Digital Discourse and Linguistic Proficiency: Sentiment Analysis of YouTube Comments in Informal English Learning Experiences
Article information
Abstract
This study investigates the potential of virtual communities of practice (VCoP) in English language education through sentiment analysis of YouTube comments on videos by Oliver Ssam (2022) and Moon Ga Young (2023). YouTube enables learners to exchange insights on English pronunciation and multilingual proficiency, fostering an interactive and collaborative learning environment through active engagement in comment sections. Sentiment analysis was conducted using the Sentiment VADER algorithm and Naïve Bayes classifier, achieving 94% and 96% accuracy rates for the respective videos. The study identified five key emotional themes among Korean English learners: positive emotions (“jeong”), negative emotions (“han”), realization, ambivalent emotions, and tolerance. Positive emotions were the most prevalent, encouraging greater engagement in English learning through expressions of approval and admiration. The methodological implications of applying sentiment analysis to English learners’ active engagement in YouTube comments may offer valuable insights for enhancing English education. This study underscores the value of VCoP for informal learning that enhances digital literacy and supports language acquisition beyond traditional classroom settings. By integrating VCoP into formal education, educators can foster active participation and personalized learning experiences. This approach bridges the gap between informal and formal education, offering learners flexible and self-directed opportunities for language learning.
I. INTRODUCTION
Traditional methods of instruction have undergone significant transformation due to the emerging technology and internet, promoting interactive and collaborative informal learning communities in digital platforms. In the digital landscape of this era, educational and technological innovation has fundamentally changed how learners engage with educational materials, which are the key affordances and challenges of Education 4.0 and Education 5.0 (Facer & Daanen, 2007; Santos et al., 2024). They emerged as an essential component in enhancing platforms such as Internet of Things (IoT), YouTube, social media, metaverse, and over-the-top (OTT) media services employed as learning experiences (Barry et al., 2016; Bonfield et al., 2020; Chopra, 2025; Duffy, 2008; García-Peñalvo, 2018; Im & Park, 2025; Lee & Hwang, 2023; Moghavvemi et al., 2018; Rho, 2024). These have become invaluable tools in contemporary education, providing learners with numerous opportunities to engage with content and promoting deeper involvement through multimedia elements.
Previous studies on English language teaching and learning have predominantly focused on formal educational settings, emphasizing classroom-based instruction, curriculum development, and teacher-led methodologies (Ellis, 2008; Richards & Rodgers, 2014). While research on informal language learning has gained traction in recent years, much of it has centered on mobile-assisted language learning and self-directed study through digital resources (Benson, 2011, 2017; Sockett, 2014). The role of online communities of practice, particularly on social media and video-sharing platforms like YouTube, has become popular and may have benefits in English education. These platforms facilitate peer-to-peer learning, collaborative knowledge construction, and engagement with authentic linguistic input in ways that diverge from traditional pedagogical approaches (Shoufan & Mohamed, 2022; Wang & Chen, 2020). While studies acknowledge the authenticity and accessibility of YouTube content (Shoufan & Mohamed, 2022), few have investigated how its comment sections can be harnessed as pedagogical tools to foster active engagement and corrective feedback, particularly for pronunciation challenges like the English diphthong [ou] in “focus.” Millennials and Generation Z increasingly utilize YouTube comment sections as digital forums for democratic discourse, fostering critical thinking and intellectual engagement (Schmidt et al., 2017; Statista, 2025; Zimmermann et al., 2022). These interactive virtual communities of practice (VCoP) enable users to construct and exchange knowledge, articulate diverse perspectives, and refine their cognitive and linguistic proficiency. Adapted from Wenger’s (1998) framework, VCoPs are defined as online groups united by a shared purpose, such as enhancing English pronunciation in this study, where members sustain knowledge co-construction through technology-mediated, often asynchronous, interaction. Among diverse VCoPs, YouTube has undergone a remarkable transformation from a traditional entertainment hub into a multifaceted educational platform (Barry et al., 2016; Bonfield et al., 2020; Duffy, 2008; Lee, 2021).
This study aims to analyze the sentiment expressed in YouTube comments about the English learning experiences—ranging from frustration to epiphany—in Oliver Ssam (2022) and Moon Ga Young (2023)’s videos1, leveraging advanced sentiment analysis techniques employing Python and Orange 3, grounded in second language acquisition (SLA) theories such as the affective filter (Krashen, 1982), corrective feedback (Lyster & Ranta, 1997), and willingness to communicate (MacIntyre et al., 1998). By leveraging these theories, this study seeks to uncover how sentiment analysis can illuminate learners’ perceptions regarding their educational journeys of English learning. Furthermore, it intends to evaluate the role of digital platforms as a catalyst for creating interactive and engaging English learning experiences, fostering a sense of informal online community among learners, and enhancing their overall educational engagement. Interactive and collaborative informal learning communities refer to digital spaces where learners voluntarily participate in reciprocal exchanges and joint problem-solving without formal instructional oversight (Benson, 2015). Interactivity manifests through asynchronous comment-reply threads, where users pose questions (e.g., “How do you say ‘focus’ with [ou]?”) and receive responses, fostering dialogue. Collaboration occurs as learners collectively tackle linguistic issues, such as when Korean users explain the absence of diphthongs in their native phonology and suggest practice strategies, creating a supportive, peer-driven learning environment. In this study, these communities are exemplified by YouTube comment sections where Korean and global learners engage in discussions that blend linguistic inquiry with sentiment, such as 1) positive emotions (“jeong”); 2) negative emotions (“han”); 3) realization; 4) ambivalent emotions; and 5) tolerance expressed in the YouTube comments. This study aims to bridge the gap by analyzing YouTube comment sections as digital spaces for informal English learning, where learners engage in discussions, exchange insights, and develop linguistic proficiency within a community-driven environment of digital spaces, employing sentiment analysis of YouTube comment on English learning to offer methodologically and pedagogical implications for integrating VCoPs into English education.
