{"title":"Expert system for extracting keywords in educational texts and textbooks based on transformers models","authors":"Irene Cid Rico, Jordán Pascual Espada","doi":"10.1016/j.eswa.2025.127735","DOIUrl":"10.1016/j.eswa.2025.127735","url":null,"abstract":"<div><div>Automated keyword extraction is widely used for tasks like classification and summarization, but generic methods often fail to address domain-specific requirements. In education, texts are designed to help students grasp and retain key concepts needed for exercises and resolve questions. Despite the variety of existing keyword extraction algorithms, none are specifically adapted to the unique structure and purpose of educational materials like textbooks or lecture notes.Supervised methods have demonstrated their effectiveness in various domains through advanced techniques like contextual embeddings and domain-specific fine-tuning, Our study proposes a novel solution leveraging pretrained transformer models, specifically BERT, to adapt to the structure of educational materials for effective keyword extraction. Our research demonstrates that by fine-tuning BERT models to the specific characteristics of educational texts, we can achieve more accurate and relevant keyword extraction. YodkW, our adapted model, outperforms traditional algorithms in identifying the key concepts that are essential for educational purposes. Performance is quantified using the F1 score relative to text books key terms list, Preliminary results demonstrate that our approach can improve the identification of key concepts pertinent to student understanding and facilitate the automatic generation of test questions.</div></div>","PeriodicalId":50461,"journal":{"name":"Expert Systems with Applications","volume":"282 ","pages":"Article 127735"},"PeriodicalIF":7.5,"publicationDate":"2025-04-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143874004","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Zhiyong Peng , Changlin Han , Yadong Liu , Jingsheng Tang , Zongtan Zhou
{"title":"Pessimistic policy iteration with bounded uncertainty","authors":"Zhiyong Peng , Changlin Han , Yadong Liu , Jingsheng Tang , Zongtan Zhou","doi":"10.1016/j.eswa.2025.127651","DOIUrl":"10.1016/j.eswa.2025.127651","url":null,"abstract":"<div><div>Offline Reinforcement Learning (RL) aims to learn policies by using static datasets. The extrapolation error in out-of-distribution (OOD) samples can cause off-policy RL algorithms to perform poorly on offline datasets. Hence, it is critical to avoid visiting OOD states and taking OOD actions in offline RL. Several recent methods have used uncertainty estimation to distinguish OOD samples. However, errors in the uncertainty estimation make the purely uncertainty-based method unstable and require additional components to ensure sufficient pessimism. In this study, we propose a Bounded Uncertainty based Pessimistic policy iteration algorithm (BUP). The BUP pessimistically estimates the value function via bounded uncertainty, and the uncertainty bound is achieved by constraining the actor from taking highly uncertain actions. The suboptimality bound of BUP is theoretically guaranteed in linear Markov Decision Processes (MDPs), and experiments on D4RL datasets show that BUP matches the state-of-the-art performance. Moreover, BUP is simple to implement with low computational cost and does not require any additional components.</div></div>","PeriodicalId":50461,"journal":{"name":"Expert Systems with Applications","volume":"282 ","pages":"Article 127651"},"PeriodicalIF":7.5,"publicationDate":"2025-04-25","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143876887","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Aihui Ye , Runtong Zhang , Wei Cai , Yang Liu , Cui Shang , Xiaomin Zhu
{"title":"A consensus-based group decision-making method for multidisciplinary team meeting under q-rung orthopair fuzzy environment","authors":"Aihui Ye , Runtong Zhang , Wei Cai , Yang Liu , Cui Shang , Xiaomin Zhu","doi":"10.1016/j.eswa.2025.127761","DOIUrl":"10.1016/j.eswa.2025.127761","url":null,"abstract":"<div><div>In response to the complex treatment process and evolving medical needs of multimorbidity, multidisciplinary team (MDT) is dedicated to integrating the diagnosis opinions of experts and providing optimal treatment plans. Reaching consensus on disease treatment plans involves a dynamic and iterative group decision-making process, in which traditional methods for MDT meetings fail to address the standardized decision-making procedure, interactive trust relationships, and fuzzy information integration. Given the challenges, this study proposes a dynamic consensus framework based on dual-path feedback mechanism with <em>q</em>-rung orthopair fuzzy set (<em>q</em>-ROFS). A hybrid trust evolution model is first established within MDT, in which the trust degree is composed of inherent trust and preference similarity in each round. Then the opinion dynamics model is also introduced to the fuzzy environment. Based on trust evolution and opinion dynamics, the dual-path feedback mechanism is employed to provide references for preference adjustment and weight adjustment. Correspondingly, the calculation methods for consensus measure, preference similarity and alternative selection with <em>q</em>-ROFS are proposed. Additionally, a case study about vascular MDT meeting is used to illustrate the effectiveness of the proposed method. The simulation experiments are performed to verify the impact of consensus threshold, group size, individual self-confidence, and trust evolution on the proposed method. The results of the comparative analysis show that increasing the <em>q</em> value can expand the fuzzy information expression space while ensuring the consensus level, and the proposed method is superior to other methods in terms of more efficient and high-quality consensus results.