Sentiment evolution is a key component of interactions in blended learning. Although interactions have attracted considerable attention in online learning contexts, there is scant research on examining sentiment evolution over different interactions in blended learning environments. Thus, in this study, sentiment evolution at different interaction levels was investigated from the longitudinal data of five learning stages of 38 postgraduate students in a blended learning course. Specifically, text mining techniques were employed to mine the sentiments in different interactions, and then epistemic network analysis (ENA) was used to uncover sentiment changes in the five learning stages of blended learning. The findings suggested that negative sentiments were moderately associated with several other sentiments such as joking, confused, and neutral sentiments in blended learning contexts. Particularly in relation to deep interactions, student sentiments might change from negative to insightful ones. In contrast, the sentiment network built from social-emotion interactions shows stronger connections in joking-positive and joking-negative sentiments than the other two interaction levels. Most notably, the changes of co-occurrence sentiment reveal the three periods in a blended learning process, namely initial, collision and sublimation, and stable periods. The results in this study revealed that students’ sentiments evolved from positive to confused/negative to insightful.

Graph anomaly detection (GAD) has attracted increasing interest due to its critical role in diverse real-world applications. Graph neural networks (GNNs) offer a promising avenue for GAD, leveraging their exceptional capacity to model complex graph structures and relationships. However, existing GNN-based models encounter challenges in addressing the GAD’s fundamental issue—anomaly camouflage, where anomalies mimic normal instances, leading to indistinguishable features. In this article, we propose a novel approach, termed correlation information enhanced GAD (CIE-GAD). Specifically, drawing on the observation that the distribution of homophilic and heterophilic edges differs between abnormal and normal samples, we construct a hypergraph to learn the co-occurrence relationships among adjacent edges. By enhancing the extraction of sample correlation information, we effectively tackle feature similarity caused by anomaly camouflage, thereby enhancing the performance of GAD. Furthermore, we develop a spectral convolution mechanism based on node-level attention fusion, enabling the capture of multifrequency signals. This module performs adaptive fusion tailored to the unique frequency information requirements of each node, mitigating the local heterophily problem. Extensive experiments on various real-world GAD datasets demonstrate that the proposed CIE-GAD outperforms state-of-the-art methods. Notably, our approach achieves AUC-PR improvements of up to 3.47%, with an average gain of 1.5%, demonstrating its effectiveness in detecting anomalies in graph data.

Reading comprehension is a widely adopted method for learning English, involving reading articles and answering related questions. However, the reading comprehension training typically focuses on the skill level required for a standardized learning stage, without considering the impact of individual differences in linguistic competence. This paper presents a personalized support system for reading comprehension, named ChatPRCS, based on the Zone of Proximal Development (ZPD) theory. It leverages the advanced capabilities of large language models (LLMs), exemplified by ChatGPT (Chat Generative Pre-trained Transformer). ChatPRCS employs methods including skill prediction, question generation, and automatic evaluation, to enhance reading comprehension instruction. Firstly, a ZPD-based algorithm is developed to predict students' reading comprehension skills. This algorithm analyzes historical data to generate questions with appropriate difficulty. Second, a series of ChatGPT prompt patterns is proposed to address two key aspects of reading comprehension objectives: question generation, and automated evaluation. These patterns further improve the quality of generated questions. Finally, by integrating personalized skill prediction and reading comprehension prompt patterns, ChatPRCS is validated through a series of experiments. Empirical results demonstrate that it provides learners with high-quality reading comprehension questions that are broadly aligned with expert-crafted questions at a statistical level. Furthermore, this study investigates the effect of the system on learning achievement, learning motivation and cognitive load, providing further evidence of its effectiveness in instructing English reading comprehension.