Purpose High-quality book metadata improves digital libraries by supporting cataloging, searchability and automated classification. However, metadata generation from scanned books remains challenging due to limited annotated data sets and optical character recognition (OCR) limitations. This study aims to propose a deep learning-based framework for automatic metadata extraction from digitized books by integrating computer vision and natural language processing (NLP). Design/methodology/approach The framework uses MobileNetV2 to classify title pages, table of contents (ToC) pages and content pages. EfficientDet detects metadata-related regions, such as titles and author information. OCR extracts text from these regions, followed by named entity recognition (NER) and regular expressions to refine the extracted metadata. The framework was evaluated using a custom data set of 188 books published between 1800 and 2021, comprising 857 annotated pages. Findings The page classification model achieved 97.16% accuracy. For object detection, the model obtained average precision (AP) and average recall (AR) scores of 71.7% and 42.0% for title pages, 55.1% and 33.4% for ToC pages and 87.9% and 54.4% for content pages, respectively. Originality/value This study contributes a scalable framework that combines page classification, object detection, OCR, NER and rule-based refinement to reduce manual metadata extraction in digital library management. Future research can expand multilingual data sets, improve robustness to OCR noise and explore end-to-end learning for richer bibliographic metadata extraction.
Purpose The increasingly competitive job market has intensified psychological pressure on college graduates, posing new challenges for intelligent analysis and intervention within cyber-physical-social systems (CPSS). To address the limitations of manual assessment and support collaborative intelligence in employment counseling scenarios, this paper aims to propose an employment sentiment analysis method based on global and local key feature fusion. Design/methodology/approach By incorporating rough set theory, a local key information module is constructed to identify sentiment-critical words from interview data sets, enabling interpretable feature selection suitable for edge-level processing. Furthermore, convolutional neural networks are employed to extract sentiment representations at both global and local scales, while a channel attention mechanism is introduced to enhance feature discrimination without increasing model complexity, making the approach suitable for edge-artificial intelligence deployment. An overlapping pooling strategy combined with multilevel pyramids is adopted to improve robustness under resource-constrained environments. Finally, global and local sentiment information is collaboratively fused to support multidimensional sentiment analysis within CPSS. Findings Experimental results demonstrate that the proposed method effectively captures employment-related emotional semantics of college graduates and provides accurate early warning of negative sentiment, offering practical value for enabling collaborative intelligence-driven interventions in smart employment and social support systems. Originality/value This study introduces a novel employment sentiment analysis method that fuses global and local key features to improve the accuracy of detecting emotional states in college graduates seeking jobs. By integrating rough set-based key word selection, CNN-based multiscale sentiment mining, channel attention and multilevel pyramid pooling, the approach enhances feature extraction without increasing model complexity. The fusion of global text context and local word-level cues enables a more nuanced and reliable assessment, providing timely warning of negative emotions and supporting targeted manual intervention in graduate employment counseling.
PurposeSecure transmission in wireless communication systems has garnered significant attention recently. A key challenge is thwarting eavesdroppers while optimizing the performance of legitimate users and security. This paper aims to propose a novel scheme using index modulation (IM) for a space-time block code (STBC)-aided simultaneously transmitting and reflecting reconfigurable intelligent surface (STAR-RIS) system to ensure physical layer security (PLS).Design/methodology/approachSTAR-RIS is designed as an access point (AP) to adjust array element phases for signal generation. By grouping the STAR-RIS and using IM, different antenna combinations are activated for transmission. The zero-forcing (ZF) precoding method generates artificial noise (AN) based on the channel state information (CSI) of the legitimate user, which is then added to the transmitted signals. Finally, Alamouti coding is integrated into the STAR-RIS for signal transmission.FindingsTheoretical and simulation analyses of bit error rate (BER) and secrecy rate show the scheme prevents eavesdropping and enhances the secrecy rate. At a 30dB signal-to-noise ratio (SNR), the secrecy rate increases by approximately 8 bits/s/Hz compared to the no-AN scheme. Furthermore, compared to traditional STAR-RIS, at BER = 10-4, the SNR gain is approximately 2dB.Research limitations/implicationsThe good results are based on CSI in the communication system.Practical implicationsIt could be used for V2V communication networking and secured communications.Social implicationsIt gives more ideas and research on the coding in the RIS.Originality/valueDesigning STAR-RIS as an AP simplifies channel links and, combined with Alamouti code, greatly boosts diversity gain via spatial-temporal coding. Moreover, this paper uses the STAR-RIS to implement IM, enabling the additional transmission of index bits and improving spectral efficiency. Meanwhile, adding AN to signals, a key design aspect, prevents eavesdropping and enhances security, showing a holistic approach to optimizing wireless security.