II. LITERATURE REVIEW
Education has experienced notable transformations fueled by technological advancements and the increasing integration of digital platforms. Studies indicate that YouTube, social media, and OTT platforms significantly contribute to informal learning by offering access to a vast array of user-generated content. These platforms enable learners to engage with content in personalized online learning and raise a “sense of community and presence” (Johnson et al., 2023, p. 373) by participating in discussions, accessing real-world applications, and enhancing their understanding through diverse perspectives. In contrast to traditional classroom settings, where a standardized curriculum often governs learning, digital platforms promote self-directed learning, allowing students to select content that aligns with their interests and skill levels. YouTube, in particular, has proven to be highly effective in supplementing formal education, offering accessible and engaging language learning resources that cater to various skill levels. The platform’s recommendation algorithms further personalize the learning experience by suggesting relevant videos based on user’s preferences and previous interactions (Shoufan & Mohamed, 2022; Wang & Chen, 2020).
The concept of online communities of practice has gained prominence in the digital age (see Figure 1, Wenger, 2010, as cited in Bates, 2014). According to Wenger (2015), YouTube is positioned within the online learning domain of communities of practice (CoPs) in Figure 1. As a VCoP tool, YouTube facilitates the sharing of personal experiences and expertise, particularly in the context of English language learning. It serves as a virtual platform where learners share insights on English pronunciation and multilingual proficiency. Through active participation in video comment sections, users engage in informal learning and alignment within a VCoP, contributing to a dynamic and interactive learning community. These communities enable educators and learners to engage in continuous professional development and knowledge sharing beyond traditional classroom boundaries. In such informal online communities, participants can enhance their digital literacy skills as well as language learning and adapt to evolving educational technologies (Johnson & Majewska, 2022). CoPs consist of individuals connected by shared concerns, interests, challenges, or passion for a specific topic, fostering learning and expertise development through continuous interaction (Wenger, 2010; Wenger et al., 2002). While many CoPs emerge informally, formal recognition is more likely when they are strategically utilized within organizations. The advent of the internet and information and communication technologies (ICT) has expanded CoPs beyond geographical constraints, facilitating the rise of VCoPs (Wenger, 2015). VCoPs are defined as technology-supported communities that enable knowledge sharing and collective learning among geographically dispersed members (Ardichvili et al., 2003; Dubé et al., 2005). Many traditional CoPs have transitioned into online environments, while new VCoPs continue to emerge as organizations and industries become increasingly virtual. Research suggests that while tacit knowledge sharing—central to CoP theory—occurs in VCoPs, it may be less effective than in traditional CoPs (King, 1996).
Tools That Support Communities of Practice (Wenger, 2010, as cited in Bates, 2014)
Incorporating digital platforms like YouTube, social media, metaverse environments, and OTT media services into educational settings has transformed traditional learning experiences. These platforms offer interactive and immersive opportunities that cater to diverse learning preferences. For instance, Duffy (2008) highlighted the educational potential of YouTube, emphasizing its capacity to provide accessible and engaging content. Similarly, Wang and Chen (2019) pointed out that Taiwanese EFL students used YouTube to supplement their English learning, valuing its flexibility and engagement but recognizing its limitations for exam preparation. They suggest integrating YouTube content with formal instruction for a more effective learning experience. Barry et al. (2016) found that the “YouTube Generation” or “Generation Connected” (Gen C), primarily Millennials, increasingly integrates social media into anatomy education. Their study revealed that 78% of medical and radiation therapy students used YouTube for anatomy learning, highlighting the need for blended learning approaches with academic and ethical guidance. Chopra (2025) explores how OTT platforms, traditionally used for entertainment, can enhance student learning experiences. Analyzing interviews with 35 undergraduate and postgraduate students from various countries, this study identifies how OTT platforms can offer more engaging, personalized, and interactive learning experiences compared to traditional digital learning tools. Im and Park (2025) show that vlogs foster Korean EFL pre-service teachers’ translanguaging practices and multimodal literacy. Furthermore, immersive technologies such as the metaverse provide experiential learning experiences (Dede, 2020; Dede & Richards, 2012; Lee & Hwang, 2023).
