Castledown Journals
Not a member yet
968 research outputs found
Sort by
High-accuracy, privacy-compliant multilingual sentiment categorization on consumer-grade hardware: A monte carlo evaluation of locally deployed large language models
No full textThis study presents a comprehensive evaluation of multilingual sentiment categorization performance using locally deployed large language models (LLMs) on consumer-grade hardware, focusing on GDPR-compliant implementation scenarios. Through extensive Monte Carlo validation involving 947,700 classifications over 702 iterations, we demonstrate significant performance capabilities across English, Italian, and Japanese languages while operating within consumer hardware constraints. Using lightblue/suzume-llama-3-8B-multilingual-orpo-borda-half on a Python-based llama-cpp framework on consumer NVIDIA GPU hardware, English achieved 96.3% accuracy (95% CI: 0.963–0.964), with Italian and Japanese showing strong performance at 92.2% (95% CI: 0.921–0.922) and 90.7% (95% CI: 0.906–0.908) respectively. Notably, our analysis demonstrates that plurality voting can achieve extremely high confidence levels across all languages, suggesting an efficient approach to improving classification reliability without requiring extensive computational resources
A systematic literature review on the use of 360-degree video-based virtual reality in English language education
No full textThis systematic review explores the use of immersive virtual reality (VR) based on 360-degree video technology in English language education, focusing on technological tools, pedagogical affordances, and their impact on language learning outcomes. Following the Preferred Reporting Items for Systematic Reviews and Meta-Analyses framework, 30 studies published between 2015 and December 2024 were analyzed. The findings indicate that 360-degree VR environments support vocabulary acquisition, reduce speaking anxiety, and foster learner engagement, especially in productive skills like speaking and writing, due to their interactive and contextual nature. However, results on listening skills were mixed, with some studies noting cognitive overload and distraction in immersive settings. Common tools included head-mounted displays, 360-degree cameras, and platforms like Google Tour Creator, which enabled realistic simulations. Despite these benefits, challenges such as high implementation costs, technical limitations, and the complexity of content creation were noted. Suggested solutions include the use of standardized assessment tools, clearer technical guidelines, and increased institutional collaboration. Overall, 360-degree VR holds considerable promise for enhancing English language learning, provided it is supported by effective pedagogical integration and adequate technical infrastructure
Language learning with Replika: Fostering grammar correction and conversational confidence
No full textThere are many beneficial English language learning tools. One of these tools is chatbots, powered by Artificial Intelligence (AI) models. Replika (https://replika.ai/) is an AI chatbot that communicates with users naturally, allowing learners to chat in English anytime and anywhere. It can be used in language learners’ speaking activities. It can also be used as a self-study tool for learners. Although Replika started as a chatbot and has age verification, learners can use it as a tool for learning how to communicate with others in English. In this research, three students used Replika for two weeks to see how their communicative competence changed through interacting with Replika. The study was conducted from the middle of October to the beginning of November 2021, and the second experiment was conducted at the beginning of February 2023. The researcher asked the students to engage in written chat with Replika for at least one hour weekly for a period of two weeks to see how Replika helped their communicative ability. This research was also conducted for one month. The author chatted with students before and after using Replika to determine any changes. The chat session was held two to three times in total. The author will propose how the chatbot affects learners’ language learning
ChatGPT or “CheatGPT”? EFL teachers’ perceptions of scoring AI-assisted English writing through the lens of assessment literacy
No full textThe rapid integration of artificial intelligence (AI) into education has reshaped teaching and learning practices, bringing unprecedented challenges to traditional approaches to writing assessment. Grounded in the three-dimensional framework of teacher assessment literacy, which encompasses the conceptual, praxeological, and socio emotional dimensions, this study explores nine Chinese EFL teachers’ perceptions of assessing AI-assisted writing assignments. Specifically, it focuses on issues of academic integrity, rubric compatibility, and criterion development in the AI era. Adopting a qualitative design, data were collected through semi-structured interviews to examine teachers’ beliefs, practices, and emotional stances toward AI-mediated assessment. The findings reveal that while teachers acknowledge AI’s potential to enhance learner engagement and higher-order thinking, they express persistent concerns about academic integrity and the inadequacy of existing scoring rubrics. Their attitudes toward AI-assisted writing are influenced by varying levels of digital literacy and pedagogical commitment, with most maintaining traditional scoring orientations that prioritize linguistic accuracy and originality. The study highlights the need to strengthen EFL teachers’ assessment literacy, particularly digital and ethical competencies—through professional development initiatives. It also underscores the necessity of updating institutional guidelines to ensure valid, fair, and pedagogically aligned writing assessment in the AI era
