International Journal of Continuing Engineering Education and Life-Long Learning (9 papers in press)

Special Issue on: OA AI and Digital Technology driven Innovation in Continuing Engineering Education and Lifelong Learning Part Two

• Open Access

  Pathways for the digital transformation of ideological and political education in engineering education in the context of industry 4.0 ( Free Full-text Access )

  by Huibin Wei 
  Abstract: Industry 4.0 gives a new twist to the engineering professional competence framework. Digital transformation of engineering ideological and political education is not to simply add nonscientific factors but to reconstruct the dynamic balance of digital mechanism and ideological and political education. This paper constructs a four-dimensional driving spiral model to reveal the mechanism of technology empowerment. It designs a comprehensive ideological and political system to embed ethical competence throughout the entire cycle. Relying on three paths: scenario simulation, ethical sandbox, and agent platform, empirical measurements have shown that the efficiency of emotional transfer is 78%, the optimization rate of decision-making has increased by 75.5%, the transparency of the process has risen to 89%, and the comprehensive efficiency index has reached 106.8%. In response to tensions such as dataism and algorithm dependence, an innovative technical humanistic balance degree constraint index has been proposed.
  Keywords: Industry 4.0; engineering ideological and political education; digital transformation; four-dimensional driving model; ethical decision-making.
  DOI: 10.1504/IJCEELL.2026.10081318
   
  

Special Issue on: OA Immersive Technologies Redefining Education Transforming Learning in the Digital Age

• Open Access

  Using deep learning to analyse automatic grading and personalised feedback of student assignments ( Free Full-text Access )

  by Xuan Zhao 
  Abstract: Homework assessment faces challenges including lengthy scoring time, subjective variability, and lack of targeted feedback. This study constructs a Transformer-based deep learning model to overcome these limitations. The model uses a dual-channel encoder to extract deep semantic representations of student homework content and scoring criteria, with cross-channel attention enabling accurate alignment. A scoring sub-network based on knowledge-point sensitivity factors produces scores in a unified representation space. A conditional Transformer decoder then generates tailored feedback conditioned on these scores, integrating homework understanding, criterion alignment, scoring, and feedback generation. Results show a Pearson correlation of 0.930 for automatic scoring, an average error coverage index of 0.90 (most samples near 1.00) for feedback, and a BERT score of 0.89, indicating accurate weakness identification and semantically consistent feedback.
  Keywords: automatic rating; personalised feedback; deep learning; dualchannel transformer; semantic alignment; knowledge tracing.
  DOI: 10.1504/IJCEELL.2026.10080845
   
  

• Open Access

  Multidimensional application scenarios and key technical challenges of AI enabling digital transformation in distance education ( Free Full-text Access )

  by Ping Yang 
  Abstract: The rapid development of artificial intelligence (AI) technology is triggering a profound change in education, from text to content, and the application of AI technology in distance education is becoming increasingly widespread. This paper discusses the innovative applications of AI-based intelligent devices in distance education, including intelligent teaching assistants, personalised learning path design, intelligent assessment and feedback systems, and the construction of virtual experimental and practice environments. By analysing the existing research results and practice cases, this paper summarises the significant advantages of AI technology in enhancing the efficiency of distance education, optimising the learning experience, and promoting educational equity, and proposes future research directions and challenges. As the functions of AI devices are enriched and improved, they will be better suited to the characteristics of distance education and drive innovation in teaching scenarios.
  Keywords: artificial intelligence; distance education; smart devices; digitalisation of education.
  DOI: 10.1504/IJCEELL.2026.10081242
   
  

• Open Access

  5G and smart sports: an online sports teaching and training mechanism based on blockchain and edge computing ( Free Full-text Access )

  by Wenyu Zhang, Lihao Guan, Dan Huang 
  Abstract: The rapid development of the internet has promoted education reform, with online teaching gaining prominence, especially during the COVID-19 pandemic. This paper studies an online physical education (PE) system based on blockchain and edge computing (EC). It introduces the online teaching framework and smart sports concept, then proposes combining blockchain with edge technology. Simulation experiments compare task completion time and energy consumption. Results show that when task volume ranges from 800-1,800, the common method’s completion time drops from 4.6 s to 3.5 s, while mobile edge computing (MEC) remains stable at 4 s. For tasks of 500-1,500, MEC energy consumption stays within 9-10, whereas the common method fluctuates between 614, proving MEC’s greater stability. Thus, the blockchain-EC-based system enhances data processing and significantly improves online education effectiveness.
  Keywords: online sports; teaching and training; blockchain service; edge technology; edge computing; data security; mobile edge computing; MEC.
  DOI: 10.1504/IJCEELL.2026.10081264
   
  

Special Issue on: OA Intelligent Learning Ecosystems AI, Metaverse and Emerging Technologies for Continuing Engineering Education Part Two