Research on English language learning from the late 1980s to the early 2000s has highlighted how successful language learners employ both direct and indirect strategies, categorized as cognitive, metacognitive, and socio-affective approaches (O’Malley & Chamot, 1990; Oxford, 1990; Weinstein & Mayer, 1986; Wenden & Rubin, 1987). In addition to cognitive factors, affective influences play a crucial role in second language acquisition (SLA). Krashen’s (1985) Affective Filter Hypothesis suggests that motivation, attitude, anxiety, and self-esteem can either facilitate or hinder language learning. A lower affective filter, characterized by strong self-esteem, a clear study purpose, and manageable anxiety, allows for more significant language input, whereas a higher affective filter can impede learning. Similarly, Stern (1983) emphasized that affective factors are as significant as cognitive ones in SLA. Previous studies have primarily focused on language learners’ negative emotions, such as anxiety and nervousness, and language learning strategies aimed at achieving native-like proficiency. However, this study explores how VCoPs and digital environments such as YouTube have transformed English learning settings for language learners. These online communities play a pivotal role in facilitating learning English pronunciation through authentic materials, such as YouTube videos, while encouraging learners to participate actively. By leaving comments and engaging in discussions, EFL learners and users become active contributors to the VCoP, fostering collaborative learning and deeper engagement with pronunciation and language acquisition. This study will conduct a sentiment analysis of comments on two YouTube video clips (Moon Ga Young and Oliver Ssam) related to English pronunciation learning, providing insights into learners’ emotional responses and engagement in the VCoP. The following research questions guide this study:
1. How do sentiment analysis techniques applied to YouTube comments reflect learners’ perceptions and experiences of English learning through digital platforms?
2. What are the predominant emotional tones found in YouTube comments on English learning videos, and how do they influence engagement and learning outcomes?
3. How do social media platforms, including YouTube, facilitate dynamic engagement and personalized English learning experiences in informal VCoP for English learners, and what are the pedagogical implications for formal and non-formal language teaching stakeholders?
III. METHOD
1. Data Segregation Process
In this study, YouTube comments were collected from two video clips that reflect Koreans’ emotions toward English pronunciation. Two videos, one by Oliver Ssam and the other by Moon Ga Young, were chosen after careful discussion among the authors due to their representativeness of English pronunciation and the ability to elicit diverse emotional responses critical to the study’s sentiment analysis goals. Oliver Ssam’s video, a widely viewed language tutorial on practical Korean conversational skills, was selected for its high engagement, with over 1,560,000 views and approximately 6,000 comments. The authors picked the video about the pronunciation of ‘focus,’ which can be heard differently by a native speaker, and it provokes diverse emotions in Oliver Ssam’s clip. Similarly, Moon Ga Young’s video of her multilingual abilities as a Korean actress was chosen for its motivational tone and cultural relevance to Korean-speaking audiences. One of the guests sitting next to Miss Moon showed admiration for English pronunciation with a German accent. The comments on this video reflected complex emotions such as admiration, inspiration, and self-doubt, providing a complementary dataset to evaluate English in a global context.
The data was extracted from an API provided by Google, which provided access to comment threads from selected videos while adhering to Google’s terms of service. The dataset included both English and Korean comments, with the latter requiring translation to align with the English-based model. Korean comments were translated to English using Google Translate, a widely used tool for its robustness in handling informal text. For text preprocessing, the Natural Language Toolkit (NLTK) was employed for data cleaning, using regular expressions via Google Colab’s module to remove URLs, non-textual metadata (e.g., timestamps), and irrelevant special characters from comments while preserving typos, grammatical errors, and emojis to maintain the authenticity of commenters’ emotions (Bird et al., 2009). After that, the authors removed comments unrelated to the video content, including advertisements and religious publicity. The total number of comments in Oliver Ssam’s clip was 6,342 on the extraction date, and 306 comments were eliminated. In Moon Ga Young’s clip, the total number of comments was 1,156 and no comments were deleted. After that, the long comments were summarized by text classification using a machine learning algorithm. For emotion classification, this study extracted core emotions with the “monologg/bert-base-cased-goemotions-original” model, enabling a more nuanced understanding of emotional expressions in textual data. Fine-tuned on the GoEmotions dataset, comprising Reddit comments labeled with 27 emotion categories plus a neutral class, the model is well-suited for analyzing social media text similar to YouTube’s comment ecosystem. Its multi-label classification capability allows it to assign multiple emotions to a single comment, reflecting the complexity of user reactions. However, existing Korean corpora tend to be limited in scale, with fewer instances, and encompass a narrow spectrum of emotions, restricting their utility for comprehensive emotion analysis (Jeon, 2022). Consequently, the GoEmotions dataset is frequently employed for analyzing Korean text by applying machine translation, offering a robust foundation for studies like this one, which examines emotional expressions in translated YouTube comments (Demszky et al., 2020). After identifying the primary emotions, the authors deliberated before deciding which emotions were closest to the subscribers’ intentions or thoughts.
