electronic payments are facilitated through QR codes. All four apps allow users to pay for daily living expenses and accumulate and redeem reward points. Except for PxPay Plus, the other three apps can generate QR codes for public transportation rides. Additionally, iPass Money and EasyPay can be linked to Taiwan’s two major stored-value cards: iPass and EasyCard.

3. Methodology

Figure 1 illustrates the research methodology and the steps involved in this study. As mentioned in the Section 1, this study primarily targets Android phone users. Android applications are downloaded from the Google Play Store, so user reviews for this study were gathered from there. The reviews were then preprocessed to eliminate those that could distort the analysis results. To identify the topics discussed in the user reviews, this study employed BERTopic. Sentiment scores derived from the reviews served as a quantitative measure of user satisfaction. Probability values for each topic across all reviews were treated as the independent variable, while user satisfaction was considered the dependent variable. Multinomial logistic regression was utilized to identify the key factors influencing satisfaction (RQ1). Additionally, a correspondence analysis was conducted to examine the differences in how these factors affect user satisfaction and to identify which factors contribute the most and least to satisfaction (RQ2).

3.1. Data Collection and Sentiment Analysis

We gathered user reviews of apps from the Google Play Store for four specialized electronic payment institutions: iPASS Corporation, Jkopay Co., Ltd, PXPay Plus Co., Ltd., and EASYCARD CORPORATION. The timeframe for collecting reviews in this study covers a recent period, spanning from September 2018 to February 2025. In the next data preprocessing step, we needed to filter out reviews where the user ratings are inconsistent with the review sentiment. Therefore, we first needed to calculate the sentiment score for each review.

3.2. Data Preprocessing

Initially, 12,614 reviews were collected, and after filtering, 10,669 reviews remained. It is important to note that overly short reviews can negatively impact the accuracy of topic modeling [41]. Therefore, we first removed reviews with fewer than 10 words. Additionally, we found inconsistencies in user ratings; some users gave high ratings to apps while expressing negative sentiments in their reviews, and vice versa. The gap between user ratings and sentiment may also result from a misjudgment of sentiment analysis. This inconsistency could adversely affect our analysis in subsequent steps, so we also decided to remove these reviews. Reviews with sentiment scores above the average positive score were classified as “strong positive”, while those exceeding the average negative score were classified as “strong negative”. Then, we removed the reviews with the lowest user ratings that expressed strong positive sentiments and those with the highest user ratings that conveyed strong negative sentiments.
Some examples of removed reviews are listed below:
  • One review stated, “I was unable to log into my account; it kept spinning in circles, then got stuck and crashed. The same issue occurred after I uninstalled and reinstalled the app”. This review clearly expressed dissatisfaction, and the sentiment analysis accurately judged it strongly negative. However, the user rated it 5 stars.
  • Another review mentioned, “Saving paper money is both environmentally friendly and convenient; it considers multiple factors and keeps pace with modern times”. This review conveyed a positive sentiment, and the sentiment analysis assessed it as strongly positive, but the user rated it 1 star.
  • A different review noted, “After I installed the app and opened it, it indicated that an update was needed. When I clicked on the update, the app crashed”. Although this review expressed dissatisfaction, the sentiment analysis incorrectly categorized it as strongly positive.
  • Lastly, one review stated, “The previous login issue has been completely resolved!” While this review offered a positive evaluation, the sentiment analysis mistakenly interpreted it as negative.
Table 3 displays the initial and final number of reviews for each app, along with details on the number of reviews removed under different filtering conditions. We used the total number of reviews for each app as a baseline to calculate the relative frequency of each rating. Figure 2 is a score-based bar chart that illustrates the relative frequencies of ratings for each app. It is clear that the number of 1-star reviews exceeds that of other ratings, indicating that most users tend to express their dissatisfaction. Figure 3 shows an app-based bar chart that displays the distribution of user ratings (from 1 to 5) for each app. The relative number of low-rated reviews appears to be fairly consistent across all apps.

3.3. Topic Analysis

Topic modeling is an effective method for automatically extracting themes from a large volume of unstructured review text. Various studies have widely used this technique to investigate the factors influencing user satisfaction [21]. This study employed BERTopic for topic modeling on app user reviews, using the embedding model “BAAI/bge-large-zh-v1.5” [48], designed to handle Chinese documents. Moreover, the number of topics was set to 10, which is neither too small to create ambiguous topics nor too large to lead to overfitting [49]. This study analyzes 10 topics, yielding a coherence score of C_v = 0.6261. Based on experience, C_v scores typically range from 0.3 to 0.6 [50,51], indicating that the selection of 10 topics in this study is acceptable. The result of BERTopic consists of clusters of keywords. Researchers have traditionally manually reviewed the content of each keyword group to assign topic names. In this study, we leverage ChatGPT Plus to assist in the labeling process, as its underlying generative large language model has demonstrated strong language understanding capabilities. We prompted the clusters of keywords to ChatGPT Plus and requested it to derive a generalized concept from each set and assign an appropriate topic name. After manually examining the names generated by ChatGPT Plus, we found that they accurately reflected the meanings of the corresponding keyword sets. Consequently, we decided to adopt the topic names produced by ChatGPT Plus. To assess the effectiveness of generating topic labels using ChatGPT Plus, we invited a domain expert to review the keywords and reviews and subsequently annotate the topic labels. Afterwards, we employed two verification methods to determine whether the topic labels generated by ChatGPT Plus aligned with the manual annotations. The first method was manual verification. We invited two raters (Rater-1 and Rater-2) to evaluate the semantic similarity between the labels produced by the domain expert and those generated by ChatGPT Plus. The raters used a scoring system from 1 to 10, where 1 indicates very low semantic similarity and 10 signifies high semantic similarity. To assess inter-rater reliability, we applied Spearman’s rank correlation coefficient for statistical analysis. The results indicate that the average scores for Rater-1 and Rater-2 were 8.3 and 8.4, respectively, with a Spearman’s rank correlation coefficient of 0.727, demonstrating a high positive correlation. The two-tailed significance test yielded a p-value of 0.017, suggesting significant consistency between the two raters at a significance level of α = 0.05. The second method involved language model verification. We utilized the Sentence Transformer model (with the embedding model intfloat/multilingual-e5-large) to convert each set of topic labels generated by the domain experts and ChatGPT Plus into semantic vectors, from which we calculated the cosine similarity. The results revealed that the similarity scores for each set of topic labels were all above 0.9, with an overall average similarity of 0.942. This indicates that the semantic meanings of the labels generated by both humans and ChatGPT Plus are highly consistent. In summary, whether evaluated through manual scoring or semantic vector analysis, the topic labels produced by ChatGPT Plus exhibit a high degree of semantic alignment with manual annotations. This suggests that ChatGPT Plus is highly effective and has significant potential for practical applications in topic tagging tasks.
Then, we examined the relationships among various user satisfaction measures through regression analysis. The sentiment score, originally a continuous value ranging from −1 to 1, was categorized into a discrete five-point sentiment scale, which served as the dependent variable. The independent variables consisted of the probabilities of ten topics, indicating the likelihood that a review is related to a specific topic. Since the dependent variable is discrete and has more than two categories, we employed multinomial logistic regression to assess the significance of each topic on user satisfaction. Additionally, multinomial logistic regression was chosen because it does not require the variables to be normally distributed or to have a linear relationship with one another.

