Conversation Breakdown and Institutional Discourse in Ghanaian ESL Classrooms: A Conversation Analytic Investigation

Francis Bukari — 2026-01-15

This study responds to the growing pedagogical interest in optimizing communicative competence within English as a Second Language (ESL) classroom interaction by investigating a critical, yet under-explored, domain, the interactional trouble sources that initiate conversation breakdown. Grounded in a Conversation Analytical (CA) framework, the research methodology utilizes a hybrid approach: CA modeling for the micro-analysis of recorded data and content analysis for the qualitative interview data. The empirical base consists of 52 h of recorded ESL classroom discourse extracted from the Ghana Senior High School corpus of academic spoken English database collected by the researchers and research assistants, and augmented by interviews with practicing ESL teachers. A systematic analysis of the interactional sequences showed a pronounced presence of both etic (analyst-defined) and emic (participant-oriented) conversational trouble sources. The findings delineate six salient categories of trouble sources, namely, mishearing/non-hearing, vagueness, topic transition, information deficit, and lexical inappropriacy. These trouble sources demonstrably impeded interactional flow. Notably, the research establishes that the origins of these trouble sources are multi-layered, transcending mere surface-level linguistic (phonology, syntax, lexis) deficiencies to include institutional factors such as instructional ambiguity, procedural misalignments, disciplinary actions, and culturally situated vocabulary choices. This evidence mandates that future ESL research accord greater significance to the impact of institutional discourse (especially, classroom discourse) features as a primary generator of interactional trouble.

Detecting Student Inattention Using Deep Learning and Behavioral Analysis

Fatima Zedan — 2026-01-07

Student inattention in classrooms negatively impacts learning outcomes and academic performance, posing a significant challenge for educators. Traditional methods of monitoring engagement rely on subjective teacher observations, which can be inconsistent, labor-intensive, and prone to bias. To address these limitations, this paper presents an AI-driven framework that uses deep learning and behavioral analysis to detect student inattention in real time. The proposed system integrates computer vision techniques including facial expression recognition, posture analysis, head pose estimation, and eye-gaze analysis, employing convolutional neural networks (CNNs) to extract spatial features and recurrent neural networks (RNNs) to model temporal patterns. The framework was evaluated using annotated classroom video data collected from real teaching sessions, capturing natural student behavior under typical classroom conditions. Experimental results demonstrate that the proposed approach achieves high accuracy in distinguishing attentive from inattentive states, outperforming traditional machine learning baselines while maintaining real-time performance. Beyond detection, the system provides actionable insights for educators by highlighting patterns of disengagement across time and students. By combining CNN-based spatial analysis with RNN-based temporal modeling, the framework offers an objective, scalable, and practical solution for monitoring classroom engagement, enabling timely interventions, personalized instruction, and improved learning outcomes.

Journal of International Education and Practice

Conversation Breakdown and Institutional Discourse in Ghanaian ESL Classrooms: A Conversation Analytic Investigation

Detecting Student Inattention Using Deep Learning and Behavioral Analysis