Multiple Languages Research Articles

A corpus-assisted discourse study of parental concerns regarding multilingual child-rearing

ABSTRACT Many parents have concerns about raising their children with multiple languages. However, there is a paucity of previous research regarding parental concerns about multilingual child-rearing, particularly in multilingual societies. We address this gap with a corpus-assisted discourse study of parental concerns regarding multilingual child-rearing in Quebec, Canada. We created two corpora comprising 641 parents’ responses to an open-ended survey question regarding their main concerns about raising multilingual infants/toddlers (English corpus: 12,502 words, French corpus: 9,360 words). We examined frequencies, collocations, concordance lines, and longer segments to investigate the nature and strength of different concerns. Our results revealed that two previously-attested concern types – cognition concerns and exposure-fluency concerns – were most prominent. The results also provided more nuanced insights into the nature of these concerns. Moreover, the results revealed two additional concern types: concerns regarding trilingual/heritage language transmission and concerns about the effect of multilingualism on children’s identity and social/emotional well-being. These had not previously been attested. Our research makes a theoretical contribution by advancing knowledge about parental concerns and how they can contribute to the study of family language policies. Additionally, our findings may serve as the basis for improving support for multilingual families.

Repertoire by Women Composers: A Guide for Beginning and Intermediate-Level Singers

Abstract: Although vocal music has one of the longest recorded histories of any art form, the songs of women composers remain largely excluded from the traditional canon. Recently, performers and educators have been working to change the way we decide which music and musicians are part of the classical repertoire. This article will address repertoire by women composers to suit early-stage singers in the first years of study. Repertoire presented will explore multiple languages, time periods, and styles within the Western-classical pedagogic tradition. The authors offer insight into each selection’s benefits. Online and traditional resources are suggested for further exploration.

Multilingual Realities and English Teacher Construction in Indonesian Pesantren: A Narrative Inquiry

This study aimed to explore the construction of English teacher identity within the multilingual and religiously oriented environment of Indonesian pesantren through a narrative inquiry approach. Applying narrative inquiry, the study examined how teachers negotiate their religious and professional identities while balancing multiple language ideologies embodied in this educational context. Data were collected through in-depth semi-structured interviews and reflective journals, providing rich narratives of teachers' experiences, challenges, and strategies in navigating their roles. Three key themes emerged: balancing language ideologies, the interplay between religious and professional identities, and multilingual practices in the classroom. The findings reveal the dynamic, context-responsive nature of teacher identity development within pesantren, highlighting how educators align religious values with pedagogical goals. This study addresses a critical gap by focusing on identity formation in faith-based multilingual education settings, offering new insights beyond secular or formal institutions. The implications contribute to broader discussions on teacher identity, multilingual education, and the intersection of language, religion, and teaching practices.

Challenges and Techniques in Multilingual Fake News Detection

The spread of false information across language and cultural barriers poses a significant threat to global societies. This study investigates the obstacles in detecting misinformation in various languages, including linguistic diversity, limited datasets, cultural nuances, and the ever- changing strategies of disinformation propagators. These complex issues require advanced methods to effectively identify and combat false information across multiple languages. The research evaluates several prominent machine learning approaches, such as Support Vector Machines (SVM), Random Forest, K-Nearest Neighbors (KNN), Logistic Regression, Decision Trees, and Natural Language Processing (NLP) techniques. By examining their effectiveness in multilingual contexts, this study sheds light on the advantages and drawbacks of each method. It underscores the need for continuous innovation and collaborative efforts in developing robust detection systems that can adapt to the ever-changing landscape of misinformation. These initiatives aim to promote informed public discourse and improve information integrity across diverse cultural contexts. Key Words: Multilingual Fake News Detection, Machine Learning, Machine Learning algorithms, Natural Language Processing, Misinformation.

Part C Early Intervention Procedural Safeguard Notices: Are They Supporting Parents to Understand Their Rights?

A family-centered early intervention (EI) approach mandated by the Individuals with Disabilities Education Act (IDEA) Part C, means involving parents fully in EI and ensuring they know their legal rights and responsibilities for participating. However, many parents report difficulty accessing information and decision-making in EI. IDEA mandates all jurisdictions provide parents with a procedural safeguard notice (PSN) outlining parents’ rights and advocacy processes. Yet, it is unclear to what extent PSNs are accessible, particularly to parents with marginalized identities. This study analyzed PSNs from 56 Part C jurisdictions for the components required by law, understandability to the general public, and availability in multiple languages. The PSNs generally did not include all required information, necessitated grade-levels above high school, minimally attended to plain language, and were infrequently available in non-English languages, suggesting that many parents are prevented from information about EI. PSNs, therefore, are a source of bias in Part C.