3.4. Satisfaction and Dissatisfaction Analysis

After identifying the factors influencing user satisfaction, this study employed correspondence analysis to examine the relationship among these factors and user sentiments, as demonstrated in several previous studies [21,29]. Correspondence analysis is a statistical method that explores the relationship between two categorical variables without assuming a causal relationship between them. The results of this analysis are presented as a visualization of a contingency table for the two categorical variables. This graphical representation aids in understanding which factors contribute to positive user satisfaction and which may lead to negative experiences. In this study, a sentiment scale rating of ‘1’ indicates the most negative user satisfaction, while a ‘5’ represents the most positive user satisfaction.

4. Results and Discussion

4.1. Factors Impacting User Satisfaction

The second column of Table 4 presents the 10 sets of keywords extracted by BERTopic. The original user reviews are in Chinese. Table 4 presents the results of translating Chinese keywords and topic names into English for topic modeling. The original language results can be found in Appendix A. BERTopic does not produce meaningful topic names for each set of keywords. In most studies, researchers manually review the keyword content and related reviews to assign topic names. In this study, we used ChatGPT Plus to suggest appropriate topic names. The third column of Table 4 displays the topic names generated by ChatGPT Plus, translated into English. While English translations are provided, the meanings of these topics may not be easily understood by individuals lacking local background knowledge. To address this, we offer interpretations of these topics in Table 5.
This study employs a multinomial logistic model to assess the influence of various topics on user satisfaction. Table 6 presents the goodness-of-fit statistics, illustrating how well the model aligns with the data. A statistically significant p-value for the Pearson chi-square statistic (i.e., p < 0.05) indicates that the model does not fit the data well. In this study, Table 6 reports a p-value of 0.997, which shows that our model fits the data effectively. Table 7 provides the model fitting information, which assesses whether the inclusion of the independent variables significantly improves the model. The p-value reported in Table 7 is close to 0, signifying that the full model with the 10 topics predicts the dependent variable significantly better than a model without any variables. Additionally, Table 8 features a likelihood ratio test that evaluates the overall effect of the nominal variables to identify which independent variables are statistically significant. From Table 8, it is clear that Topic 6 has a p-value of 0.298, whereas the p-values for the other topics are all below 0.05. Based on the current model testing results, we cannot provide sufficient statistical evidence to support the significant impact of Topic 6 on satisfaction. In summary, apart from Topic 6, the other topics are important factors influencing user satisfaction. Topic 6 refers to users making life payments through electronic payment apps. However, in Taiwan, people can also pay their living expenses at convenience stores. According to the Ministry of Economic Affairs’ statistics, the number of convenience stores in Taiwan has been consistently increasing, reaching 13,706 by the end of December 2023. This translates to an average of one convenience store for every 1703 people, giving Taiwan the second-highest per capita convenience store density in the world, just behind South Korea [52]. In other words, paying living expenses at convenience stores is highly convenient in Taiwan. Electronic payment methods are merely an alternative for fulfilling these expenses. This could explain why Topic 6 has no significant impact on overall satisfaction.
Table 9 shows a comparison of Categories 1 to 4 with the reference category (scale 5). According to the results of multiple logistic regression analysis, each topic has varying degrees and directions of influence on the user’s emotional score, which ranges from 1 to 5 points:
-
Topic 0 had a significant positive impact on Category 2 (β = 3.618, p < 0.001; odds ratio = 37.27). This means that when users’ reviews include this topic, the probability of them expressing an emotion rated at 2 points increased significantly. This indicates that it is a strong negative topic.
-
Topic 1 significantly reduced the likelihood of users expressing an emotion rated at 1 point (β = −5.317, p < 0.01; odds ratio = 0.005) and 3 points (β = −8.207, p < 0.001; odds ratio ≈ 0). This suggests that this topic is associated with high satisfaction and has a positive influence.
-
The influence of Topic 2 is inconsistent; it significantly reduces the probability of users showing a 3-point emotion (β = −8.769, p < 0.01), while significantly increasing the probability of showing a 4-point emotion (β = 5.093, p < 0.05). This indicates that it has a mixed effect on satisfaction evaluation and can be regarded as a neutral-to-positive topic.
-
Topic 3 has a significant positive impact in all categories, with an extremely high odds ratio (up to 9.2 × 1015), indicating a strong correlation with unsatisfactory evaluations and categorizing it as a significant and strong negative topic.
-
The coefficients of **Topic 4** are negative and significant across all categories, with an odds ratio close to 0, meaning that its presence almost excludes low emotion values, making it an extremely positive topic.
-
Topic 5 significantly increases the odds of users expressing emotions of 1 point (odds ratio = 14,489.137, p < 0.001) and 3 points (odds ratio = 4.06 × 105, p < 0.001), indicating it is a strong negative topic.
-
Topic 6 significantly increases the odds of showing a 1-point emotion (β = 2.754, p < 0.05; odds ratio = 15.702). The other categories show similar but non-significant trends, suggesting it is a potentially negative topic.
-
Topic 7 has a significant positive impact only in Category 3 (β = 4.332, p < 0.01; odds ratio = 76.07), with the other categories showing no significant impact, classifying it as a moderately negative topic.
-
Topic 8 significantly increases the probabilities of 1–3 points (odds ratio = 3023.205, 384.615, 460.681, p < 0.001, 0.01, 0.05), which is indicative of a highly negative topic.
-
Topic 9 significantly reduces the probabilities of 1 and 2 points (β = −19.548, −6.957, both p < 0.001; odds ratio ≈ 0), indicating a strong correlation with high satisfaction and categorizing it as a positive topic.
Based on these results, Topics 3, 5, and 8 exhibit the strongest negative effects, while Topics 4 and 9 demonstrate highly positive effects.