The relative importance of self-determined motivation for engagement in generative AI-supported English language learning: Insights from Japanese university students
No full textBoth motivation and engagement play pivotal roles in second language (L2) learning within the fields of applied linguistics and digital applied linguistics. While previous studies in applied linguistics have accumulated findings on the motivation-engagement nexus, there remains a paucity of empirical research in digital applied linguistics grounded in robust theoretical frameworks for understanding the comprehensive relationship of both constructs due to its relatively recent emergence. The current study addresses the gap by employing a self-determination theory to investigate the relationship between self-determined motivation and engagement in L2 learning with the use of generative artificial intelligence (GenAI). Data were collected using a questionnaire from a total of 98 university students learning English as a foreign language in Japan. Results from correlation analysis showed that autonomous forms of regulation were strongly associated with three dimensions of engagement (i.e., behavioral, cognitive, and emotional). Furthermore, a series of multiple regression analysis with supplemental use of dominance analysis and random forest plots revealed that autonomous motivation accounted for nearly 80% of the variance in behavioral, cognitive, and emotional engagement, respectively. The findings highlight the key role of autonomous motivation, especially identified regulation, in facilitating active participation in L2 learning with GenAI. This article concludes with theoretical and pedagogical implications, emphasizing the viability of SDT in conceptualizing the motivation-engagement nexus in GenAI-supported L2 learning and the critical roles of teachers in helping students recognize the value of L2 learning with GenAI
Online interaction learning: A way towards sustainability and buoyant EFL students
No full textThe ability of students to deal with common academic difficulties is referred to as academic buoyancy. It can also be described as the capacity of students over obstacles and difficulties that are typical of the everyday school experience. The purpose of this research was to examine the connections between the academic buoyancy of Iranian university-level EFL students and their online interaction learning model. Online interaction learning seems to be a strong predictor of academic buoyancy. Academic buoyancy surveys and the online interaction learning model were given to students majoring in English translation and teaching English as a second, third, and final year of college. Both structural equation modeling and correlational studies found a statistically significant association between academic buoyancy, online interaction learning, and all of their components. The best predictors of academic buoyancy, according to regression analysis, were the components of course material and instructor performance. Taken together, the findings add to what is widely accepted about the online engagement tendency of university students and sheds light on the academic buoyancy constructs in higher education. However, more research on the role of buoyancy in relation to achievement and online tactics to increase individual students’ achievement and overall online interaction learning model is warranted based on the findings of the present study
Exploring ethnic identity and heritage language proficiency among second-generation Hausa Saudis
No full textThis article reports a part of a larger study which examines the sociolinguistic dynamics within the Hausa community in Saudi Arabia. It focuses on how second-generation members of the Hausa Saudi community perceive their ethnic identity and investigates the relationship between their proficiency in the Hausa heritage language and their sense of ethnic identity. Data were collected through an online questionnaire completed by 103 participants. The findings reveal that participants reported moderately high levels of ethnic identity (M = 3.14, SD = 0.63). Responses regarding ethnic self-identification indicated a strong preference among participants for identity labels that emphasize their Hausa family roots and their "Arabness" (e.g., Hausa Arab, Arab Hausa). These labels illustrate how individuals negotiate the relationship between their Hausa origin and the predominantly Arabic-speaking environment in which they were born and currently reside and identify. Correlation analysis showed a weak yet significant positive relationship between Hausa language proficiency and ethnic identity (r = .24, p < .05). The conclusion suggests that while heritage language contributes to shaping ethnic identity, other factors, including ancestry, race, and religion, may also play pivotal roles