</div></div>","PeriodicalId":50461,"journal":{"name":"Expert Systems with Applications","volume":"282 ","pages":"Article 127761"},"PeriodicalIF":7.5,"publicationDate":"2025-04-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143864800","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
{"title":"Generalizing fuzzy k-nearest neighbor classifier using an OWA operator with a RIM quantifier","authors":"Mahinda Mailagaha Kumbure, Pasi Luukka","doi":"10.1016/j.eswa.2025.127795","DOIUrl":"10.1016/j.eswa.2025.127795","url":null,"abstract":"<div><div>This paper proposes a new fuzzy k-nearest neighbor (FKNN) method, called the ordered weighted averaging (OWA) with regular increasing monotone quantifier-based fuzzy k-nearest neighbor (OWARIM-FKNN) classifier. The proposed method aims at enhancing the classification performance of the KNN rule-base variants, especially the local mean-based approaches, while dealing with outlier and data uncertainty issues. In the proposed method, the OWA operator is used to generalize the multi-local mean vectors from each class. The resulting <span><math><mi>k</mi></math></span> multi-local OWA vectors are then used to create the class representative pseudo nearest neighbors. Lastly, the new sample is classified into the class with the highest membership degree measured using the weighted distance between the new sample and the pseudo nearest neighbor. The classification performance of the proposed method was examined using one artificial and twenty-seven real-world data sets compared with the results obtained from eight related KNN variants. Experimental results showed that the proposed OWARIM-FKNN classifier achieves the highest average accuracy of 87.59% with an average confidence interval of <span><math><mo>±</mo></math></span>0.64, outperforming all baseline methods. Using the Friedman and Nemenyi tests, the analysis further confirms that the proposed method shows statistically significant performance improvements.</div></div>","PeriodicalId":50461,"journal":{"name":"Expert Systems with Applications","volume":"282 ","pages":"Article 127795"},"PeriodicalIF":7.5,"publicationDate":"2025-04-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143869519","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Miao Zhang , Zhenlong Fang , Tianyi Wang , Shuai Lu , Xueqian Wang , Tianyu Shi
{"title":"CCMA: A framework for cascading cooperative multi-agent in autonomous driving merging using Large Language Models","authors":"Miao Zhang , Zhenlong Fang , Tianyi Wang , Shuai Lu , Xueqian Wang , Tianyu Shi","doi":"10.1016/j.eswa.2025.127717","DOIUrl":"10.1016/j.eswa.2025.127717","url":null,"abstract":"<div><div>Traditional Reinforcement Learning (RL) suffers from challenges in replicating human-like behaviors, generalizing effectively in multi-agent scenarios, and overcoming inherent interpretability issues. These tasks become even more difficult when they require a deep understanding of the environment, coordination of agents’ intentions and driving styles across various scenarios, and the overall optimization of safety, efficiency, and comfort in dynamic environments. Recently, Large Language Model (LLM) enhanced methods have shown promise in improving generalization and interoperability. However, these approaches primarily focus on single-agent scenarios and often neglect the necessary coordination among multiple road users. Therefore, in this paper, we introduce the Cascading Cooperative Multi-agent (CCMA) framework, designed to address these challenges by enhancing human-like behaviors and fostering multi-level cooperation across diverse multi-agent driving tasks, ultimately improving both micro and macro-level performance in complex driving environments. Specifically, the CCMA framework integrates RL for individual interactions, a fine-tuned LLM for regional cooperation, a reward function for global optimization, and the Retrieval-augmented Generation mechanism to dynamically optimize decision-making across complex driving scenarios. Our experiments demonstrate that our CCMA method not only enhances human-like behaviors and interpretability, but also outperforms other state-of-the-art RL methods in multi-agent environments. These findings highlight the significant impact of cascading coordinated communication and dynamic functional alignment in advanced, human-like multi-agent autonomous driving environments. Our project page is <span><span>https://miaorain.github.io/rainrun.github.io/</span><svg><path></path></svg></span>.</div></div>","PeriodicalId":50461,"journal":{"name":"Expert Systems with Applications","volume":"282 ","pages":"Article 127717"},"PeriodicalIF":7.5,"publicationDate":"2025-04-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143876883","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Ho Anh Thu Nguyen , Duy Hoang Pham , Byeol Kim , Yonghan Ahn , Nahyun Kwon