4.2. Correspondence Analysis

Figure 4 illustrates the correspondence analysis of the contingency table (Table 10) between the sentiment scale (ranging from 1 to 5 points) and the discussed topics (labeled t0 to t9). Points that are closer together have a higher likelihood of appearing in reviews simultaneously, indicating a stronger relevance between those topics. From the visual representation in Figure 4, it is clear that “enhancement of account access mechanisms” (t1) and “application installation and updates” (t3) are the two topics closest to each other. This proximity implies that these topics are often mentioned together, which makes sense—after installing or updating an application, users typically need to register or log in. Additionally, the distance of each point from the origin (0,0) signifies distinctiveness. A point that is farther from the origin indicates a greater contribution to the overall corresponding structure of the data. In Figure 4, “convenient payment for various living expenses” (t6) and “optimization of balance inquiry features” (t7) are the closest points to the origin, suggesting that these topics contribute less to explaining the overall structure. Furthermore, the topics and sentiment scores in the same quadrant are positively associated. If the angle connecting the topic and sentiment score to the origin is sharper, it indicates a stronger relationship between them. Figure 4 reveals that “perceived usability and promotional engagement” (t4) is strongly positively associated with a sentiment score of 5. This highlights that users express extremely high satisfaction when discussing this topic, meaning consumers are very pleased with the convenience and promotions provided by electronic payment services. The keyword set associated with Topic 4 includes terms such as “convenient”, “easy to use”, and “practical”, along with several keywords highlighting “many discounts”. Reviews related to this topic reveal that many users appreciate its convenience. For instance, one user stated, “It is very convenient to use overall. You can take a ride, pay, and top up with one device if you have an Internet connection”. Another user mentioned, “Many payment application channels are available, and withdrawals are instant. It is very convenient, and I would recommend it”. Moreover, users frequently expressed their appreciation for promotional activities in their reviews. One user noted, “There are often promotional activities for account payments, which is great”. Another shared, “It is convenient to withdraw deposits, and there are many promotional activities for consumption, which is fantastic!” In conclusion, the ease of use of electronic payment systems is a crucial factor in user satisfaction. Additionally, incorporating more promotional activities could further enhance this satisfaction. The topic “diverse payment channels and loyalty point systems” (t2) shows a strong positive association with a sentiment score of 4. This indicates that electronic payment services that offer reward points also enhance overall user satisfaction. Although “multiple convenient payment” (t6) also falls within the same quadrant as sentiment score 4 and the angle between them and the origin is small, it is important to note that t6 is very close to the origin, indicating low distinctiveness for this topic.
On the other hand, “enhancement of account access mechanisms” (t1) and “application installation and updates” (t3) both have a strong positive association with a sentiment score of 1, while “app stability and crash fixes recovery” (t9) has a strong positive association with a sentiment score of 2. The commonality among these three topics is their relation to the use of electronic payment applications, indicating that users are experiencing significant frustration with the use of electronic payment applications. Topic 1 includes several keywords, such as “login”, “register”, and “password”, which are related to account access. Notably, the word “terrible” also appears in several keywords related to this topic, indicating that users are unhappy with their experience regarding account access. Examining reviews related to Topic 1, several users expressed their frustrations: “It’s really bad; the design is very unsatisfactory. The system keeps prompting me that I’ve registered, but there is no membership center for me to re-register or continue. Every time, I get stuck at the point where I’ve already activated my membership, but when I enter my email, it says there’s no membership information. It’s truly terrible. I’ve never used such a frustrating and ineffective system”. Another user noted, “There are many restrictions on member registration and the naming rules of accounts and passwords, which makes it inconvenient for users. The accounts and passwords I usually use do not comply with the rules. Do you think it’s safe to have so many regulations? Each company has its own set of rules, but how can users remember so many different accounts and passwords?” A third review mentioned, “I’ve activated fingerprint recognition, but it frequently prompts me to enter a four-digit password. This is unnecessary and a waste of time. The entire process is extremely inconvenient. In a word, it’s extremely bad”. For Topic 3, keywords related to application installation and updates include “cannot update”, “cannot be used”, and similar to Topic 1, it also features several instances of “terrible”. Users shared their experiences, saying, “After installing it, the app keeps spinning in circles when I try to access the homepage. No wonder the rating is only 2”. Another user commented, “It gets worse with each update. Not only is the transaction record slow to update, but it is also incorrect. Some ride records still haven’t been posted after two days. This was not the case before”. The keywords and comments related to Topics 1 and 3 clearly demonstrate that poor user experience on electronic payment mobile apps can lead to negative impressions among users.

4.3. Theoretical and Practical Contributions and Managerial Implications

This study makes several theoretical contributions. First, it employs text mining to analyze user reviews of Taiwan’s electronic payment apps, identifying factors that influence user satisfaction. Traditional questionnaire surveys, which rely on established theoretical models, limit the range of factors that can be explored. In contrast, using topic modeling to extract insights from user reviews allows for a broader understanding of users’ feelings without being limited to pre-determined factors. Secondly, the pre-processing of user reviews in this study differs from approaches taken in other research. Many studies simply discard short comments; however, this study recognizes that user ratings may sometimes contradict the feelings expressed in their reviews. Such inconsistencies can skew the analysis, so this study also excludes these conflicting reviews during the pre-processing stage. Finally, this study utilizes BERTopic for topic modeling, whereas most other studies typically tend to utilize LDA. Moreover, this study incorporates generative AI technology based on large language models to automatically interpret keyword clusters and assign relevant topic labels—an approach not commonly seen in analyses of user reviews. This study utilized ChatGPT Plus to help label each keyword set after topic modeling, demonstrating the potential to lessen the human burden in traditional naming, which typically relies on manual keyword set reviews.
The findings of this study offer practical implications for electronic payment providers in Taiwan. First, it is clear that convenience and ease of use are the primary concerns for users, which aligns with previous studies on user reviews of payment and banking apps [34,36,41,45]. The study reveals that users are most satisfied when electronic payment platforms integrate various services and offer discounts. Therefore, service providers should focus on developing innovative features that enhance user convenience and frequently provide appealing discounts to increase user engagement. Secondly, users have expressed significant dissatisfaction with the usability of mobile applications and the system instability caused by updates. Electronic payment service providers need to take decisive action to address these two issues in order to improve user satisfaction. The most effective action plan involves tackling the problems related to account access, as well as the installation and updating of applications from a technical perspective. However, genuine enhancement of user satisfaction must also come from the management level. It is crucial to establish a cross-departmental team and utilize project-management tools to ensure that all user issues, defect reports, and feature suggestions can be recorded, tracked, and assigned to the appropriate team member. Additionally, a unified feedback tracking system should be implemented, supported by regular seminars. During these seminars, team members can discuss the most pressing user pain points and brainstorm innovative solutions. The progress of these initiatives, along with key performance indicators (KPIs) and any changes in user satisfaction after implementing solutions, should be continuously monitored and optimized. Moreover, the company’s marketing and public relations department should inform users of upcoming updates through push notifications and announcements. They should also publish guidelines on how to address potential issues and establish a dedicated customer service channel on social media. This approach will enable users to receive timely assistance if they encounter problems with account access or application updates. Finally, electronic payment service providers can prioritize their improvement projects based on the findings of this study. If their primary goal is to tackle the main sources of user dissatisfaction, providers should first focus on enhancing the stability of their mobile apps. Alternatively, if their primary aim is to retain existing customers, they should consider integrating electronic payments with other services, enabling users to accomplish multiple tasks on a single platform. The more convenient electronic payment services make life for their users, the more likely those users are to remain loyal to the services.

4.4. Potential Cultural or Societal Aspects Unique to Taiwan

This study analyzes user reviews of electronic payment applications to understand user satisfaction with these services. In Taiwan, some unique cultural and social factors may also shape users’ attitudes and behaviors towards electronic payment services.
First, the high levels of digital literacy and mobile device penetration among the Taiwanese have facilitated the widespread use of electronic payment systems. According to the 2020 Communications Market Survey Results Report by the Taiwan Institute of Economic Research, the household smartphone ownership rate among individuals aged 16 and above is 96.8% [53]. Additionally, the government continues to strengthen the development of the digital economy. The “Digital Nation and Innovative Economic Development Plan” (DIGI+) promotes solutions in application fields and amends relevant regulations to build an ecosystem that supports mobile payments [54].
Second, Taiwanese consumers prefer free or discounted offers such as free trials and gifts. Electronic payment services in Taiwan are often free, and service providers frequently collaborate with merchants or banks to offer additional discounts to encourage the use of electronic payments. This strategy has effectively increased the usage rates of these services. Liang and Li confirm that price value has replaced perceived value as a defining feature of Taiwan’s mobile payment market [55].
Third, Taiwan has long operated a system where prizes are awarded through invoices, which encourages people to request invoices and boosts government tax revenue. This unique system is closely linked to electronic payments. Consumers can apply for mobile barcodes via the Ministry of Finance’s electronic invoice integration service platform or through electronic payment apps. This integration allows consumers to easily store invoices, make payments, and automatically check and claim prizes using the electronic payment app, thereby promoting the habit of using electronic payments among Taiwanese consumers.