Synthesis of current research on the affective dimensions of online English language education: Theories and praxis
No full textThe affective domain can impact the way teachers teach and students’ motivation and engagement in learning in online contexts. This study focuses on the affective dimensions of online English language education by offering a comprehensive overview of the diverse types of emotions experienced by both students and teachers in online English language environments in terms of (1) the types of emotions and participants studied in specific virtual environments, (2) the theoretical approaches and pedagogical practices for studying emotions in online English language education, and (3) the methodological design features of the studies. Thematic analysis of the 40 selected studies highlighted these feelings as the focus, namely boredom, emotion regulation, emotional labour, emotional intelligence, stress, emotions towards L2 learning, foreign language enjoyment (FLE), FLE and foreign language anxiety/foreign language classroom anxiety, foreign language speaking anxiety, and positive and achievement emotions. The findings revealed the use of established theoretical frameworks such as control-value theory, the emotion regulation scale, the FLE and foreign language anxiety scales and other instruments, as well as questionnaires, interviews and multiple instruments to study emotions in online English language education. Educators can consider the emotionally resilient pedagogical practices identified in this study such as emotional scaffolding strategies and technologyassisted language learning or mobile-based language learning for emotion regulation in teaching and learning English online
EFL learners’ emotional challenges and the associated coping mechanisms in online and hybrid learning contexts: A narrative inquiry
No full textThe COVID-19 pandemic necessitated an abrupt transition from traditional classroom settings to online learning, which had a profound impact on learners' emotional experiences. This study explored the emotional challenges, particularly anxiety and frustration, faced by university English as a Foreign Language (EFL) learners during this period and their subsequent transition back to hybrid or traditional settings. The study employed narrative inquiry with 33 undergraduate EFL learners, followed by semi-structured interviews with four participants in a state university context in western Iran. The results revealed five key themes, including emotional responses to technology challenges, development of coping strategies, impact of external factors, instructor influence on students, growth in confidence and adaptation. The findings highlight how students gradually transformed initial negative emotions into resilience and confidence through coping strategies. The study underscores the critical role of supportive instructor behavior and well-designed hybrid learning environments in reducing emotional challenges and fostering adaptability. The results emphasize the need for institutional support to address learners' emotional and technological needs, which ensure effective transitions in evolving educational contexts
Revisiting Hong’s (2010) model of CALL integration through conceptual replication
No full textThe integration of computer-assisted language learning (CALL) has held increasing interest among the community over the last three decades. One paper which has taken stock of this interest and modelled the factors that impact integration has been written by Hong (2010). While Hong’s model has significantly contributed to understanding technology integration in language classrooms, several methodological shortcomings limit its validity. Specifically, Hong’s (2010) reliance on a limited number of studies and the lack of systematic evidence to support the centrality of CALL teacher education raise concerns about the model’s robustness. Furthermore, the literature review process used to position the individual and contextual factors remains unclear, leaving the strength of the evidence base uncertain. Given these shortcomings, this study conceptually replicated the literature review and subsequent spherical model from Hong (2010) to determine the validity of this model and understand the current complexities of CALL integration. The conceptual replication was based on 138 studies from eight Q1-ranked CALL journals. From this literature, we found that Hong’s three spheres: CALL teacher education, individual teacher, and contextual factors, continued to represent the majority of CALL integration factors validly. However, we found a much broader range of relevant sub-factors within these three broad spheres. For CALL teacher education factors, we found teachers desired pedagogically-linked training. For individual teacher factors, time management was a pertinent factor. For contextual factors, we found an intricate split between institutional and country-specific factors and a mixed picture of country and institutional responsibility. This latter mixed picture included inflexible teaching methods