2. Sentiment Analysis: VADER and Naïve Bayes
Sentiment VADER (Valence Aware Dictionary and sEntiment Reasoner) was applied using the Python package SentimentIntensityAnalyzer. It determines whether the degree of YouTube comments is positive (1), negative (-1), or neutral. The cutoff value is 0.05, and the compound value is a normalized sentiment score that ranges from -1 to 1. When a value is between -0.05 and 0.05, it is neutral; a value over 0.05 is positive; a value lower than -0.05 is negative (Hutto & Gilbert, 2014). After applying Sentiment VADER, word clouds for Figure 6 and Figure 7 were initially generated using Orange 3, but due to lower image resolution, the authors created higher-quality word clouds using https://www.freewordcloudgenerator.com/ based on core emotions extracted from YouTube comments in the dataset. After processing the VADER algorithm with emotional data, Naïve Bayes was performed. Naïve Bayes is a machine learning algorithm that predicts the sentiment of comments by applying a preprocessed dataset. The dataset is divided into 80% for the training model and 20% for the testing model (Novendri et al., 2020). Naïve Bayes classifier is conducted using the Python package scikit-learn, and it shows correct and incorrect positive/negative comments by the classifier (Chaithra, 2019).
Word Cloud Derived From Comments on Oliver Ssam’s by Emotion Classification
Word Cloud Derived From Comments on Moon Ga Young’s by Emotion Classification
IV. FINDINGS AND DISCUSSION
VADER algorithm was used for comments of two YouTube clips (Moon Ga Young; Oliver Ssam). Table 1 shows the results of Sentiment VADER for two videos. Both clips showed more positive comments than negative comments. Figures 2 and 3 imply column charts for high rates of positive sentiment in both clips, but the rate of negative sentiment is much higher in Oliver Ssam’s than in Moon Ga Young’s.
Results of Sentiment VADER
Column Chart of Sentiment VADER Result (Oliver Ssam)
Column Chart of Sentiment VADER Result (Moon Ga Young)
All the sentiment distributions of Oliver Ssam’s clip are in Figure 4. The positive score ranges from 0.0 to 1.0. It reveals most of the positive comments in Oliver Ssam’s are low positive sentiment scores. For example, one comment on YouTube says, “Dedicated English teacher taught English featuring family members long time concept teaching lot Koreans want to learn accurate English pronunciation learned channel irreplaceable,” with a positive score of 0.177, a negative score of 0.0, a neutral score of 0.823, and a compound score of 0.5106. The average positive sentiment score is 0.1461. The negative sentiment distribution of Oliver Ssam’s clip ranges from 0.0 to 1.0. It shows a similarity to the positive sentiment distribution with a prevalence of low negative sentiment scores. An example of a low negative sentiment score is “The difficulty level of English is extremely high because there are no principles of pronunciation.” Its negative score is 0.261, its positive score is 0.0, the neutral score is 0.739, and the compound score is -0.5574. The average negative score is 0.1070. The mean of the neutral score is 0.7469, and it implies neither extremely positive nor negative. The compound score ranges from -0.9958 to 0.9965, and the average score is 0.0669.
Histogram of VADER Sentiment Analysis (Oliver Ssam)
Figure 5 shows the four sentiment distributions of Moon Ga Young’s clip. The positive score ranges between 0.0 and 1.0. It demonstrates that the majority of Moon Ga Young’s positive comments have low positive sentiment scores. For instance, “I thought it was just me, but I feel somewhat comforted when I hear this logic. When you learn and use a certain language a lot, you gradually forget the language you knew before” illustrates a positive score of 0.173, a negative score of 0.069, a neutral score of 0.759, and a compound score of 0.5515. The mean positive score of 0.2323 is higher than that of Oliver Ssam’s clip, which has a score of 0.1461.
Histogram of VADER Sentiment Analysis (Moon Ga Young)
Similar to positive sentiment distribution, there is a preponderance of low negative sentiment scores in negative sentiment distribution. It ranges from 0.0 to 1.0, and one of the comment examples is, “The perspective on education may differ from person to person, but can you say it’s bad? There are many people who are full of holes even in just one language, so I don’t know if it’s possible to generalize because it’s different for everyone depending on the degree.” It indicates a negative score of 0.094, a positive score of 0.0, a neutral score of 0.906, and a compound score of -0.6956. The average negative score in Moon Ga Young’s clip is 0.0581, lower than the 0.1070 shown in Oliver Ssam’s clip. The neutral score has a mean of 0.7096, which reflects that it is neither very positive nor very negative. The average compound score is 0.3033, with a range of -0.9831 to 0.9972.
Table 2 indicates the confusion matrix linked to the Naïve Bayes classifier’s application to the test data. In Oliver Ssam’s clip, the accuracy of the classifier is 94%. Out of 2895 positive comments, 2808 were predicted correctly as positive. With 2259 negative comments, 2169 were correctly predicted as negative. In Moon Ga Young’s clip, the accuracy is 96%, and out of 694 positive comments, all 694 comments were predicted correctly as positive. Of 174 negative comments, 162 were correctly predicted as negative.
