The increasing use of mobile fitness applications (MFAs) globally, fueled by the COVID-19 pandemic and the rise of smartphones, makes understanding user perspectives crucial for improving these services. The global market for MFAs shows significant growth, with millions of active users and billions of dollars in annual revenue. Previous research has largely relied on user surveys to explore MFA adoption motivations, using theories like the extended Unified Theory of Acceptance and Use of Technology (UTAUT2) and Expectation-Confirmation Theory (ECT). However, these studies lack the exploration of user experience (UX) gleaned from online reviews. This study aims to address this gap by employing natural language processing (NLP) techniques to analyze online review comments to identify UX elements that influence user satisfaction. The research questions focus on determining the key factors influencing user satisfaction with MFAs and examining the feasibility of using UX elements in online reviews to assess this satisfaction. The study uses data from the Google US Play Store, a readily accessible source of user opinions.

The concept of UX, defined as a person's perceptions and responses to a product or service, is a significant predictor of user satisfaction. Key components of UX include usability (ease of use), usefulness (effectiveness in achieving goals), and affection (emotional response). Numerous studies have explored the influence of these components on user satisfaction across various applications, including mobile reservation services and e-learning platforms. These studies employed theoretical frameworks like TAM and ECT. While UX has been well-researched in various contexts, there is a need for more studies specifically analyzing the UX of fitness applications using online review data. The use of online reviews offers a valuable resource for understanding user perspectives, capturing honest feedback and intrinsic feelings. Previous studies analyzing online reviews have shown perceived affection and usefulness are determinants of user satisfaction in applications like Airbnb. This study builds upon this prior research to analyze the UX of MFAs, focusing on gamification's potential role in shaping user satisfaction.

The study employed two main NLP approaches: bag-of-words and sentiment analysis, applied to data collected from online reviews of MFAs on the Google US Play Store. **Study 1: Bag-of-words** Data was collected from the top five fitness apps and the top five gamified fitness apps, totaling approximately 16,000 reviews. Pre-processing steps involved removing comments with fewer than five words, non-English words, emojis, and emoticons. Tokenization and lemmatization were performed. A dictionary-based approach was used to categorize words into three UX components: usability, usefulness, and affection. The proportion of each component in each review was calculated. Multiple regression analysis was then employed to investigate the effects of these UX elements on user satisfaction separately for gamified and non-gamified apps. **Study 2: Sentiment analysis** The same dataset was used for sentiment analysis. LIWC (Linguistic Inquiry and Word Count) was used to analyze the sentiment in the reviews. Five UX elements were extracted: hedonic values (positive emotion), user burden (negative emotion), expectation confirmation, pragmatic values (work, leisure, home), and social values. Hypotheses were formulated, predicting the impact of each UX element on user satisfaction. Multiple linear regression analysis was used to test these hypotheses for both gamified and non-gamified apps. **Additional Approaches** The study also employed LDA topic analysis and K-Means clustering to explore additional methods for extracting UX dimensions. These results were compared to the bag-of-words and sentiment analysis results. Multiple linear regression was applied to assess the explanatory power of these additional approaches on user satisfaction.

**Study 1 (Bag-of-words):** In non-gamified apps, usability, usefulness, and affection all significantly influenced satisfaction. In gamified apps, usefulness showed no significant effect, while usability and affection remained significant predictors. Users of non-gamified apps expressed higher levels of usability and affection. **Study 2 (Sentiment Analysis):** In non-gamified apps, hedonic values, user burden, pragmatic values, and social values significantly predicted satisfaction, while expectation confirmation did not. In gamified apps, all five elements significantly influenced satisfaction. Hedonic values were the strongest predictor of satisfaction in both gamified and non-gamified apps. User burden had a negative effect on satisfaction in both types of apps. **Additional Approaches:** LDA topic analysis and K-Means clustering showed weaker explanatory power for user satisfaction compared to the bag-of-words and sentiment analysis approaches.

The findings highlight the crucial role of perceived affection and hedonic values in determining user satisfaction with MFAs. The marginal effect of user burden suggests that minimizing negative experiences is important, but optimizing positive emotions is even more critical. The difference in usefulness's impact between gamified and non-gamified apps warrants further investigation. The study's findings corroborate existing research emphasizing the importance of users' emotional state and positive experiences in driving satisfaction. The different results from the various NLP methods raise questions about the optimal techniques for analyzing UX in this specific context. Future research could investigate other methods for this purpose.

This study contributes to a better understanding of user satisfaction with MFAs by employing multiple NLP techniques to analyze online review comments. Key findings indicate the significant roles of perceived affection and hedonic values, and the relatively smaller role of user burden. The study’s findings offer practical implications for developers to focus on user experience enhancement by leveraging the study’s methodology and analysis techniques. Further research could explore the influence of user demographics and expand the analysis to other app stores to enhance generalizability. Exploring more complex NLP techniques is another avenue for future study.

The study has some limitations. First, it lacked user demographic data, which could influence satisfaction levels. Second, the data was solely from the Google US Play Store, limiting the generalizability of the findings. Finally, the use of more advanced NLP methods might provide additional insights.