4.5. Platform Bias and Generalizability

This study analyzed user reviews of electronic payment applications from the Google Play Store to perform topic modeling and sentiment analysis, providing insights into the satisfaction of Android users with these services. However, previous research indicates that Android and iOS users differ in demographic characteristics, application preferences, and technical expectations. As a result, excluding iOS users from this study may overlook important differences in experiences and expectations. According to a 2024 survey by the US e-commerce platform Slickdeals, the average annual income of iPhone users is USD 53,251, while that of Android users is USD 37,040 [56,57]. This income disparity may make discounts and reward points offered through electronic payments less appealing to iOS users. Additionally, there are also notable differences in application preferences between the two groups. Android users tend to favor applications that are practical, high-performing, multimedia-oriented, and productivity-related. They value high customization and system openness, often personalizing themes, desktop layouts, and widgets to suit their tastes. In contrast, iOS users prioritize high-quality applications and a stable user experience [58], favoring simplicity, intuitive operation, and system consistency [59]. Due to the stringent review process of the App Store, iOS applications typically excel in design optimization and stability, which aligns with iOS users’ expectations for quality. Consequently, this study found that Android users are extremely dissatisfied with application usability, while iOS user reviews likely do not convey such strong negative sentiments regarding usability. Furthermore, iOS users have higher expectations for security and privacy protection, which are addressed by Apple’s rigorous application review process. Although Android users also prioritize security, they are generally more willing to accept certain risks associated with open platforms in exchange for greater freedom of use [60]. It is worth noting that the topic analysis conducted in this study did not capture aspects related to security and privacy; however, such topics may be evident in iOS user reviews.

4.6. Methodological Limitations in Text Mining

This study employs text mining techniques, specifically sentiment analysis and BERTopic-based topic modeling, to extract topics and assess sentiment polarity from user-generated comments. However, text mining technologies have inherent limitations that may affect the accuracy of the research results.
First, the informal language commonly found in online user reviews—characterized by spelling errors, slang, emoticons, and a mix of Chinese and English—can hinder the accuracy of sentiment classification and topic extraction [61]. Research has shown that employing deep-learning models alongside preprocessing methods, such as slang dictionaries, can enhance accuracy [62,63]. However, cultural differences make slang dictionaries non-universal. To accurately capture idiomatic expressions or culturally specific phrases (e.g., local Taiwanese internet slang), fine-tuning with localized corpora is still necessary.
Second, detecting sarcasm and irony, often present in user complaints or humorous posts, is particularly challenging for machines. For instance, a comment like “Great, it crashed again!” may be mistakenly interpreted as positive due to the word “great”, leading to skewed interpretations [64].
Third, the interpretability of topics heavily relies on manual annotation, which introduces subjectivity, even with the aid of large language models such as ChatGPT Plus. Although we have validated these labels through expert reviews and inter-coder reliability checks, challenges such as semantic overlap or ambiguity among topics remain.
Finally, BERTopic relies on text embedding through Transformer models like BERT. These embedding models have specialized multilingual versions that can represent semantics across multiple languages. Thus, the multilingual capability of BERTopic depends on whether the pre-trained language models it utilizes support multilingualism and how well these models can represent semantics in various contexts. Choosing an appropriate pre-trained language model that aligns with the language and context of the analyzed user reviews poses a challenge for topic modeling tasks involving multilingual texts.
Future research may explore integrating hybrid approaches, such as human-computer interaction verification, supervised sentiment models trained on localized data, or sarcasm-aware NLP techniques, to improve interpretability and robustness.

5. Conclusions

Mobile payments are financial services and must comply with local regulations. The operation of mobile payments can also vary depending on the service provider. Therefore, it is inappropriate to infer user satisfaction with mobile payments in one region based on research findings from another region. To date, there has been no research on the user satisfaction with electronic payments in Taiwan, making the results of this study particularly valuable for Taiwan’s electronic payment institutions. This study collects many user reviews of electronic payment apps for analysis, aiming to understand post-adoption user experiences comprehensively. As Zhou [39] emphasized, analyzing post-adoption user satisfaction is crucial for mobile service providers to devise strategies that convert existing customers into loyal ones. However, some limitations of the study need to be noted. In this study, we opted to capture reviews from only one platform. This choice was based on the fact that application development teams may vary across different mobile operating systems. Therefore, the results of this research are likely more beneficial for the Android app development team than for the iOS team. The volume of user reviews for different apps can vary significantly. Future research should take this discrepancy into account when conducting comparative analyses of different apps and striving to find potential solutions. This study primarily used user sentiments as a measure of user satisfaction. Sentiment scores were used without deeper contextual analysis, which may lead to misinterpretation of user complaints. Future research may benefit from incorporating other relevant variables and integrating them with user sentiments to provide a more holistic understanding of user satisfaction. In addition, the sentiment analysis in this study focuses solely on sentiment polarity and intensity, as these are sufficient for quantifying user satisfaction. Future studies could enhance this analysis by including discrete emotion detection to gain deeper insights into the various emotions expressed in user reviews. Finally, user satisfaction with electronic payment services can be influenced by various factors, including promotional activities, system upgrades, and external events. As a result, satisfaction levels may vary over time. Future research could utilize time series analysis to explore how satisfaction levels and concerns about specific issues change over time in response to particular events or service modifications.

Author Contributions

Conceptualization, S.-F.T. and C.-S.H.; methodology, S.-F.T. and C.-S.H.; software, C.-S.H.; validation, C.-S.H.; formal analysis, S.-F.T. and C.-S.H.; investigation, S.-F.T. and C.-S.H.; data curation, C.-S.H.; writing—original draft preparation, S.-F.T.; writing—review and editing, S.-F.T. and C.-S.H.; visualization, C.-S.H.; project administration, S.-F.T. and C.-S.H. All authors have read and agreed to the published version of the manuscript.

Funding

This research received no external funding.

Data Availability Statement

The original data presented in the study are openly available in FigShare at 10.6084/m9.figshare.28514378.

Conflicts of Interest

The authors declare no conflicts of interest.