Confusion Matrix
The researchers iteratively reviewed the comments and sentiment analysis results until they reached a consensus on five key themes related to two YouTube video clips. A qualitative research approach was employed to analyze how actual comments express emotions to complement the quantitative findings discussed above. The first clip featured an anecdote from Oliver’s family, a bilingual and bicultural content creator, about how the English word ‘focus’ was misheard as the swear phrase “f**k us” within his family. The second video showcased a multilingual Korean actress, Moon Ga Young, in a TV show episode that panelists praised for her fluency in Korean, German, and English. The comments in these two YouTube video clips reflected English learners’ diverse ways of expressing ambivalent feelings regarding English, discussing English education, sharing personal language learning experiences, and offering and sharing constructive feedback on this informal online community of practice (Wenger, 2010). The five key themes that emerged from the content analysis are as follows: 1. Positive emotions “jeong”; 2. Negative emotions “han”; 3. Realization: constructive feedback; 4. Ambivalent emotions: socioeconomic status; and 5. Tolerance: global English and English as a lingua franca.
1. Positive Sentiment of “Jeong” in YouTube Comments: Korean Learners of English Pronunciation
The Korean term “정(jeong)” encapsulates a wide range of emotions that arise within interpersonal relationships, such as affection, care, emotional connection, and attachment (Chung & Cho, 2006). According to Chung and Cho (2006), “jeong” extends beyond mere emotions to encompass social dimensions, including inter-individualism, reflecting the relational bonds and collective cultural identity inherent in Korean society. Ka (2010) further describes “jeong” as embodying “the feeling of endearment,” “the warmth of human connection,” “compassionate attachment,” and “a deep yearning for someone or something.” These characteristics align with the positive emotions, such as approval, admiration, and optimism, observed in the content of “Oliver Ssam” and “Moon Ga Young,” as detailed in Tables 4 and 5. These emotions, which include approval and admiration, resonate with the concept of “jeong.” For instance, Table 3 presents sample comments illustrating “jeong,” where commenters express a variety of positive sentiments. One example highlights the importance of the precise pronunciation of “focus” as significant in English learning. Another reflects admiration for the realization that “focus” can be pronounced variously, inspiring the commenter regarding English pronunciation. Additionally, a commenter expresses support and admiration for Moon Ga Young’s multilingual abilities, suggesting a sense of connection and emotional investment in her linguistic fluency. Overall, the positive emotions evident in the comments on the Oliver Ssam and Moon Ga Young videos stem from learner-oriented individuals who provide highly supportive and receptive feedback, consistent with the dynamics of “jeong.” All comments in the following tables in this paper were directly extracted from YouTube, preserving their original wording, including typos and grammatical errors. This approach ensures that the commenters’ emotions, thoughts, and attitudes toward the videos are authentically represented, free from assumptions or alterations that could distort their intended meaning.
Sample of Positive Comments From Oliver Ssam and Moon Ga Young
Frequently Shown Positive Emotions in Oliver Ssam
Frequently Shown Positive Emotions in Moon Ga Young
2. From Negative Sentiment to Engagement in YouTube VCoP: “Han” as a Dual Emotional Framework in English Education
The Korean term “한(han)” represents a multifaceted linguistic and cultural phenomenon that transcends simple definitions, embedding within it a rich tapestry of lived cultural experiences (Lee & Choi, 2003). As a sociocultural construct, “han” encapsulates a spectrum of emotions arising from unavoidable and frustrating circumstances, reflecting the process of adapting to or coping with such challenges (Kim & Ko, 2007). While Kim and Ko (2007) note that Americans tend to associate “han” with negative outcomes such as depression, exhaustion, and apathy, Koreans perceive it as encompassing both negative and positive dimensions, including self-care, resilience, and endurance. In this regard, “han” functions as a meta-mood, mediating the interplay between positive and negative emotions and simultaneously embodying sadness and hope (Kim, 2017). This duality is mirrored in the frequent negative emotions observed in the videos featuring Oliver Ssam and Moon Ga Young, as documented in Tables 7 and 8. Both videos have similar rankings of negative sentiment, such as annoyance, disapproval, confusion, and disappointment. Table 6 provides illustrative examples of YouTube comments, capturing not only pronounced negative emotions but also subtle positive sentiments such as hope, insight, and perseverance, reflecting the emotional complexity. A comment on Moon Ga Young’s video expresses confusion regarding English accents. Furthermore, a disapproval comment on Oliver Ssam’s video reveals discomfort alongside an acknowledgment of his intentions and the significance of pronunciation. This commenter, despite using the term “racism,” demonstrates an effort to engage constructively with the online community, reducing conflict while seeking to embrace cultural diversity and expand their understanding of pronunciation. Similarly, another comment reflecting confusion about the pronunciation of “focus” shows uncertainty about the correct articulation but aligns with Oliver Ssam’s perspective, suggesting an attempt at comprehension. Within this framework, comments laden with negative emotions, such as confusion and disapproval, also weave in personal experiences of understanding, hope, and engagement, resonating with the essence of “han” as a cultural concept that intertwines both negative and positive emotional currents.