PurposeEvent recognition (ER) in drone flight logs is a task in forensic investigations to extract key events from unstructured log messages. However, drone flight logs often exhibit inconsistencies in punctuation and capitalization, which make traditional ER approaches unreliable. This paper aims to propose a novel noise-resistant method that can automate the process of finding key events and extract a brief insight about the investigated artifacts quickly.Design/methodology/approachThe problem is modeled as named entity recognition (NER) to tackle both sentence boundary detection (SBD) and ER, simultaneously. A new drone flight log event detection data set is constructed and used to evaluate multiple pre-trained language models, including BERT, XLNET and RoBERTa. In addition, the authors introduce a noise-introducing oversampling technique to improve the model's robustness against log structure inconsistencies.FindingsThe results demonstrate that the proposed noise-introducing oversampling technique successfully reduces the models' sensitivity to the punctuation presence. Moreover, lowercasing the input logs naturally eliminates the challenge raised by the inconsistencies in capital letter usage. The proposed evaluation metric, Fused-F1 (F-F1), better captures the overall model's performance by accounting for the hierarchical dependency between SBD and ER tasks, compared to the standard metric. To promote reproducibility, the authors release the data set, code and evaluation scripts in a public repository.Practical implicationsAn open-source tool is released along with this paper as the implementation of the proposed framework. The tool is expected to be an intelligent assistive technology for drone forensic investigators in analyzing forensic timelines more efficiently.Originality/valueTo the best of the authors' knowledge, this paper is the first attempt to apply NER-based modeling to perform ER in the drone forensics domain.
PurposeThis study aims to address the limitations of standard cryptographic methods in securing Internet of Things (IoT) devices by introducing a compact, hybrid lightweight encryption scheme. Focusing on symmetric-key algorithms, it combines a bit permutation approach with the PRESENT cipher, which uses a substitution-permutation network (SPN). To overcome the lack of confusion in bit permutation methods, a novel instruction was added to enhance security. The scheme was tested on ARM, MSP and AVR microcontrollers, showing improved compactness and an enhanced avalanche effect. Overall, it offers a more efficient and secure encryption solution for resource-constrained IoT environments.Design/methodology/approachThe proposed hybrid lightweight encryption design combines the PRESENT cipher's S-box with new bit permutation instructions that include the confusion property, missing in previous designs. These instructions replace the original P-layer, enabling faster and more compact 64-bit permutations using only 4 lines of code instead of 23, resulting in a more efficient and memory-saving encryption scheme suitable for IoT devices.FindingsThe proposed E-PRESENT cipher outperformed the original PRESENT and PRESENT-GRP across all evaluated IoT platforms. In scenario 0, it showed superior performance by balancing code size, RAM usage and execution time equally. Scenario 1 highlighted platform-specific differences, where E-PRESENT maintained the fastest execution time and most efficient memory use, even on constrained 8-bit devices, underlining the importance of RAM in performance. Scenario 2 confirmed E-PRESENT's top ranking across all platforms. Overall, the results demonstrate that E-PRESENT is the most efficient and compact among the compared designs, making it a strong candidate for use in resource-limited IoT environments.Originality/valueThe originality of this work lies in the integration of confusion-enhanced bit permutation instructions into a lightweight symmetric encryption scheme. Unlike existing designs, the proposed E-PRESENT cipher replaces the standard P-layer of the PRESENT cipher with novel bit-level operations that are both compact and efficient. This approach not only reduces memory and computational requirements but also strengthens security through improved confusion properties. The design offers a unique balance between performance and cryptographic strength, making it highly suitable for constrained IoT environments. Its innovative structure demonstrates clear value for future lightweight cryptographic implementations.