Reverse Transfer in Chinese Adolescents: Implications for Language and Cultural Identity Preservation

This literature review examines the phenomenon of reverse transfer (RT) among Chinese adolescents, focusing on its implications for the preservation of their mother tongue and cultural identity. Reverse transfer, a subtype of cross-linguistic influence (CLI), occurs when a later-acquired language affects an earlier one, challenging the traditional view of language learning as unidirectional. The study of RT is particularly pertinent for Chinese adolescents, who are often exposed to multiple languages due to globalization and educational policies, potentially impacting their linguistic and cultural heritage. This article employs the method of a formal literature review. The review synthesizes recent research on RT among Chinese adolescents, with a focus on how additional languages, particularly English, influence their Mandarin proficiency. Findings suggest English has had a significant impact on the mother tongue of Chinese adolescents. Moreover, RT’s impact extends to cultural identity, with some adolescents navigating multicultural spaces effectively, while others experience cultural dissonance, suggesting RT’s role in shaping cultural self-perception and connections to Chinese heritage. The paper underscores the complexity of RT’s effects on language and cultural identity, highlighting the need for educational practices that support multilingual development without compromising the mother tongue. It concludes with a call for longitudinal studies to explore the long-term effects of RT on language and cultural identity among Chinese adolescents, aiming to inform language education and cultural preservation policies.

Long-term Survival After Cancer: A Scoping Review of the Health Care Landscape in Germany

Abstract Background Cancer remains a global health concern, with a significant increase in new cases projected in the coming years. Although cancer treatment and diagnosis have advanced, leading to improved survival, survivors often face complex physical, emotional, and cognitive challenges. The objective of this scoping review is to provide an overview of the current landscape of specialized long-term survivorship care services in North Rhine-Westphalia (NRW) with a focus on vulnerable groups. Methods The scoping review involves two search approaches: a manual search performed by two review authors and a systematic search performed by an information specialist following Scoping Review guidance. The protocol was published on the OSF platform. The systematic search was conducted in databases such as LIVIVIO, MEDLINE and Scopus. The manual search involves identifying key categories of survivorship services and conducting online searches based on these categories in duplicate. Data extraction and synthesis is conducted using EPPI-Reviewer software, and focuses on services characteristics, patient demographics, program format, and accessibility. Results The study has identified more than 300 survivorship care services across North Rhine-Westphalia. According to preliminary results some services offer comprehensive care, while others have significant gaps, particularly in addressing the needs of vulnerable groups. One example of these gaps is the lack of tailored services specifically designed for people with a migration background. Most of the services are not available in multiple languages or gender-specific oriented. The results will be presented on the congress in detail. Conclusions This scoping review underscores the importance of tailored survivorship care, especially for vulnerable groups. Public health actions should focus on addressing disparities in access to survivorship services, enhancing existing programs, and developing new policies to support cancer survivors. Key messages • Effective survivorship care requires attention to vulnerable groups. • The care of long-term cancer survivors should receive more attention in research.

The affordances of plurilingual speaking games for recognizing, using, and developing plurilingual competences of children and pre-service teachers

ABSTRACT Games have long been a common activity integrated by teachers in many language learning classrooms worldwide. The change in focus from the “four skills” of speaking, listening, writing, and reading to “four modes of communication”: “reception,” “interaction,” “production,” and “mediation” in the Common European Framework Companion (2018) means that teachers need to incorporate Mediation language activities, such as translation or interpretation, that fit with this shift. This study explores the perceived affordances of playing plurilingual, speaking games for children and pre-service teachers (PSTs) in a joint university-primary school project in Spain. The project aimed to support the development of plurilingual and intercultural competence of 21 pre-service teachers and 23 children. Using a case study approach, incorporating thematic analysis, teachers’ written reflections are analyzed. Findings highlight that the games afforded the learning of languages using and awareness of languages; making children’s multiple languages “visible”; boosting enjoyment, collaboration, motivation, and risk-taking which facilitated a change in children’s social roles to “teacher.” Affordances for pre-service teachers included developing pedagogical know-how for working in a multilingual classroom and awareness of children’s linguistic repertoires. Considerations for game design and research are presented.