• Open Access

  Intelligent English composition scoring method based on transfer paragraph segmentation and Pearson detection ( Free Full-text Access )

  by Haibei Chen 
  Abstract: With the growing demand for intelligent education, automated English composition scoring has gained significant research attention. Addressing limitations in cross-prompt generalisation and feature extraction, this study proposes a transfer learning-based method incorporating paragraph segmentation and Pearson detection. The approach combines BERT encoding, convolutional and bidirectional LSTM networks, augmented with attention mechanisms and cross-topic feature sharing. In performance evaluation, the proposed model achieved a weighted kappa coefficient of 0.866 and a Pearson correlation coefficient of 0.79 on the test set, significantly better than the comparison model in terms of consistency and correlation. It also showed lower prediction bias in error metrics, demonstrating higher fitting accuracy and rating stability. In three typical application scenarios, the model consistently outperformed the baseline method in terms of off topic detection accuracy and F1 score, and could more effectively capture semantic and structural features. In addition, in terms of resource consumption, the model parameter count was only 16.7M, the inference time was 28.9 ms, and the maximum processor utilisation rate was only 73.6%, demonstrating good deployment adaptability. The experimental results show that this method has good accuracy and generalisation ability in multi-scenario automatic scoring tasks, and has potential value for deployment and application in educational scenarios.
  Keywords: English composition scoring; transfer paragraph segmentation; semantic modelling; off topic detection; feature sharing.
  DOI: 10.1504/IJCEELL.2026.10079803
   
  

• Open Access

  Construction of an English academic writing teaching algorithm integrating LLM and move segmentation ( Free Full-text Access )

  by Yifang Ding, Xiaolong Ren 
  Abstract: English dominates global scientific research, but existing writing tools often fail to address disciplinary logic and conventions. Therefore, this study proposes an algorithm for assisting English academic writing teaching by integrating sentence-level segmentation with large language models. The research first uses a text feature extraction algorithm to extract local features of the text, achieving precise segmentation of academic text sentences. Then, through a thinking chain-constrained large language model, it generates writing templates, semantic error correction, and logical solutions for different disciplines. Finally, a multi-dimensional evaluation module is constructed using LightGBM to precisely fit the nonlinear relationship between text features and evaluation indicators, completing the objective quantification of text quality. Experiments achieved 0.9613 correlation with human scoring, with accuracy reaching 95.67% for empirical papers and 96.41% for reviews. The system attained 95.24% F1-score in STEM and 94.58% in humanities, while maintaining memory usage below 413.6 MB and response time under 6.27 s. This approach effectively addresses logical evaluation and feedback limitations in existing methods, offering comprehensive writing support for non-native researchers.
  Keywords: move segmentation; large language model; LLM; academic writing; teaching assistance; CNN-BiLSTM; LightGBM.
  DOI: 10.1504/IJCEELL.2026.10080251
   
  

• Open Access

  ITGAM: a co-attention and multi scale attention tandem convolutional network for multimodal sentiment analysis in English speech ( Free Full-text Access )

  by Yanling Wang 
  Abstract: This study proposes a composite structure based on the co-attention mechanism combined with a multi-scale attention cascaded convolutional network model for the multimodal sentiment analysis of spoken English. The experimental results show that the recognition accuracy of the six types of emotions ranges from 90.11% to 98.47%. The proposed model has excellent recognition and classification performance in the analysis of English oral multimodal emotion, and can effectively solve the data defects such as small sample size and missing modalities, which is helpful for the deep communication between intelligent machines and human beings.
  Keywords: multimodal sentiment analysis; co-attention mechanism; CoA; multi-scale attention tandem convolutional network; MSATC model; English speech; improved tensor fusion network; ITFN; generative adversarial network; GAN; speech emotion recognition.
  DOI: 10.1504/IJCEELL.2026.10080751
   
  

• Open Access

  English essay topic analysis using conceptual prior knowledge and Bi-GRU ( Free Full-text Access )

  by Bing Huang 
  Abstract: This paper addresses key challenges in English essay topic analysis, including cross-domain text variability, multi-topic integration, fine-grained topic discrimination, and limited adaptability to new topics. To overcome these issues, we propose a hybrid method integrating conceptual prior knowledge with an improved bidirectional GRU, enhanced by a knowledge-memory attention mechanism for dynamic knowledge selection and redundant information filtering. A CNN-LSTM module is further incorporated to strengthen semantic representation, while model-agnostic meta-learning and Bayesian optimisation with random forest construct the final adaptive framework. Experiments on student essay assessment and reading comprehension datasets show strong performance, achieving at least 93.28% local argument matching, 97.26% multi-label topic matching, 95.26% meta-test accuracy, 98.12% error attribution accuracy, and 98.41% topic hierarchy accuracy. The method also demonstrates superior efficiency, supporting accurate and practical English essay topic analysis for teaching feedback and automated multi-topic scoring.
  Keywords: conceptual prior knowledge; CPK; Bi-GRU; knowledge memory attention; English essay; topic analysis.
  DOI: 10.1504/IJCEELL.2026.10080201
   
  

• Open Access

  English machine translation based on parallel corpus mining and transfer learning ( Free Full-text Access )

  by Lipin Fang 
  Abstract: With the rapid development of neural machine translation, encoder-decoder models have become mainstream but still suffer from insufficient semantic alignment, limited long-distance dependency modelling, and inconsistent technical term translation. They often struggle with complex sentence structures and contextual understanding, affecting translation fluency and accuracy. This paper proposes an English machine translation model based on parallel corpus mining and transfer learning to enhance semantic representation and cross-lingual alignment. Experiments on a self-constructed Chinese-English parallel corpus show an error rate of 6.21, a semantic similarity of 98.22, and a technical term translation accuracy of 98.25%. The model achieves a translation latency of 63.52 s and an inference speed of 61.53 ms, demonstrating strong contextual consistency in long-text translation tasks. These results indicate that the proposed model effectively mitigates semantic deviation and improves contextual utilisation, offering a robust and efficient solution for cross-lingual natural language processing.
  Keywords: English translation; machine translation; parallel corpus mining; transfer learning; pretrained language model.
  DOI: 10.1504/IJCEELL.2026.10080200