Sample of Negative Comments From Oliver Ssam and Moon Ga Young
Frequently Shown Negative Emotions in Oliver Ssam
Frequently Shown Negative Emotions in Moon Ga Young
Figures 6 and 7 present word clouds illustrating the emotional landscape of the two videos. As “han” is the complex interplay of linguistic and cultural dimensions that defies simplistic characterization, encapsulating emotions and experiences, word clouds depict diverse sentiments, including confusion, annoyance, disapproval, disappointment, sadness, curiosity, admiration, caring, realization, optimism, and approval, among others. “Han” may elevate the affective filter, as described by Krashen (1982), by intensifying emotional barriers to language acquisition, particularly in the context of YouTube comments expressing struggles with language learning. Nonetheless, Cho (2018) asserts that “han” affects academic motivation and identity in English-dominant environments, influencing motivation in line with SLA’s willingness to communicate and pertinent for Korean learners navigating English education in Korean American students’ educational experiences. According to Kim (2020), “han” related sorrow may elevate affective filters in Korean American identity in English-language contexts, and it shapes identity and emotional expression in English-language contexts, particularly among Korean American students navigating cultural dislocation.
3. Neutral Sentiment in Active VCoP Engagement: Redefining Teacher-Student Roles in English Pronunciation Learning
The comment sections of YouTube videos on English pronunciation serve as vibrant, globally accessible platforms where Korean users of English, alongside a diverse array of international participants, engage not merely as passive consumers but as active contributors to a VCoP. This study’s sentiment analysis of these interactions reveals a recurring theme of “realization”—an affective and cognitive response where users articulate epiphanies about linguistic differences between Korean and English that underpin pronunciation difficulties (See Figure 6, Table 9, and Table 10). YouTube users emerge as active contributors, offering constructive feedback and engaging in dynamic self-regulated learning that propels their pronunciation development beyond conventional classroom boundaries. This participatory model addresses a critical gap in traditional English education, where pronunciation corrective feedback (CF) is often limited (Baker & Burri, 2016; Weekly et al., 2022), highlighting the VCoP’s potential as a transformative space for learner engagement and corrective support. Three comments from Oliver Ssam’s YouTube video clip in Table 9 exemplify these realizations: one highlights how English is a stress-sensitive language, a feature most Koreans overlook; another points out that Korean vowels differ from English due to the latter’s use of stressed versus unstressed vowels, often reduced to schwa or ellipsis; and a third observes that the Korean vowel system lacks diphthongs like the [ou] sound in “focus,” treating Korean as a syllable-timed language in contrast to the stress-timed nature of English.
Sample of Neutral Comments From Oliver Ssam
Frequently Shown Neutral Emotions in Oliver Ssam
This participatory model of English learning, emerging from this study’s VCoP of YouTube comment sections, aligns with Pica’s (1987) Interaction Hypothesis, where negotiation of meaning through confirmation checks and clarification requests drives language learning. Unlike the classroom’s hierarchical structure, feedback in the VCoP flows multidirectionally among Oliver Ssam—a native speaker, Korean English learners, Korean peers, and occasional non-Korean speakers, blurring teacher-student boundaries. This active engagement in learning English pronunciation in the YouTube comment section may be further enhanced by dynamic self-regulated learning, where learners manage their goals and strategies in response to community input, as sample comments in Table 9. Saito (2021) underscores the importance of targeting phonological features—such as stress, vowel reduction, and diphthong production—to bridge the gap between learners’ native language tendencies and English’s stress-timed rhythm, with segmental accuracy (e.g., mastering diphthongs like [ou]) driving native-like pronunciation and prosodic adjustments enhancing comprehensibility. The collaborative dissection of errors like the [ou] diphthong not only advances individual competence but also positions the VCoP as a democratized alternative to conventional instruction of English education, challenging native-speaker-centric standards and affirming Pica’s (1987) claim that negotiated interaction drives acquisition.
4. Ambivalent Emotions in Language Learning: Socioeconomic Status and VCoP Reflections
The comments on Oliver Ssam and Moon Ga Young’s videos reflect a blend of ambivalent emotions, encompassing both positive and negative sentiments. In Oliver Ssam’s video, many commenters express appreciation for the educational insights provided on pronunciation while simultaneously voicing concerns about negative attitudes from others, such as accusations of racism. Additionally, these comments often draw comparisons between Korean and English pronunciation, highlighting the challenges of achieving accuracy (See Table 11). In the field of second language acquisition, affective factors play a critical role (Gardner & MacIntyre, 1993). These scholars argue that educators can shape learners’ affective experiences by fostering positive attitudes, alleviating anxiety, boosting motivation, and nurturing autonomy and self-esteem, all of which can be supported through conducive learning environments and customized instructional approaches. In the 1990s, second language learning focused on reducing negative emotions, such as anxiety (Gardner & MacIntyre, 1993). Later, MacIntyre (2002) proposed a social-cognitive framework, emphasizing the cyclical interplay among anxiety, cognition, and behavior, and advocating for supportive learning contexts to mitigate language anxiety and cultivate positive emotions. Such environments, enriched with empathy, can enhance learners’ motivation and confidence.
Sample of Ambivalent Comments From Oliver Ssam and Moon Ga Young
In the comments on Moon Ga Young’s video, viewers exhibit mixed emotions toward her, including envy of her multilingual proficiency and self-comparisons tied to her family background. These reactions suggest that commenters evaluate their language abilities not only in terms of effort or inherent aptitude but also in relation to socioeconomic factors. For instance, one commenter in Table 11 notes a shared social background with Moon Ga Young but attributes differences in language fluency to personality traits. This observation aligns with the concept of a “second language ego,” distinct from the first language ego, which emerges in second language learning contexts. Dörnyei (2014) underscores the significance of individual differences—such as personality, motivation, and language aptitude—in determining success in second language acquisition. Within this framework, the ambivalent emotions expressed in the comments on Oliver Ssam and Moon Ga Young’s videos can be interpreted as a mix of frustration and reflection on second language learning challenges, interwoven with comparisons of socioeconomic circumstances, as well as expressions of motivation tempered by understanding and critique.