Co-producing COVID-19 health education resources with ethnic minority communities in the UK

Abstract Background Effective communication of relevant health information is challenging in elderly minority ethnic groups due to language barriers and cultural differences. During the COVID-19 pandemic, much of the public health guidance around COVID-19 was in dominant languages, lacking cultural nuance. This study aimed to co-produce culturally, linguistically, and age-appropriate health education resources in COVID-19, tailored to the needs of older adults from ethnic minority backgrounds. Methods This study involved three stages: 1) Semi-structured interviews with Indian and Nepalese older adults (≥65 years), their families (≥18 years) and healthcare professionals (HCPs) (≥18 years) engaging with these communities; 2) Co-production of COVID-19 information resource, and c) Implementation of co-produced resources among the communities. Results 24 participants took part in the study. There was wide variation in participants’ knowledge of COVID-19; some were knowledgeable, while others had limited information. Participants highlighted the need for information in multiple formats and languages and discussed the importance of culturally avenues for information dissemination. Drawing on these findings and two co-production workshops held with the stakeholders, culturally sensitive information leaflets (electronic and paper) were co-prodcued in Hindi and Nepalese using age-appropriate images and infographics. The leaflets were distributed widely to the Nepalese and Indian communities, healthcare trusts and local authorities using the networks of community members, HCPs, and religious leaders. Conclusions The project highlights the significance of participatory methods in the development of culturally relevant public health interventions applicable to the cultural practices of ethnic minority groups. The co-production approach in this study could influence information provision for future public health campaigns for other marginalised groups. Key messages • Health information should be co-produced with a representative sample of the community members and relevant stakeholders. • There is a need on an ongoing basis to identify and fill the information gaps, address misinformation and develop targeted messages.

Validated translation of telehealth and health app usability questionnaires to 8 European languages

Abstract International projects such as the Joint Action on strengthening eHealth including telemedicine and remote monitoring for health care systems for cancer prevention and care (eCAN) are implemented in countries with different languages, requiring validation of translations of standardized tools. Evaluation frameworks for telemedicine services necessarily include the patient’s perspective, i.a. on the service usability. Due to the lack of validated questionnaires in multiple European languages we sought to develop translation for the languages of the eCAN pilot countries. Based on literature review, we selected mHealth App Usability Questionnaire (MAUQ) and Telehealth Usability Questionnaire (TUQ) and developed a conceptual definition for each item. Next, we performed forward and back translation from English into the target languages, expert review, cognitive interview (CI) with 5-10 cancer or chronic disease patients per language/per questionnaire. Qualitative CI results were scored by 2 independent researchers for consistence of each item with conceptual definitions for final review by native speaker expert. We chose 8 languages (ES, GR, HU, IT, LT, NL, PT, SI) for validation. CI involved 48 respondents (29 females, 60%; 18 aged>60, 38%;23 higher education, 48%) for TUQ and 48 (22 females, 46%; 16 aged>60, 33%; 23 higher education, 48%) for MAUQ. Overall 872 (87%) of TUQ and 728 (84%) of MAUQ item answers were classified as consistent. In more than one language we identified concepts difficult to understand: consistent navigation when moving between screens, distinction between easy to use/learn, social settings. The validation process allowed us to identify opportunities to improve translation and highlighted concepts generally more difficult to understand. We provided validated translation of MAUQ and TUQ in 8 European languages for use in eCAN pilots and future telemedicine evaluation studies. Key messages • eCAN successfully validated translations of telemedicine usability questionnaires for 8 languages. • Multilingual translation validation can provide insights into concepts difficult across languages.