Appendix A

Table A1 displays Chinese keyword sets along with their respective topic names resulting from topic analysis.
Table A1. Results from topic modeling in Chinese.
Table A1. Results from topic modeling in Chinese.
Topic No.KeywordsLabel
Topic 0謝謝, nfc, wallet, 悠遊付, app, pay, 更新, 11, Samsung, 嗶乘車, 10, 悠遊卡, 請問, Xperia, Sony電子票證與行動支付無縫整合
Topic 1無法登入, 爛透了, 無法註冊, 忘記密碼, 爛死了, 到底, 有夠爛, 登不進去, 更新後, 忘記帳號, 換了手機, 是怎樣, 無法使用, 11, 一直顯示錯誤帳號存取優化
Topic 2pay, line, px, app, 街口, 付款碼, ipass, 臺灣, 中華民國, pxpay, 街口支付, 取消px, 為什麼沒有街口幣了, 街口幣, 跟line多元支付與點數整合
Topic 3無法更新, 爛死了, play, 應用程式未安裝, 10, 垃圾軟體, 一直要我更新, 爛透了, android, 無法使用, 謝謝, 超爛, nokia, 等待中, 1後應用程式安裝與更新
Topic 4好用, 很方便, 使用方便, 很棒, 很好用, 非常方便, 非常好用, 方便好用, 超方便, 不錯, 實用, 還不錯, 超讚, 太厲害了這個軟體, 優惠多使用便利性與優惠體驗
Topic 5不方便, 有點不方便, 不好用, 很爛, 爛程式, point卡, 不能綁中國信託, 不能綁定信用卡, 不能用信用卡儲值, 中國信託這麼大間, 才發現, 搞毛, 支援的銀行太少, 很不方便, 全支付支付綁定與儲值優化
Topic 6生活繳費, 電費, 水費, 謝謝, 顯示, 繳費, 停車費, 外帶外, 沒辦法繳停車費, 根本沒辦法繳費, 點選繳費, 牌照稅, 電號, 系統搶修查無資料, 要繳停車費多元便捷繳費
Topic 7希望這個app也能夠有這個功能可以靠卡馬上查詢餘額, 有一些app都可以靠卡感應立刻查詢餘額, 儲值, 餘額, 支付工具管理, 好爛, 不好用, 希望可以改善, 加油, 16, 隔天進入app查尋金額, 難怪評價不好, 難道不是應該針對, 高層座領高薪這種問題也無法解決, 難道就無法改善嗎餘額查詢優化
Topic 8從相簿選擇, 按鍵, 超麻煩, 爛爆了, 07, 06, 差評, 爛透了, 是怎樣, 問題是隻是買個海外東西不懂其用意, 問題是我根本沒用過, 問題是提供帳戶之後是可以用此app轉帳的, 阿所以我本人的身分證字號就只能卡在以前已經取消的電話號碼裡, 除非用你們的app來付款後才會有免提領手續費機會變相強迫使用你們app, 重來沒有用過街口的我剛下載想要使用看看直接連驗證碼都拿不到使用者體驗與操作便利性
Topic 9android, 一直閃退, 會閃退, 11, 10, 閃退, 開啟即閃退, 14, 瘋狂閃退, 無法開啟, 還是一樣閃退, 三星, 會一直閃退, 重新下載好幾次了, app一直閃退應用穩定性與崩潰修復

