This study introduces a benchmark for evaluating information retrieval in the Slovak language, leveraging a question-answering dataset for fine-tuning and assessment of sentence transformers. The dataset, named Retrieval SkQuAD, is integrated into two widely recognized evaluation frameworks: BEIR (Benchmarking Information Retrieval) and MTEB (Massive Text Embedding Benchmark). Retrieval SkQuAD (Slovak Question Answering Dataset) comprises 19,000 manually annotated answers to 1,134 questions, with each answer assigned a relevance score, and includes information on the usefulness of documents in generating responses. Unlike question answering datasets, this resource provides a nuanced assessment of partial relevance across multiple documents. We fine-tuned several sentence transformers and BERT-based models specifically for retrieving documents containing correct answers within the Slovak Wikipedia. Our fine-tuning process incorporates adversarial questions as hard negatives, leading to significant improvements in retrieval accuracy. Experimental results demonstrate that our approach advances state-of-the-art performance for Slovak information retrieval.

A deep learning model with an imbalanced mammography dataset can bias models toward common benign BI-RADS categories and reduce recognition of less frequent malignant or high-risk categories. To address this issue, we propose B2M (Benign Share Benefit to Malignant), a model-agnostic framework for imbalance-aware multi-class BI-RADS classification in C-View mammography. B2M uses a two-phase training strategy that combines dual sampling with feature-space mixing. In Phase I, the model is trained with dual sampling, integrating instance-based and class-balanced sampling to increase minority-class representation while preserving majority-class diversity. In Phase II, the model is fine-tuned with feature-space mixing using samples from the two sampling streams. A soft-target regularization objective supervises the mixed features using labels from both streams, encouraging smoother decision boundaries across BI-RADS categories. We evaluated B2M on an imbalanced mammography cohort from the C-View EMBED dataset using stratified 5-fold cross-validation across multiple CNN backbones. C-View is a synthesized 2D mammographic image generated from 3D digital breast tomosynthesis data, capturing DBT-derived structural information while requiring less memory and computation than processing the full 3D image volume. Among these experiments, ResNeXt-50 with B2M achieved the highest balanced accuracy and Macro-F1 scores compared with the evaluated oversampling and mixing-based methods. This improvement requires an offline training-time overhead of approximately 2.81×, but it does not increase inference cost. Overall, the results suggest that B2M may be useful for imbalanced multi-class BI-RADS classification in C-View mammography. However, the findings are based on the EMBED cohort, and further validation, including external and prospective evaluation, is needed before clinical use.

Wind noise is a pervasive and non-stationary form of interference in outdoor audio recordings, posing a significant challenge for speech enhancement systems. To address this problem, this paper proposes a speech separation-based wind noise reduction framework termed Dual-mask permutation-invariant training (DMPIT). Building upon the dual-masking concept, the key contribution of DMPIT lies in embedding the dual-mask structure within a permutation-invariant training (PIT) framework and reformulating the loss function to better align with speech-oriented noise reduction objectives. Specifically, two supervised masks are jointly optimized: a speech mask that directly estimates the target speech from the mixture and a noise mask that isolates the wind noise component. Assuming that the mixture consists solely of speech and wind noise, the training process computes the loss using both estimated components and the corresponding clean speech. Since wind noise is not a signal of interest, the estimated noise is subtracted from the mixture to obtain a residual speech signal, which is then used to refine the direct speech estimate. The final enhanced speech output is produced by fusing the direct and residual speech estimates through a weighted combination. The experimental results demonstrate that DMPIT consistently outperforms conventional single-mask and single-channel wind noise reduction methods in terms of speech quality and noise suppression.

Chatbots are based on predefined answers and rules, making it difficult to truly understand the user's emotions and achieve the same contextual capabilities as humans. Therefore, it has become a big challenge for chatbots to recognize the narrator's emotions and respond with empathetic sentences matching the topic. This research proposes an emotion classifier model to make the generated response sentences more empathetic and increase the accuracy of emotional prediction. To solve this problem, we propose a pre-trained model based on word vectors to re-train historical situation sentences, thereby improving the fluency of response sentences. The research results show that the CER model proposed in this study is better than other baseline models, reaching 41.31 % in the prediction of emotion classification. In addition to generating more empathetic response sentences, it also makes the generated response sentences more fluent. The CER model's response generation performance is closer to human response behavior, which relies on COMET to extract common-sense knowledge in each situation to help generate empathetic responses.