Development of a risk-stratified shared decision-making tool for colorectal cancer screening

Abstract European guidelines stress the importance of informed decision-making in colorectal cancer (CRC) screening. However, tailored tools that facilitate decision-making for underserved people and general practitioners (GPs) in Flanders are lacking. This study aimed to develop a tailored shared decision-making (SDM) tool for CRC screening, specifically a video decision aid, using a co-creative process. The development involved two rounds of focus group discussions with underserved people and five local quality group meetings with GPs in various parts of Flanders, Belgium. Qualitative data were analyzed using a Modified-Grounded Theory Analytical (M-GTA) approach to identify key themes. Findings revealed four core themes. The first theme, ‘Recognizing the Barriers to Screening’, identifies challenges such as a lack of time, staff, and tools for GPs, as well as language and literacy barriers for underserved people. The second theme, ‘Learning from Existing Tools’, highlights successful elements like visuals and simple language and limitations such as low accessibility and poor navigation. The third theme, ‘Determining the Content and Design,’ reveals that GPs seek user-friendly tools for at-home viewing, while underserved people value personalized and clear information about CRC risk factors and screening costs. The final theme, ‘Identifying Strategies for Integration,’ stresses the importance of a structured, collaborative care approach for the successful integration of the tool. The proposed modifications for the SDM tool include enhancing its accessibility with multiple language options, using simple and relatable visuals, and providing personalized risk discussions for underserved people and GPs. Nonetheless, further testing is needed to confirm the effectiveness of the SDM tool in supporting preventive talks and informed decision-making in CRC screening among these target groups. Key messages • A structured, collaborative care approach is needed for successful SDM tool integration during preventive talks with underserved people. • Further testing is needed to confirm the tool’s effectiveness in supporting informed decision-making in CRC screening for underserved people.

How to measure the general Environmental Health Literacy

Abstract Background Many environmental factors, both natural or man-made, are associated with health risks. To make informed decisions about these potential risks, certain skills and competencies are necessary. The concept of Environmental Health Literacy (EHL) which integrates aspects of Science Literacy, Health Literacy and Environmental Literacy encompasse these competencies. To assess to what extent individuals or groups are environmental health literate, a general Environmental Health Literacy Scale is necessary. To the best of our knowledge, there is no validated instrument to measure general EHL. Methods Two phases of scale construction were conducted: 1) Item development with deductive methods which includes a literature review and assessment of existing scales resulting in two suitable studies for scale construction. Included items were translated to German language. 2) Scale development with pre-testing of items followed by applying potential items (n = 38) to a sample of of 454 students. Internal consistency (Cronbach’s alpha), Kaiser-Meyer-Olkin Criterion (KMO), Bartlett’s test, Velicer’s minimum average partial (MAP) test for number of underlying factors and Exploratory Factor Analysis using Principal Component Analysis with oblimin rotation and stepwise item reduction were assessed. Results Velicer’s MAP identified 3 factors, Cronbach’s alpha = .92, KMO = .92, Bartlett’s Test revealed significant correlation. Seven items were removed from the scale due to low factor loadings (< .3). Factor 1 (access) explains 18% of variance, factor 2 (knowledge) 12% and factor 3 (reflection) 10%, respectively. Overall, the model explains 40% of variance. Model fit represented by RMSEA index = .05. Conclusions A new general EHL scale has been developed and proven to be suitable for a students sample. This marks the first step towards a validated tool for multiple purposes. Cultural adaptation as well as multiple language validations are in process. Key messages • The process towards a general Environmental Health Scale has started. • The development of a validated EHL scale will improve understanding on whether the general population or subgroups are able to gain, understand and act upon information regarding environmental health.

GestureGPT: Toward Zero-Shot Free-Form Hand Gesture Understanding with Large Language Model Agents

Existing gesture interfaces only work with a fixed set of gestures defined either by interface designers or by users themselves, which introduces learning or demonstration efforts that diminish their naturalness. Humans, on the other hand, understand free-form gestures by synthesizing the gesture, context, experience, and common sense. In this way, the user does not need to learn, demonstrate, or associate gestures. We introduce GestureGPT, a free-form hand gesture understanding framework that mimics human gesture understanding procedures to enable a natural free-form gestural interface. Our framework leverages multiple Large Language Model agents to manage and synthesize gesture and context information, then infers the interaction intent by associating the gesture with an interface function. More specifically, our triple-agent framework includes a Gesture Description Agent that automatically segments and formulates natural language descriptions of hand poses and movements based on hand landmark coordinates. The description is deciphered by a Gesture Inference Agent through self-reasoning and querying about the interaction context (e.g., interaction history, gaze data), which is managed by a Context Management Agent. Following iterative exchanges, the Gesture Inference Agent discerns the user’s intent by grounding it to an interactive function. We validated our framework offline under two real-world scenarios: smart home control and online video streaming. The average zero-shot Top-1/Top-5 grounding accuracies are 44.79%/83.59% for smart home tasks and 37.50%/73.44% for video streaming tasks. We also provide an extensive discussion that includes rationale for model selection, generalizability, and future research directions for a practical system etc.