References

  1. Barnard, Y.; Bradley, M.D.; Hodgson, F.; Lloyd, A.D. Learning to use new technologies by older adults: Perceived difficulties, experimentation behaviour and usability. Comput. Hum. Behav. 2013, 29, 1715–1724. [Google Scholar] [CrossRef]
  2. Yagmur, S. A literature review: Usability aspects of ubiquitous computing. In Proceedings of the 2016 International Conference on Platform Technology and Service (PlatCon), Jeju, Republic of Korea, 15–17 February 2016; IEEE: Piscataway, NJ, USA, 2016; pp. 1–6. [Google Scholar] [CrossRef]
  3. Choi, B.; Kim, D. Does the proliferation of smartphones reduce consumer search costs? The case of the Korean gasoline market. Asian Econ. J. 2024, 38, 153–176. [Google Scholar] [CrossRef]
  4. Bezovski, Z. The future of the mobile payment as electronic payment system. Eur. J. Bus. Manag. 2016, 8, 127–132. [Google Scholar]
  5. Kim, S.H.; Park, H.S. The effects of mobile application quality on satisfaction and intention to pay mobile application. J. Inf. Syst. 2011, 20, 81–109. [Google Scholar]
  6. Kuo, T.; Tsai, G.Y.; Lu, I.Y.; Chang, J.S. Relationships among service quality, customer satisfaction and customer loyalty: A case study on mobile shopping APPs. In Proceedings of the 17th Asia Pacific Industrial Engineering and Management System Conference, Taipei, Taiwan, 7–10 December 2016; pp. 7–10. [Google Scholar]
  7. Hammouri, Q.; Almajali, D.A.; Nusairat, N.; Saraireh, S. Determinants of Users’ Satisfaction with Mobile Apps. Int. J. Adv. Sci. Technol. 2020, 29, 14613–14624. [Google Scholar]
  8. Lin, Y.C. New Policies Change Industry Rules, New Competitors Are Eyeing the Market, and the Electronic Payment Elimination Round Begins. 2024. Available online: https://www.credit.com.tw/newcreditonline/Epaper/PersonalityContent.aspx?sn=381&unit=595 (accessed on 24 January 2025).
  9. FSC. Statistics of Credit Card, Cash Card & Electronic Payment Institutions Business Operation as of November 2024; Press Release. 14 January 2025. Available online: https://www.fsc.gov.tw/en/home.jsp?id=54&parentpath=0,2&mcustomize=multimessage_view.jsp&dataserno=202501140004&dtable=News (accessed on 24 January 2025).
  10. Liu, H.-H.; Tsai, D.-R. Discovering the Potentials and Directions of Mobile Payment Applications Development an Exploratory Study in Taiwan. In Proceedings of the 3rd International Conference on Computer Science and Service System, Bangkok, Thailand, 13–15 June 2014; Atlantis Press: Dordrecht, The Netherlands, 2014; pp. 549–552. [Google Scholar] [CrossRef]
  11. Kuo, R.Z. Why do people switch mobile payment service platforms? An empirical study in Taiwan. Technol. Soc. 2020, 62, 101312. [Google Scholar] [CrossRef]
  12. Tsai, J.; Ye, X.; Wang, C. Consumer Behavioral Intension for Using Mobile Payment for E-commerce in Taiwan. J. Internet Technol. 2020, 21, 1821–1828. [Google Scholar]
  13. Davis, F.D. Perceived usefulness, perceived ease of use, and user acceptance of information technology. MIS Q. 1989, 13, 318–340. [Google Scholar] [CrossRef]
  14. Davis, F.D.; Bagozzi, R.P.; Warshaw, P.R. User acceptance of computer technology: A comparison of two theoretical models. Manag. Sci. 1989, 35, 982–1002. [Google Scholar] [CrossRef]
  15. Adams, D.A.; Nelson, R.R.; Todd, P.A. Perceived usefulness, ease of use, and usage of information technology: A replication. MIS Q. 1992, 16, 227–247. [Google Scholar] [CrossRef]
  16. Straub, D.; Limayem, M.; Karahanna, E.E. Measuring system usage: Implications for IS theory testing. Manag. Sci. 1995, 41, 1328–1342. [Google Scholar] [CrossRef]
  17. Van Laarhoven, P.J.; Pedrycz, W. A fuzzy extension of Saaty’s priority theory. Fuzzy Sets Syst. 1983, 11, 229–241. [Google Scholar] [CrossRef]
  18. Ng, W.-K.; Chen, S.; Chen, W.-H.; Chen, C.-L.; Jiang, J.-L. Mobile Payment Innovation Ecosystem and Mechanism: A Case Study of Taiwan’s Servicescapes. J. Theor. Appl. Electron. Commer. Res. 2024, 19, 633–653. [Google Scholar] [CrossRef]
  19. Chen, P.S.; Lee, J.J.; Huang, S.J.; Lin, C.L. A Study on Consumers’ Usage Behavior of the Bank’s Mobile Payment APPs in Taiwan. Bus. Econ. Res. J. 2020, 11, 793–805. [Google Scholar] [CrossRef]
  20. Sheth, J.N.; Newman, B.I.; Gross, B.L. Why we buy what we buy: A theory of consumption values. J. Bus. Res. 1991, 22, 159–170. [Google Scholar] [CrossRef]
  21. Kumar, A.; Bala, P.K.; Chakraborty, S.; Behera, R.K. Exploring antecedents impacting user satisfaction with voice assistant app: A text mining-based analysis on Alexa services. J. Retail. Consum. Serv. 2024, 76, 103586. [Google Scholar]
  22. Blei, D.M.; Ng, A.Y.; Jordan, M.I. Latent dirichlet allocation. J. Mach. Learn. Res. 2003, 3, 993–1022. [Google Scholar]
  23. Newman, D.; Noh, Y.; Talley, E.; Karimi, S.; Baldwin, T. Evaluating Topic Models for Digital Libraries. In Proceedings of the Annual Joint Conference on Digital Libraries, Gold Coast, QLD, Australia, 21–25 June 2010; pp. 215–224. [Google Scholar]
  24. Albanese, N.C. Topic Modeling with LSA, pLSA, LDA, NMF, BERTopic, Top2Vec: A Comparison. 2022. Available online: https://towardsdatascience.com/topic-modeling-with-lsa-plsa-lda-nmf-bertopic-top2vec-a-comparison-5e6ce4b1e4a5/ (accessed on 24 January 2025).
  25. Grootendorst, M. BERTopic: Neural topic modeling with a class-based TF-IDF procedure. arXiv 2022, arXiv:2203.05794. [Google Scholar]
  26. Egger, R.; Yu, J. A topic modeling comparison between lda, nmf, top2vec, and bertopic to demystify twitter posts. Front. Sociol. 2022, 7, 886498. [Google Scholar] [CrossRef]
  27. Pal, S.; Biswas, B.; Gupta, R.; Kumar, A.; Gupta, S. Exploring the factors that affect user experience in mobile-health applications: A text-mining and machine-learning approach. J. Bus. Res. 2023, 156, 113484. [Google Scholar] [CrossRef]
  28. Wu, J.J.; Chang, S.T. Exploring customer sentiment regarding online retail services: A topic-based approach. J. Retail. Consum. Serv. 2020, 55, 102145. [Google Scholar] [CrossRef]
  29. Kumari, V.; Bala, P.K.; Chakraborty, S. A text mining approach to explore factors influencing consumer intention to use metaverse platform services: Insights from online customer reviews. J. Retail. Consum. Serv. 2024, 81, 103967. [Google Scholar] [CrossRef]
  30. Geetha, A.V.; Mala, T.; Priyanka, D.; Uma, E. Multimodal Emotion Recognition with deep learning: Advancements, challenges, and future directions. Inf. Fusion 2024, 105, 102218. [Google Scholar]
  31. Saadi, I.; Taleb-Ahmed, A.; Hadid, A.; El Hillali, Y. Driver’s facial expression recognition: A comprehensive survey. Expert Syst. Appl. 2023, 242, 122784. [Google Scholar] [CrossRef]
  32. Ismail, S.N.M.S.; Aziz, N.A.A.; Ibrahim, S.Z.; Mohamad, M.S. A systematic review of emotion recognition using cardio-based signals. ICT Express 2024, 10, 156–183. [Google Scholar] [CrossRef]
  33. Tian, L.; Lai, C.; Moore, J.D. Polarity and Intensity: The Two Aspects of Sentiment Analysis. arXiv 2018, arXiv:1807.01466. [Google Scholar]
  34. Yi, J.; Kim, J.; Oh, Y.K. Uncovering the quality factors driving the success of mobile payment apps. J. Retail. Consum. Serv. 2024, 77, 103641. [Google Scholar] [CrossRef]
  35. Basu, B.; Sebastian, M.P.; Kar, A.K. What affects the promoting intention of mobile banking services? Insights from mining consumer reviews. J. Retail. Consum. Serv. 2024, 77, 103695. [Google Scholar] [CrossRef]
  36. Luo, J.M.; Hu, Z.; Law, R. Exploring online consumer experiences and experiential emotions offered by travel websites that accept cryptocurrency payments. Int. J. Hosp. Manag. 2024, 119, 103721. [Google Scholar] [CrossRef]
  37. Wang, W.-J.; Chen, C.-J.; Lee, C.-M.; Lai, C.-Y.; Lin, H.-H. KeyMoji; Computer Program; Droidtown Linguistic Tech: Taipei, Taiwan, 2020; Available online: https://api.droidtown.co/ (accessed on 24 January 2025).
  38. Ha, J.; Nam, C.; Kim, S. A systematic review of mobile payment literature: What has been studied and what should be studied? Telecommun. Policy 2024, 48, 102795. [Google Scholar] [CrossRef]
  39. Zhou, T. An empirical examination of users’ post-adoption behaviour of mobile services. Behav. Inf. Technol. 2011, 30, 241–250. [Google Scholar] [CrossRef]
  40. Chatterjee, S. Explaining customer ratings and recommendations by combining qualitative and quantitative user generated contents. Decis. Support Syst. 2019, 119, 14–22. [Google Scholar] [CrossRef]
  41. Alrizq, M.; Alghamdi, A. Customer satisfaction analysis with Saudi Arabia mobile banking apps: A hybrid approach using text mining and predictive learning techniques. Neural Comput. Appl. 2024, 36, 6005–6023. [Google Scholar] [CrossRef]
  42. Balcıoğlu, Y.S. Analyzing Customer Sentiments and Trends in Turkish Mobile Banking Apps: A Text Mining Study. Dumlupınar Üniv. Sos. Bilim. Derg. 2024, 80, 49–69. [Google Scholar] [CrossRef]
  43. Shankar, A.; Tiwari, A.K.; Gupta, M. Sustainable mobile banking application: A text mining approach to explore critical success factors. J. Enterp. Inf. Manag. 2022, 35, 414–428. [Google Scholar] [CrossRef]
  44. Verkijika, S.F.; Neneh, B.N. Standing up for or against: A text-mining study on the recommendation of mobile payment apps. J. Retail. Consum. Serv. 2021, 63, 102743. [Google Scholar] [CrossRef]
  45. Perea-Khalifi, D.; Irimia-Diéguez, A.I.; Palos-Sánchez, P. Exploring the determinants of the user experience in P2P payment systems in Spain: A text mining approach. Financ. Innov. 2024, 10, 2. [Google Scholar]
  46. Lang, C.; Li, M.; Zhao, L. Understanding consumers’ online fashion renting experiences: A text-mining approach. Sustain. Prod. Consum. 2020, 21, 132–144. [Google Scholar] [CrossRef]
  47. Laws & Regulations Database of the Republic of China (Taiwan). The Act Governing Electronic Payment Institutions (19 January 2023). Available online: https://law.moj.gov.tw/ENG/LawClass/LawAll.aspx?pcode=G0380237 (accessed on 24 January 2025).
  48. Xiao, S.; Liu, Z.; Zhang, P.; Muennighoff, N.; Lian, D.; Nie, J.Y. C-pack: Packed resources for general Chinese embeddings. In Proceedings of the 47th International ACM SIGIR Conference on Research and Development in Information Retrieval, Washington, DC, USA, 14–18 July 2024; pp. 641–649. [Google Scholar]
  49. Alharbi, A.; Hai, A.A.; Aljurbua, R.; Obradovic, Z. AI-Driven Sentiment Trend Analysis: Enhancing Topic Modeling Interpretation with ChatGPT. In Proceedings of the IFIP International Conference on Artificial Intelligence Applications and Innovations, Corfu, Greece, 27–30 June 2024; Springer Nature: Cham, Switzerland, 2024; pp. 3–17. [Google Scholar]
  50. Mimno, D.; Wallach, H.; Talley, E.; Leenders, M.; McCallum, A. Optimizing semantic coherence in topic models. In Proceedings of the 2011 Conference on Empirical Methods in Natural Language Processing, Edinburgh, UK, 27–31 July 2011; pp. 262–272. [Google Scholar]
  51. Hadiat, A.R. Topic Modeling Evaluations: The Relationship Between Coherency and Accuracy. Master’s Thesis, University of Groningen, Groningen, The Netherlands, 2022. [Google Scholar]
  52. Chiang, M. Convenience Store 2023 Sales Estimated at Record-Breaking NT$413 Billion; Press Release. (15 March 2024). Available online: https://en.rti.org.tw/news/view/id/2010862 (accessed on 15 April 2025).
  53. Taiwan Institute of Economic Research. 2020 Communications Market Survey Results Report. February 2021. Available online: https://www.ncc.gov.tw/chinese/files/21021/5190_45724_210217_2.pdf (accessed on 15 April 2025).
  54. National Science and Technology Council. DIGI+: Digital Nation and Innovative Economic Development Program; Press Release. 9 September 2019. Available online: https://english.ey.gov.tw/News3/9E5540D592A5FECD/659df63b-dad4-47e3-80ab-c62cb40a62cd (accessed on 15 April 2025).
  55. Lian, J.W.; Li, J. The dimensions of trust: An investigation of mobile payment services in Taiwan. Technol. Soc. 2021, 67, 101753. [Google Scholar] [CrossRef]
  56. Dhawan, H. Android vs. Apple: How Consumer Demographics Shape Mobile App Strategies. 2024. Available online: https://www.neuronimbus.com.au/blog/android-vs-apple-how-consumer-demographics-shape-mobile-app-strategies/ (accessed on 11 June 2025).