During semiconductor manufacturing, wafer defect patterns emerge in an uncontrolled environment, making immediate recognition challenging. To enhance the classification accuracy in pattern recognition, deep learning (DL) techniques are employed to address the issue of defective pattern identification. The inherent differences among various wafer defect types lead to imbalanced testing datasets, thereby reducing the efficiency of classification models in delivering accurate predictions. This study proposes a convolutional neural network (CNN) combined with a data transformation method at the feature extraction stage, incorporating Density-based features and Radon-based features to enhance the representation of key patterns. Subsequently, multiple classifiers are integrated using voting, bagging, and AdaBoost strategies to improve classification performance and model generalization. The primary objective is to mitigate data imbalance and selection bias by employing random sampling and assigning distinct weights to individual classifiers. The contributions of this study are threefold. First, we introduce a novel feature extraction strategy that leverages Features extracted by the CNN model, Density-based features and Radon-based features, effectively capturing structural and spatial characteristics in wafer defect patterns. Second, we propose an ensemble learning framework integrating multiple classifiers with optimized weighting mechanisms to enhance classification robustness. Third, we provide empirical evidence demonstrating that our weighted soft voting approach achieves superior performance, attaining a classification accuracy of 95.09% and an F1 score of 0.95. These results validate the effectiveness of our method in improving the reliability of wafer defect classification, which is crucial for advancing automated defect inspection in semiconductor manufacturing.

This study addresses industrial appearance defect detection challenges by enhancing the YOLOv5 algorithm using the NEU-DET dataset. Leveraging optimized Spatial Pyramid Pooling (SPPCSPS), the model exhibits improved accuracy in recognizing various defect types, mitigating issues of omission and misdetection. Experimental results demonstrate a 1.94% average accuracy improvement over the original YOLOv5. The proposed modifications enhance the algorithm's ability to detect defects, contributing to more accurate and efficient industrial inspection processes.

In recent years, embedded system technologies and products for sensor networks and wearable devices used for monitoring people’s activities and health have become the focus of the global IT industry. In order to enhance the speech recognition capabilities of wearable devices, this article discusses the implementation of audio positioning and enhancement in embedded systems using embedded algorithms for direction detection and mixed source separation. The two algorithms are implemented using different embedded systems: direction detection developed using TI TMS320C6713 DSK and mixed source separation developed using Raspberry Pi 2. For mixed source separation, in the first experiment, the average signal-to-interference ratio (SIR) at 1 m and 2 m distances was 16.72 and 15.76, respectively. In the second experiment, when evaluated using speech recognition, the algorithm improved speech recognition accuracy to 95%.

This study addresses the challenges and limitations of surface defect detection on industrial parts. Traditional manual inspection methods are inefficient, error-prone, and difficult to meet the requirements of mass production. Therefore, machine vision technology becomes an alternative. However, existing methods still have problems with leakage and false detection, especially in the NEU-DET data set, which is prone to leakage and false detection when identifying tiny defects like the background. Inspired by this, this research aims to introduce the Coordinate Attention (CA) mechanism and add it to the bottleneck layer, replace all C3 modules in the backbone network with CA _ C3 modules, and optimize the original Spatial Pyramid Pooling (SPP) method, improved and named SPPCSPCGroup, and optimizes the original loss function (QFocalLoss) by adding a penalty factor to specific categories that are prone to misdiagnosis to improve recognition accuracy. Experiments show that the average accuracy (mAP) of the original YOLOv5 is 75.7% on the NEU-DET data set, and the average accuracy (mAP) we obtained is 79.1 %, which is increased by 3.4% through the improved defect detection method proposed in this study. The innovation of this study is to improve the efficiency and accuracy of surface defect detection in industrial applications by improving existing methods and providing reliable technical support for quality control in the industrial production process.

With the rapid advancement of modern hardware technology, breakthroughs have been made in many areas of artificial intelligence research, leading to the direction of machine replacement or assistance in various fields. However, most artificial intelligence or deep learning techniques require large amounts of training data and are typically applicable to a single task objective. Acquiring such large training datasets can be particularly challenging, especially in domains like medical imaging. In the field of image processing, few-shot image segmentation is an area of active research. Recent studies have employed deep learning and meta-learning approaches to enable models to segment objects in images with only a small amount of training data, allowing them to quickly adapt to new task objectives. This paper proposes a network architecture for meta-learning few-shot image segmentation, utilizing a meta-learning classification weight transfer network to generate masks for few-shot image segmentation. The architecture leverages pre-trained classification weight transfers to generate informative prior masks and employs pre-trained feature extraction architecture for feature extraction of query and support images. Furthermore, it utilizes a Feature Enrichment Module to adaptively propagate information from finer features to coarser features in a top-down manner for query image feature extraction. Finally, a classification module is employed for query image segmentation prediction. Experimental results demonstrate that compared to the baseline using the mean Intersection over Union (mIOU) as the evaluation metric, the accuracy increases by 1.7% in the one-shot experiment and by 2.6% in the five-shot experiment. Thus, compared to the baseline, the proposed architecture with meta-learning classification weight transfer network for mask generation exhibits superior performance in few-shot image segmentation.

This study proposes an automated checkout system utilizing improved YOLOv5 technology to classify product images effectively. Traditional self-checkout systems often rely on barcode scanning, leading to time-consuming processes and potential errors. Our system eliminates manual barcode scanning by analyzing product images for quick identification and classification, achieving an impressive classification accuracy of 99.1 % in experimental testing. This research contributes to significant improvements in retail automation, enhancing checkout speed, accuracy, and convenience for both customers and retailers.

With the continuous development of technology, automated checkout systems are becoming increasingly popular in the retail industry. This research aims to use the deep learning technology YOLOv5 model to classify product images efficiently and then implement an automated checkout system without scanning barcodes. Our research method achieved terrific results in testing, with a classification accuracy of 98.7%.