How COVID-19 pandemic and Ukrainian crisis shaped mental health services for medical students in Poland?

The mental health of medical students is a crucial concern in contemporary medical education. This manuscript reports the availability and shape of psychological support for medical students in Poland in the backdrop of the COVID-19 pandemic and the Ukrainian crisis. A survey comprising 10 questions was distributed to 22 medical schools in Poland, with a 59% response rate. The findings reveal that psychological support programs are predominantly managed by internal university units. Funding sources vary, with some universities utilizing internal budgets and others leveraging European funds or external projects. The support modalities are primarily hybrid, catering to student preferences. Limitations exist in the number of consultations, although some universities provide extended support based on student needs. Assistance is offered in multiple languages, with additional programs like stress-reduction workshops and mindfulness training being available in several universities. A notable increase in demand for psychological support post-COVID-19 was reported, with some universities establishing their programs during or after the pandemic. The commentary highlights the rising need for mental health services among medical students and underscores the importance of flexible, inclusive, and well-funded support. The effective functioning of these programs aligns with the broader objective of fostering a resilient and emotionally balanced healthcare workforce.

Implementation of the Telemetric Integration of the BIM-RFID in Context of Access Control

Building Information Modelling (BIM) integration with the Internet of Things (IoT) is progressing. The high level of BIM maturity involves using sensor data to manage processes or objects. The article presents the process of creating a telemetry connection between the BIM model and a Radio-Frequency Identification (RFID) sensor in the context of gaining access to various parts of a building. The process of creating a connection using an experimental set based on a microcontroller board for RFID reader support is described. The set was programmed using multiple programming languages and artificial intelligence. The article presents a unique process of connecting an RFID reader with BIM using a simple model that can be replicated in other contexts (e.g., gaining access to different parts of a construction site). Previous research shows that the unidirectional connection of IoT sensors with BIM models is not difficult. Instead, real-time bidirectional and stable connection (telemetry) is problematic. The authors undertook to fill this research gap using a proprietary IoT kit, programming sequence, and lightweight communication protocol. The paper makes a significant contribution to the discussion and understanding of BIM-IoT technology integration. The article also includes the limitations and possibilities for further development of such a connection.

A Review of Integrated Curriculum Models in Language Teaching

Integrated curriculum models in language education aim to combine multiple disciplines and language skills into a cohesive framework, fostering connections between language learning and real-world contexts. While these models offer a holistic approach, their implementation has raised concerns about effectiveness, teacher readiness, and challenges in cross-subject integration. This review evaluates the current literature on integrated curriculum models in language teaching, examining their impact on language proficiency and learner engagement. The study reviews empirical studies, case reports, and theoretical discussions, focusing on both qualitative and quantitative research that assesses outcomes in language classrooms. Key themes include cross-curricular integration, teacher collaboration, and the influence of cultural and contextual factors. Findings suggest that integrated curriculum models improve language proficiency by encouraging language use in diverse contexts, helping learners develop both language and critical thinking skills. However, successful implementation requires significant teacher collaboration and curriculum alignment. Teacher preparedness and institutional support are also crucial for achieving positive outcomes. In conclusion, integrated curriculum models hold great potential to enhance language education, but their success relies on effective implementation and collaboration between educators. Further research is needed to assess the long-term effects and scalability of these models in different educational settings.

Evidence of community structure in phonological networks of multiple languages.

Thousands of phonological word forms known to a speaker can be organised as a lexical network using the tools of network science. In these networks, nodes represent words and edges are placed between phonological neighbours. Previous work has shown that phonological networks of various languages have similar macrolevel network properties. The present study aimed to investigate if phonological networks of different languages also have similar mesolevel properties, specifically, the presence of robust community structure. Prior community detection analyses revealed robust community structure for English. Community detection analyses conducted on French, German, Dutch, and Spanish networks indicate that all networks showed strong evidence of community structure-mesolevel clustering of word forms whereby larger communities tended to contain shorter, frequent words with many phonological neighbours. Words of the same community tended to share similar phonotactic structures. Results suggest that the organisation of phonological word forms in language are governed by similar principles that could have important implications for lexical processing. (PsycInfo Database Record (c) 2024 APA, all rights reserved).