  57. Kapias, M. Android and iPhone Users: Differences in Behavior. 2025. Available online: https://fireart.studio/blog/android-and-iphone-users-differences-in-behavior/ (accessed on 11 June 2025).
  58. AppMySite, Inc. Android vs. iOS Users: A Detailed Behavioral Comparison. 2025. Available online: https://www.appmysite.com/blog/android-vs-ios-users-a-detailed-behavioural-comparison/ (accessed on 11 June 2025).
  59. Testers HUB: Software Testing Company. The Role of User Experience in iOS and Android App Testing Services. 2025. Available online: https://www.linkedin.com/pulse/role-user-experience-ios-android-app-testing-services-testers-hub-6yedf (accessed on 11 June 2025).
  60. Ahmad, D.M.; Javed, P. Security comparison of Android and iOS and implementation of user approved security (UAS) for android. Indian J. Sci. Technol. 2016, 9, 1–7. [Google Scholar] [CrossRef]
  61. Dia, H.; Pettersson, N. Evaluating the Accuracy of Sentiment Analysis Models when Applied to Social Media Texts. Bachelor’s Thesis, KTH Royal Institute of Technology, Stockholm, Sweden, 2024. [Google Scholar]
  62. Ahmad Azrir, A.H.I.; Naveen, P.; Haw, S.C. Improving Sentiment Analysis of Shopee Reviews with Informal Language and Slang. J. Logist. Inform. Serv. Sci. 2024, 11, 151–169. [Google Scholar]
  63. Ganie, A.G. Presence of informal language, such as emoticons, hashtags, and slang, impact the performance of sentiment analysis models on social media text? arXiv 2023, arXiv:2301.12303. [Google Scholar]
  64. Maynard, D.G.; Greenwood, M.A. Who cares about sarcastic tweets? Investigating the impact of sarcasm on sentiment analysis. In Proceedings of the Ninth International Conference on Language Resources and Evaluation Conference (LREC’14), Reykjavik, Iceland, 26–31 May 2014; ELRA: Paris, France, 2014. [Google Scholar]
Figure 1. The research method and process of this study.
Figure 1. The research method and process of this study.
Bdcc 09 00165 g001
Figure 2. Score-based bar chart. (The numbers 1 to 5 represent the ratings given by users).
Figure 2. Score-based bar chart. (The numbers 1 to 5 represent the ratings given by users).
Bdcc 09 00165 g002
Figure 3. App-based bar chart. (The numbers 1 to 5 represent the ratings given by users).
Figure 3. App-based bar chart. (The numbers 1 to 5 represent the ratings given by users).
Bdcc 09 00165 g003
Figure 4. Correspondence analysis.
Figure 4. Correspondence analysis.
Bdcc 09 00165 g004
Table 1. Specialized electronic payment institutions in Taiwan.
Table 1. Specialized electronic payment institutions in Taiwan.
Company NameNumber of UsersNumber of Users (%)Cumulative Number of Users (%)
iPASS Corporation6,639,69624.4524.45
Jkopay Co., Ltd.6,603,25224.3248.77
PXPay Plus Co., Ltd.5,197,23019.1467.91
EASYCARD CORPORATION3,300,52612.1680.07
All Win Fintech Company Limited2,176,1208.0188.08
icash Corporation1,522,0905.6193.69
O’Pay Electronic Payment Co., Ltd.1,073,8383.9597.64
GAMA PAY Co., Ltd.565,2322.0899.72
ezPay Co., Ltd.75,1610.28100.00
Table 2. Features of the four electronic payment apps in Taiwan.
Table 2. Features of the four electronic payment apps in Taiwan.
App NameiPASS MoneyJiekou
Payment
PXPay PlusEasy Wallet
Features
Payment
Transfer
Deposit/withdraw
Rewards and redemption
Living payment
Riding public transport
Binding a stored-value card
✓ indicates that the app has the feature, and ✗ indicates that the app does not have the feature.
Table 3. Number of reviews before and after filtering.
Table 3. Number of reviews before and after filtering.
iPASS MoneyJiekou PaymentPXPay PlusEasy Wallet
Initial Review Count68646499196360
Final Review Count41736557005897
Number of Reviews Removed269994219463
   Number of short reviews267883201311
   Reviews with the lowest rating but strong positive sentiment18115102
   Reviews with the highest rating but strong negative sentiment130350
Average Rating2.93002.47692.13282.2816
Table 4. Results from topic modeling.
Table 4. Results from topic modeling.
Topic No.KeywordsLabel
Topic 0thanks, nfc, wallet, Easy Wallet, app, pay, update, 11, Samsung, beep ride, 10, Easy Card, May I ask, Xperia, SonySeamless integration of e-tickets and mobile payments
Topic 1unable to log in, completely awful, Unable to register, forgot password, so bad, seriously, utterly terrible, can’t log in, after the update, Forgo account, Changed phone, what’s the deal, not working, 11, constant error messageEnhancement of account access mechanisms
Topic 2pay, line, px, app, Jiekuo, payment QR code, ipass, Taiwan, R.O.C, pxpay, JKOPay, Cancel px, Why is JKO coin no longer available?, JKO coin, with lineDiverse payment channels and loyalty point systems
Topic 3unable to update, really terrible, play, app not installed, 10, garbage app, keeps asking me to update, absolutely terrible, android, unable to use, thank you, super bad, nokia, pending, after 1Application installation and updates
Topic 4easy to use, very convenient, convenient to use, great, works well, extremely convenient, highly practical, convenient and easy to use, super convenient, pretty good, practical, quite good, awesome, This app is amazing, plenty of dealsPerceived usability and promotional engagement
Topic 5inconvenient, a bit inconvenient, hard to use, terrible, poorly designed app, point card, can’t link CTBC bank, unable to bind credit card, can’t top up with a credit card, CTBC is such a major bank, just found out, what the heck, too few supported banks, very inconvenient, PXPayPayment linking and top-up optimization
Topic 6daily bill payment, electricity bill, water bill, thank you, display, bill payment, parking fee, take-out, unable to pay parking fee, totally unable to make payments, click to pay bills, vehicle license tax, meter number, system under maintenance and no data found, need to pay parking feeConvenient payment for various living expenses
Topic 7hope this app can also support instant balance check via card tap. some apps allow instant balance inquiry via card sensing, top-up, balance, payment method management, terrible, not user-friendly, hope it can be improved, keep it up, 16, checked the balance in the app the next day, no wonder the ratings are bad, shouldn’t they be addressed, even the high-paid executives can’t solve this kind of issue, can’t this be improvedOptimization of balance inquiry features
Topic 8select from photo album, button, extremely troublesome, absolutely terrible, 07, 06, negative review, terrible experience, what’s going on, the issue is just bought something overseas and doesn’t get the point of this, the problem is I’ve never even used it before, the thing is that being able to transfer money with this app after my account has been provided, so my ID number is stuck with a deactivated phone number, the withdrawal fee is waived only if the payment is made with the app, I’ve never used JKOPay before and I cannot receive a verification code after downloading itUser experience and interaction efficiency
Topic 9android, keeps crashing, app crashes, 11, 10, crash, crashes immediately upon opening, 14, constant crashing, unable to open the app, still crashes the same way, Samsung, keeps crashing repeatedly, reinstalled several times, app keeps crashingApp stability and crash recovery
Table 5. Topic interpretations.
Table 5. Topic interpretations.
Topic No.Topic NameInterpretation
Topic 0Seamless integration of e-tickets and mobile paymentsThis refers to integrating electronic tickets with mobile payment apps, enabling users to complete transactions via their smartphones when they are using public transportation, shopping, or dining out.
Topic 1Enhancement of account access mechanismsThis is related to user account management and access processes, including signup, signing in and out, and password setting
Topic 2Diverse payment channels and loyalty point systemsThis refers to combining electronic payment with various rewards, like membership points, cash back, and discount coupons, ensuring a unified and convenient user experience.
Topic 3Application installation and updatesThis involves downloading and installing the electronic payment app on a mobile device and updating the app to its latest version.
Topic 4Perceived usability and promotional engagementThis means the electronic payment app is user-friendly and easy to navigate, allowing users to get started effortlessly. It also offers a reward mechanism, providing additional value and benefits while ensuring a smooth experience.
Topic 5Payment linking and top-up optimizationThis refers to the process of binding a credit card or bank account to an electronic payment app and recharging a payment instrument, such as an electronic ticket or stored-value card, through an electronic payment app.
Topic 6Convenient payment for various living expensesThis means that users can make various payments in life through electronic payment apps, such as water bills, electricity bills, taxes, etc.
Topic 7Optimization of balance inquiry featuresThis refers to users’ operational experience and efficiency when viewing their accounts, payment tools, or e-wallet balances in electronic payment apps.
Topic 8User experience and interaction efficiencyThis refers to the user’s experience when interacting with the electronic payment app and the smoothness of the electronic payment operation process.
Topic 9App stability and crash recoveryThis refers to ensuring that the electronic payment app continues to operate normally and can quickly recover in case of an unexpected crash.
Table 6. Goodness-of-fit.
Table 6. Goodness-of-fit.
Chi-SquaredfSig.
Pearson36,169.69836,9080.997
Deviance22,559.39236,9081.000
Table 7. Model fitting information.
Table 7. Model fitting information.
ModelModel Fitting CriteriaLikelihood Ratio Tests
−2 Log LikelihoodChi-SquaredfSig.
Intercept Only2.744 × 104
Final2.256 × 1044.877 × 104400.000
Table 8. Likelihood ratio tests.
Table 8. Likelihood ratio tests.
EffectModel Fitting CriteriaLikelihood Ratio Tests
−2 Log Likelihood of Reduced ModelChi-SquaredfSig.
Intercept2.293 × 104373.71840.000
Topic 02.260 × 10435.88740.000
Topic 12.261 × 10452.81340.000
Topic 22.259 × 10434.77940.000
Topic 32.310 × 104540.57740.000
Topic 42.446 × 1041.902 × 10340.000
Topic 52.275 × 104187.31740.000
Topic 62.256 × 1044.90240.298
Topic 72.258 × 10415.80940.003
Topic 82.259 × 10426.25540.000
Topic 92.287 × 104309.00540.000
The chi-square statistic is the difference in −2 Log-likelihoods between the final model and a reduced model. The reduced model is formed by omitting an effect from the final model. The null hypothesis is that all parameters of that effect are 0.
Table 9. Results of multinomial logistic regression.
Table 9. Results of multinomial logistic regression.
Independent VariableCategory 1Category 2Category 3Category 4
CoefficientOdds RatioCoefficientOdds RatioCoefficientOdds RatioCoefficientOdds Ratio
Topic 00.8292.290 3.618 ***37.272 −1.3680.255 1.3493.853
Topic 1−5.317 **0.005 −0.8630.422 −8.207 ***0.000 1.3163.729
Topic 2−0.7000.497 4.44184.835 −8.769 **0.000 5.093 *162.813
Topic 336.761 ***9.232 × 101517.953 ***6.267 × 1077.591 **1980.046 6.136 **462.115
Topic 4−31.457 ***0.000 −24.776 ***0.000 −24.778 ***0.000 −12.730 ***0.000
Topic 59.581 ***14,489.137 1.5224.580 12.914 ***4.061 × 105−1.6680.189
Topic 62.754 *15.702 1.4744.366 0.6982.010 1.6645.279
Topic 7−0.7320.481 1.2393.451 4.332 **76.071 1.1453.143
Topic 88.014 ***3023.205 5.952 **384.615 6.133 *460.681 2.0437.713
Topic 9−19.548 ***0.000 −6.957 ***0.001 −1.0320.356 −0.7270.483
Note: The dependent variable is 5-scale sentiment polarity. The categories of sentiment polarity are as follows: Category 1: 1, Category 2: 2, Category 3: 3, Category 4: 4, and Category 5: 5. Category 5 was used as the reference category. * p < 0.05, ** p < 0.01, and *** p < 0.001.
Table 10. Contingency table.
Table 10. Contingency table.
Topicst0t1t2t3t4t5t6t7t8t9
Scales
165551219452418289921476955
21025483329444752341201648891
31207548752710538381132
458518721520418911696983054
52875311054607645149410
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.