In the research into semantic sentiment analysis, researchers commonly use some factor rules, such as the utilization of emotional keywords and the manual definition of emotional rules, to increase accuracy. However, this approach often requires extensive data and time-consuming training, and there is a need to make the system simpler and more efficient. Recognizing these challenges, our paper introduces a new semantic sentiment analysis system designed to be both higher in quality and more efficient. The structure of our proposed system is organized into several key phases. Initially, we focus on data training, which involves studying emotions and emotional psychology. Utilizing linguistic resources such as HowNet and the Chinese Knowledge and Information Processing (CKIP) techniques, we develop emotional rules that facilitate the generation of sparse representation characteristics. This process also includes constructing a sparse representation dictionary. We can map these back to the original vector space by resolving the sparse coefficients, representing two distinct categories. The system then calculates the error compared to the original vector, and the category with the minimum error is determined. The second phase involves inputting topics and collecting relevant comments from internet forums to gather public opinion on trending topics. The final phase is data classification, where we assess the accuracy of classified issues based on our data training results. Additionally, our experimental results will demonstrate the system’s ability to identify hot topics, thus validating our semantic classification models. This comprehensive approach ensures a more streamlined and effective system for semantic sentiment analysis.

With the breakthroughs in deep learning technology, human pose estimation has emerged as a critical research area in computer vision. Its wide-ranging applications include posture assessment, posture correction, sports coaching, and injury prevention. This study explores the application of human pose estimation techniques in sports science, utilizing OpenPose and YOLOv5 models to capture basketball players' limbs and the shooting postures of basketball players when shooting. Based on these measured data, the accuracy of their shooting techniques will be analyzed based on standards established based on academic research and the experience of coaching professionals. Finally, the machine learning classification model was used to distinguish the quality of the basketball shooting process, achieving an accuracy rate of 87.69%.

With the breakthroughs in deep learning technology, human pose estimation has emerged as a critical research area in computer vision. This study explores the application of human pose estimation techniques in sports science, utilizing OpenPose and YOLOv5 models to capture basketball players’ limbs and shooting postures during shots to analyze the accuracy of their shooting techniques. Ultimately, this research aims to provide in-depth insights into basketball shooting techniques and promote the advancement of physical education.

Deep learning (DL) methods are widely employed in the semiconductor manufacturing process to enhance pattern recognition and classification accuracy, specifically for addressing defect patterns. However, the classification performance of the current models is hindered by the imbalanced distribution of defect data within the test dataset. To tackle this issue, this study presents a feature extraction approach utilizing data transformation, and ensemble learning techniques aiming to enhance the model's classification performance. The primary objective of this study is to mitigate selection and imbalance problems in the dataset through random sampling and assigning distinct weights to individual classifiers. The results demonstrate that the proposed method achieves an impressive accuracy rate of 95.09%, thus substantiating its efficacy in improving the robustness of both the classification model and wafer classification.

Automatic generation of questions about the given context is useful for the adaptation of question-answering systems or to support education. We trained and evaluated a model that generates a question in the Slovak language. We have designed an automatic metric where an additional question-answering model is used to evaluate the generated questions. We calculated how many questions have confidence greater than the given threshold. For generating questions, we used contexts from the Slo-vak question-answering dataset. The fine-tuned Slovak T5 model did generate 38% of the questions that the evaluation model could answer with confidence greater than 50%. We coop-erated with partners from Taiwan during these experiments in the frame of a bilateral project and we plan to transfer the knowledge to the Chinese language later.

By providing enhanced knowledge retrieval capabilities, real-time decision support, and efficient information exchange, question answering (QA) systems play a crucial role in driving productivity, efficiency, and innovation in Industry 4.0. Today's most reliable knowledge-based QA systems require a large knowledge base, which tends to consume more reasoning time. In order to improve the inference speed and response accuracy of the system, this paper adds a Reranker between the Retriever and Reader of the traditional two-stage mechanism. This study uses pretrained Roberta to perform system retrieval and improve data processing and training methods. Experiments on Chinese Wikipedia show that the proposed system significantly reduces the system response time and improves the accuracy and scope of the response.

Named Entity Recognition is a fundamental task in information extraction, which locates and classifies defined named entities in unstructured text. Chinese NER is more difficult than English NER. Since there are no separators between Chinese characters, incorrectly segmented entity boundaries will cause error propagation in NER. In this study, named entity recognition is constructed and applied in the Chinese medical domain, where Chinese medical datasets are labeled in BIO format. The Chinese HealthNER Corpus contains 33,89 sentences, of which 2531 sentences are divided into the validation set and 3204 sentences are divided into the test set. This study uses PyTorch Embedding + BiLSTM + CRF, RoBERTa + BiLSTM + CRF, BERT Classifier, and BERT + BiLSTM + CRF for training and compares their model performance. Finally, the BERT + BiLSTM + CRF achieves the best prediction performance with a precision of 91.30%, recall of 89.46%, and F1 score of 90.53%