5. Tolerance in Global English: Multilingual Accents in English as a Lingua Franca
The last category, tolerance, refers to the capacity to navigate and withstand the emotional challenges inherent in the language learning process. This concept entails intricate psychological mechanisms, such as intuitive and deliberative reasoning, shaped by both individual characteristics and contextual variables (Verkuyten et al., 2022). As evidenced in Table 12, several comments highlight the frustration and difficulty associated with achieving correct pronunciation, underscoring the significant effort required to master English articulation. Some commenters reflect on instances where President Trump and a Japanese speaker struggled to pronounce “focus on,” sarcastically noting that “focus” poses a particular challenge for non-native speakers. Meanwhile, comments in Table 13 related to Moon Ga Young’s video express approval and insight regarding her German-influenced English pronunciation. Tyler, a panel commentator featured in Moon Ga Young’s clip, also expresses a preference for her European-accented English, signaling a subtle rejection of standardized English pronunciation norms. This suggests a need for greater tolerance of varied English pronunciations within the framework of global English as a lingua franca. These observations collectively challenge the rigidity of conventional English pronunciation standards and advocate for a broader, more inclusive perspective that embraces and accommodates diverse pronunciation patterns in the globalized context of English as a lingua franca.
Sample of Global Comments From Oliver Ssam
Sample of Global Comments From Moon Ga Young
In this paper, it is important to recognize that the VCoP can be an effective tool as an informal and voluntary platform for English education. There were similar studies applying the YouTube platform with formal English education (Duffy, 2008; Wang & Chen, 2020), but this research performed sentiment analysis with YouTube comments in an informal situation. The classifier has predicted the sentiment of each comment, as indicated by the confusion matrix generated from its performance. It is observed that the Naïve Bayes classifier demonstrates strong performance in sentiment prediction when combined with the Sentiment VADER lexical method for analyzing Oliver Ssam and Moon Ga Young’s videos. The result shows that the accuracy of the classifier in Moon Ga Young’s video is higher than that of Oliver Ssam, especially in terms of positive sentiment. The brevity and specificity of the comments in Moon Ga Young’s video surpassed those in Oliver Ssam’s overall.
The sentiment analysis of YouTube comments revealed a categorization into five distinct sentiments. Notably, the positive sentiment, which bears a resemblance to the concept of “jeong,” emerged as the most prevalent sentiment among all comments. This sentiment facilitates the formation of interpersonal connections, encompassing emotions such as affection, care, emotional bonding, and attachment in the process of language learning in an informal VCoP setting. In contrast, negative sentiments akin to “han” reflect commenters’ personal experiences, evoking feelings of annoyance, disapproval, confusion, understanding, hope, and engagement. These sentiments incorporate cultural concepts that encompass both negative and positive emotional dimensions, delivering Korean English learners’ frustration and hope for acquiring English proficiency. Neutral sentiments were observed in the context of VCoP, where commenters redefined teacher-student roles in English pronunciation learning. For instance, commenters actively shared their methods for correctly pronouncing the word “focus.” Next, ambivalent emotions, which combine both positive and negative feelings, were also evident, reflecting the authentic expressions of commenters towards language learning. The commenters demonstrate their personal experiences, particularly sharing their perspectives on socioeconomic status. The last theme of sentiment in comments is tolerance. It conveys criticism of standard English pronunciation and advocates for tolerance of diverse English pronunciations in Global English as a lingua franca (Berns, 2008; Dewey & Jenkins, 2010; Jenkins, 2014; Walker, 2021).
Positive emotions such as personal fulfillment, social bonding, and delight played an essential foreign language learning achievement (Aydın & Tekin, 2023). Moreover, according to Wang and Chen (2020), a self-regulated online learning environment makes a positive impact on the learning objectives, boosting their motivation and giving them access to additional learning materials. Previous empirical studies have indicated positive emotions associated with autonomous motivation, learning behavior, and achievement (Alamer & Lee, 2019; Lee & Drajati, 2019). For example, autonomous motivation can enhance students’ perceived value, such as appraisal in online learning (Liu et al., 2020; Luo et al., 2021). Moreover, language learners’ self-explanation and engagement in the YouTube comments section may facilitate effective and complex learning and cognitive strategies (Tan et al., 2025) and foster students’ achievement in foreign language learning (Wang et al., 2022).
Sentiment analysis of YouTube comments holds significant implications for language learning, particularly within the context of VCoP. The YouTube comments section serves as an informal English learning platform where individuals can voluntarily express their emotions, fostering motivation to learn languages. Even when confronted with negative emotions, such as confusion over English pronunciation or challenges in the language learning process, participants can openly share their thoughts and emotions within VCoP. Thus, sentiment analysis of YouTube comments in this study may enhance understanding of English learning processes.