{"title":"Developing an automated framework for eco-label information categorization using web crawling and Natural Language Processing techniques","authors":"Ho Anh Thu Nguyen , Duy Hoang Pham , Byeol Kim , Yonghan Ahn , Nahyun Kwon","doi":"10.1016/j.eswa.2025.127688","DOIUrl":"10.1016/j.eswa.2025.127688","url":null,"abstract":"<div><div>Eco-labels are extensively employed to assess the environmental performance of building materials. However, their management is often fragmented across disparate online databases with inconsistent data structures, presenting significant challenges for efficient information acquisition and management. This study explores the application of web crawling techniques, Natural Language Processing (NLP), and machine learning (ML) models to collect and categorize eco-label information, with the objective of advancing the automation of information management processes. The results demonstrate that the categorization models exhibit high performance, achieving F1-scores exceeding 0.95 on the test set and at least 0.76 when validating datasets incorporating temporally updated information. However, the limited availability of data for certain eco-labels, such as Forest Stewardship Council certification and Green Screen, substantially degrades model performance with updated data. Notably, traditional ML models leveraging manual feature engineering outperform deep learning models with automatic feature extraction when applied to web-crawled data. Furthermore, the TF-IDF feature extraction technique surpasses other n-gram-based approaches, with model performance declining as n-gram length increases. This study establishes a systematic framework that informs the selection of reliable data sources, feature engineering strategies, and ML algorithms for integrating web crawling, thereby enhancing the automation of eco-label information management.</div></div>","PeriodicalId":50461,"journal":{"name":"Expert Systems with Applications","volume":"282 ","pages":"Article 127688"},"PeriodicalIF":7.5,"publicationDate":"2025-04-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143869521","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Qiaohong Chen, Zhenyang Xu, Xian Fang, Qi Sun, Xin Wang
{"title":"Selective Guidance Network with edge and texture awareness for polyp segmentation","authors":"Qiaohong Chen, Zhenyang Xu, Xian Fang, Qi Sun, Xin Wang","doi":"10.1016/j.eswa.2025.127772","DOIUrl":"10.1016/j.eswa.2025.127772","url":null,"abstract":"<div><div>Colorectal polyp segmentation plays a crucial role in preventing colorectal cancer through colonoscopic image screening. However, most existing methods overlook uncertain regions in colonoscopic images, particularly the blurred boundary areas where polyps closely resemble colon fold structures. To address this challenge, we propose the Selective Guidance Network with edge and texture awareness for polyp segmentation (SGNet). SGNet consists of three essential modules, namely the Edge and Texture Awareness Module (ETAM), the Prior Enhancement Module (PEM), and the Hierarchical Feature Fusion Module (HFFM). ETAM integrates Laplacian operators with spatial attention mechanisms to enhance feature perception, allowing for precise extraction of polyp boundaries and adaptive amplification of texture patterns. PEM strengthens multi-scale contextual perception through dilated convolutions while refining backbone features through dual prior-driven feature rectification. HFFM employs multi-level attention gating to achieve cross-scale feature integration while effectively combining low-level edge cues with high-level semantic representations. Experimental results on five public datasets demonstrate that SGNet outperforms 16 state-of-the-art methods across six evaluation metrics, highlighting its superior segmentation performance and robustness.</div></div>","PeriodicalId":50461,"journal":{"name":"Expert Systems with Applications","volume":"282 ","pages":"Article 127772"},"PeriodicalIF":7.5,"publicationDate":"2025-04-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143874006","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Laura de Diego-Otón , Álvaro Hernández , David Fuentes , Rubén Nieto , Víctor M. Navarro
{"title":"Architectural strategies for enhanced NILM classification and anomaly detection: Addressing limited data scenarios","authors":"Laura de Diego-Otón , Álvaro Hernández , David Fuentes , Rubén Nieto , Víctor M. Navarro","doi":"10.1016/j.eswa.2025.127756","DOIUrl":"10.1016/j.eswa.2025.127756","url":null,"abstract":"<div><div>Non-intrusive load monitoring (NILM) enables appliance-level behaviour analysis by examining the aggregated electrical consumption signals. These techniques hold significant potential for applications ranging from electrical load management to remote human health monitoring. Despite its potential, NILM faces challenges in adapting to evolving appliance baselines, including the integration of new devices or the replacement of existing ones. These challenges aggravate when dealing with a large number of appliances, or even more if there are overlapping energy consumption profiles, thus reducing the effectiveness of load monitoring techniques. In