JST and rJST: joint estimation of sentiment and topics in textual data using a semi-supervised approach

ABSTRACT This paper demonstrates the performance of the Joint Sentiment Topic model (JST) and the reversed Joint Sentiment Topic model (rJST) in measuring sentiment in political speeches, comparing them against a set of popular methods for sentiment analysis: widely used off-the-shelf sentiment dictionaries; an embeddings-enhanced dictionary approach; Latent Semantic Scaling, a semi-supervised approach; and a zero-shot transformer-based approach using a large language model (GPT-4). The findings reveal JST’s superiority over all non-transformer-based approaches in predicting human-coded sentiment in multiple languages and its ability to replicate known sentiment trends in legislative speech. rJST, meanwhile, provides valuable topic-specific sentiment estimates, responsive to political dynamics and significant events. Both models are, however, outperformed by transformer-based models like GPT-4. Additionally, the paper introduces the ’sentitopics’ R-package, designed to facilitate the use of JST and rJST in computational text analysis workflows. This package is compatible with popular text analysis tools, making the models accessible for applied researchers in communication science.

Validity, test-retest reliability, and responsiveness of the Indonesian version of FACIT-COST measure for subjective financial toxicity.

Financial toxicity describes the impairment of financial wellbeing in patients due to the burden of cancer diagnosis and care. The COST: A Functional Assessment of Chronic Illness Therapy Measure of Financial Toxicity (FACIT-COST) is the most widely used cancer-specific measure of subjective financial toxicity, having been validated in multiple languages, but not in Indonesian. This study aimed to validate the Indonesian version of FACIT-COST in a breast cancer sample. A single-center prospective cohort study was performed in Indonesia. Female breast cancer patients aged ≥ 18 undergoing treatment at baseline were invited to participate and followed for up to six months. The survey included the official Indonesian version of FACIT-COST (v2) which was administered to the patients by interviewers. Clinical information (e.g., metastasis status, disease duration) was provided based on medical records. The following measurement properties of FACIT-COST were tested: distributional characteristics, structural validity (principal component [PCA] and confirmatory factor analyses [CFA]), internal consistency reliability (Cronbach's alpha and McDonald's omega), known-groups validity (Mann-Whitney U or Kruskal-Wallis H test), test-retest reliability, and responsiveness to change. Overall, 300 female patients participated at baseline. No patients reported the best or worst possible FACIT-COST total scores. The PCA proposed a two-factor model structure for the instrument, which was confirmed by the CFA (RMSEA = 0.042, SRMR = 0.049, CFI = 0.99, TLI = 0.99). The internal consistency reliability of the two factors was considered adequate (Cronbach's alpha = 0.774-0.882, McDonald's omega = 0.786-0.888). The FACIT-COST total score significantly discriminated across the following known-groups: age, education, residential setting, income, employment, metastasis status, number of symptoms, and financial coping strategies. The FACIT-COST demonstrated excellent test-retest reliability (intraclass correlation coefficient = 0.96) and satisfactory responsiveness to change (standardized response mean and effect size ranges=|0.39| to |0.92|). This is the first study to validate the FACIT-COST in patients with breast cancer and to present the measurement properties of the Indonesian version of FACIT-COST. The Indonesian FACIT-COST demonstrates acceptable psychometric performance and shows potential as a valid measure of subjective financial toxicity. The instrument may serve as a valuable tool for informing health policies that focus on providing resource support to improve cancer care in Indonesia.

Quantifying the impact of ASR-based instruction: What does the iSpraak platform learner data show?

Computer-assisted Pronunciation Training (CAPT) tools have become increasingly dependent on Automatic Speech Recognition (ASR) technology to provide automated corrective pronunciation feedback to learners. The extent to which ASR-based tools measurably improve second language (L2) pronunciation is of great interest to language educators globally, and Computer-assisted Language Learning (CALL) researchers. Studies to date have largely been conducted by research practitioners with small-to-medium sized samples at single institutions. The findings and conclusions drawn from such small-scale data collection might be significantly bolstered by analysing the vast stores of learner data from large CAPT platforms. This study is informed by a sizable eight-year dataset from iSpraak, an open-source pronunciation tool designed to model and evaluate L2 speech. Quantitative analysis of anonymised learner interactions with this application reveals significant gains in intelligibility measures across multiple languages. Results also suggest that the extent of ASR’s ability to improve learner pronunciation may be L2 dependent.