V. CONCLUSION
This study reveals two key implications methodologically and pedagogically against the backdrop of its findings. Firstly, the sentiment analysis on VCoP highlights a novel approach to leveraging VCoP for English education. The active engagement of users in video comment sections fosters informal learning and alignment within VCoP, thereby cultivating a dynamic and interactive learning environment. These communities facilitate continuous professional development and knowledge sharing among educators and learners, extending beyond the confines of traditional classrooms. Moreover, participation in such informal online communities enhances participants’ digital literacy skills, supports language acquisition, and enables adaptation to emerging educational technologies (Johnson & Majewska, 2022)
Second, this study demonstrates the pedagogical potential of VCoP for Korean English learners, with sentiment analysis of YouTube comments revealing diverse emotional drivers–positive “jeong,” provocative “han,” neutral engagement “realization,” ambivalent socioeconomic reflections of learning English, and tolerance for multilinguals’ English accents—that may enhance English pronunciation learning through high engagement and interactivity on digital platforms, particularly when addressing authentic English productive skills (Lee, 2021; Lee et al., 2023). This participatory culture facilitates the development of virtual learning communities, where individuals collaboratively negotiate meaning and contribute to a broader discourse on digital literacy and education, serving as language educational content venues and fostering interactive language learning environments (Pires et al., 2022; Tan, 2013). These findings may inform targeted English teaching strategies, including structured YouTube comment prompts to cultivate “jeong” to inspire and improve peer support for English pronunciation practice, teacher-facilitated discussions on Learning Management Systems (LMS) to reflect on neutral “realization” sentiments as meta-cognitive language learning strategy, and exposure activities using free YouTube videos featuring diverse English accents to foster tolerance for global English. These language teaching strategies may promote immersive learning by combining visual, audio, and textual elements of digital resources for English language learning, leading to better reading and speaking production (Gilakjani, 2012). Ong and Quek (2023) found that improving these interactions is crucial, as their mixed-methods study of secondary students highlighted the importance of social engagement and effective teaching strategies in online classrooms. Personalization is also a key factor, as students prefer the ability to customize content to their preferences and provide feedback on their learning progress. This research found that YouTube comments created VCoP beyond the classroom for language learning. However, the shift to online learning has reduced teacher-student interactions, which may contribute to ineffective learning. According to Shoufan and Mohamed (2022), their systematic review of YouTube in education highlights concerns about content quality and the lack of a direct correlation between production strategies and learning outcomes. While most studies emphasize YouTube’s role as an educational tool, offering free, accessible, and engaging resources, they also underscore its challenges. These challenges can be better addressed by integrating YouTube into a structured, pedagogy-driven language learning framework. As the boundary between formal and informal language learning continues to blur (Benson, 2011, 2017; Sockett, 2014), English teachers, acting as guides and facilitators, can mitigate these challenges by delivering instruction within a guided learning environment, fostering more effective language learning design, even in informal contexts. Therefore, active engagement and interaction within VCoPs for language learning, supported by teachers as guides and facilitators, can enhance collaborative knowledge-sharing, foster deeper language acquisition, and leverage students’ affective factors and sentiment analysis results to optimize learning outcomes.
This study is not without limitations. The anonymity of YouTube commenters results in a lack of demographic information, including gender, age, and nationality. As YouTube comments are not a form of an interview, the analysis might be interpreted without confirmation to verify the commenters’ emotions or intentions. Furthermore, though public, YouTube comments are often posted with an expectation of limited visibility within the platform’s community, not as data for academic research (Fiesler, 2019). This study anonymized all data by removing usernames to protect commenters’ privacy, but the assumption that public posting implies consent is ethically contentious. In an ethical stance in analyzing YouTube comments, the absence of explicit consent remains a limitation, highlighting the need for clearer guidelines on social media data use in research (Tanner et al., 2023). Moreover, machine-translated Korean YouTube comments introduce limitations, as the English-based “monologg/bert-base-cased-goemotions-original” model may not fully capture cultural nuances or emotional subtleties altered during translation. Errors in translating context-specific Korean expressions, such as honorifics or idioms, could affect the model’s emotion classification accuracy. To address this, a reliable translation tool was used, but future studies could employ Korean-specific models or bilingual datasets to enhance performance.
Future studies could extend this research by measuring informal second language engagement using Arndt’s (2023) questionnaire, alongside pre- and post-tests, to assess the impact of implementing informal English learning digital platforms in formal English classrooms. Furthermore, it illustrated that YouTube comments can serve as an authentic and effective interpersonal platform for accessing diverse sentiments. A key advantage of this sentiment analysis research is its ability to facilitate active engagement among commenters on informal platforms, which can be effectively integrated into formal educational settings (Lin, 2022). In an era where the boundaries between informal and non-formal education are becoming increasingly blurred, VCoP emerges as a powerful tool and may provide an easily accessible and voluntary environment for English education, offering a unique opportunity for learners to engage in language learning in a flexible and self-directed manner.
Notes
After the first mention of the two YouTube videos used in this paper, Oliver Ssam (2022) and Moon Ga Young (2023), we will omit the year and refer to them as Oliver Ssam and Moon Ga Young.