Share and Cite

MDPI and ACS Style

Tu, S.-F.; Hsu, C.-S. Exploring Factors Impacting User Satisfaction with Electronic Payment Services in Taiwan: A Text-Mining Analysis of User Reviews. Big Data Cogn. Comput. 2025, 9, 165. https://doi.org/10.3390/bdcc9070165

AMA Style

Tu S-F, Hsu C-S. Exploring Factors Impacting User Satisfaction with Electronic Payment Services in Taiwan: A Text-Mining Analysis of User Reviews. Big Data and Cognitive Computing. 2025; 9(7):165. https://doi.org/10.3390/bdcc9070165

Chicago/Turabian Style

Tu, Shu-Fen, and Ching-Sheng Hsu. 2025. "Exploring Factors Impacting User Satisfaction with Electronic Payment Services in Taiwan: A Text-Mining Analysis of User Reviews" Big Data and Cognitive Computing 9, no. 7: 165. https://doi.org/10.3390/bdcc9070165

APA Style

Tu, S.-F., & Hsu, C.-S. (2025). Exploring Factors Impacting User Satisfaction with Electronic Payment Services in Taiwan: A Text-Mining Analysis of User Reviews. Big Data and Cognitive Computing, 9(7), 165. https://doi.org/10.3390/bdcc9070165

Article Metrics

Back to TopTop