real-world scenarios, the scarcity of labelled data further intensifies these issues, increasing the risk of overfitting. This limits the ability of NILM models to generalise and perform effectively on unseen data. To address these limitations, this work presents some methods for accurately classifying known appliances while identifying unknown ones by using features derived from electrical current signals. The framework includes a feature extraction stage that explores neural networks with both supervised and unsupervised learning techniques to derive latent representations. Additionally, the appliance distinction stage optimises data distribution for recognised known appliances and evaluates two distinct approaches (a supervised method and a semi-supervised one) for detecting unseen appliances. Experimental evaluations demonstrate promising results, achieving over 95% accuracy for the supervised feature extraction method and 83% for the unsupervised one in classifying known appliances, even under limited data conditions. Furthermore, both approaches performed well in detecting unseen appliances, with detection rates exceeding 90% for the supervised classification method and 70% for the semi-supervised method for certain categories.</div></div>","PeriodicalId":50461,"journal":{"name":"Expert Systems with Applications","volume":"282 ","pages":"Article 127756"},"PeriodicalIF":7.5,"publicationDate":"2025-04-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143876884","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Mingwei Wang , Sheng Li , Chong Cheng , Maolin Chen , Wei Liu , Zhiwei Ye
{"title":"A Tri-evolutionary mechanism with information interaction of differential evolution for hyperspectral band selection","authors":"Mingwei Wang , Sheng Li , Chong Cheng , Maolin Chen , Wei Liu , Zhiwei Ye","doi":"10.1016/j.eswa.2025.127611","DOIUrl":"10.1016/j.eswa.2025.127611","url":null,"abstract":"<div><div>Band selection is one of the most important tasks for hyperspectral imaging (HSI) dataset, and fewer bands are extracted with satisfactory classification accuracy. As a combinatorial optimization problem, evolutionary algorithm (EA) is widely used in the field, each band is represented by a dimension. However, reducing the number of selected bands for a vast number of band combinations presents challenges. In this paper, a Tri-evolutionary mechanism with information interaction of differential evolution (TEDE) is proposed for hyperspectral band selection, all bands are divided into two parts based on their correlation with each band and label, two populations are independently trained on these parts, and the selected bands are combined to reconstruct the band combination. Subsequently, a new population is generated to obtain the optimal band subsets satisfying the criteria of information interaction. Experimental comparisons with EA-based, Co-evolution-based and newly proposed band selection methods are conducted to evaluate the performance of Tri-evolution, which demonstrate that TEDE outperforms other approaches in multiple metrics, notably reducing the number of selected bands while improving classification accuracy.</div></div>","PeriodicalId":50461,"journal":{"name":"Expert Systems with Applications","volume":"282 ","pages":"Article 127611"},"PeriodicalIF":7.5,"publicationDate":"2025-04-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143876886","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
{"title":"TransCORALNet: A two-stream transformer CORAL networks for supply chain credit assessment cold start","authors":"Jie Shi, Arno P.J.M. Siebes, Siamak Mehrkanoon","doi":"10.1016/j.eswa.2025.127581","DOIUrl":"10.1016/j.eswa.2025.127581","url":null,"abstract":"<div><div>Supply chain credit assessment is critical for financial decision-making due to limited historical data for new borrowers and the domain shift between segment industries. Existing models often struggle with challenges such as domain shift, cold start, imbalanced classes, and lack of interpretability. This paper proposes an interpretable two-stream transformer CORAL network (TransCORALNet) for supply chain credit assessment, designed to address these challenges. The two-stream domain adaptation architecture with correlation alignment (CORAL) loss serves as the core model and is equipped with a transformer, which provides insights into the learned features and allows efficient parallelization during training. Thanks to the domain adaptation capability of the proposed model, the domain shift between the source and target domains is minimized. Furthermore, we employ Local Interpretable Model-agnostic Explanations (LIME) to provide additional insights into the model predictions and identify the key features contributing to supply chain credit assessment decisions. Experimental results on a real-world dataset demonstrate the superiority of TransCORALNet over several state-of-the-art baselines in terms of accuracy. The code is available on GitHub.<span><span><sup>1</sup></span></span></div></div>","PeriodicalId":50461,"journal":{"name":"Expert Systems with Applications","volume":"282 ","pages":"Article 127581"},"PeriodicalIF":7.5,"publicationDate":"2025-04-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143869642","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}