Although most of the research on large language models (LLMs) focuses on their development and validation against datasets, significant gaps remain in their application to real-world knowledge-intensive tasks. This research addresses key challenges in using LLMs for extracting and synthesizing knowledge from unstructured sources, focusing on applications where the validity and consistency of the results are critical. We propose Ontology-based Adaptive Knowledge System (OAKS) as a holistic approach to manage the complexities of acquiring unstructured knowledge, varying user expertise, and dynamic query formulation. This research provides practical value for enabling the domain user community to leverage their technical documentation and expertise and accelerate ongoing working projects through improved literature review and cross-disciplinary insight discovery. Validated through empirical studies, our findings offer insight into best practices for the deployment of OAKS, bridging the gaps between AI capabilities and real-world needs in knowledge acquisition and research.

Magnetic resonance imaging (MRI) is a powerful imaging method that provides rich anatomical information in clinical applications, leading to more accurate diagnosis and pathological analysis. However, MRI acquisition is limited by the hardware performance and scan time, making it challenging to obtain complete high-quality images. In recent years, MRI reconstruction algorithms using deep learning (DL) have demonstrated good capabilities in improving image quality and accelerating image acquisition. Thus, a new complex-valued dual-domain dilated convolution neural network (C3DNet) providing fast and accurate MRI reconstruction is proposed in this paper. Unlike existing DL-based methods, the developed C3DNet uses complex-valued convolution to extract complex-valued features from the k-space and image-domain data separately and perform dual-domain feature fusion. Additionally, we use dilated convolution to expand the features’ receptive field and thus capture contextual information. To fully use the prior knowledge, we utilize two data consistency (DC) methods and apply them several times to both the k-space and image-domain feature maps. Experimental results demonstrate that our proposed method outperforms several state-of-the-art algorithms regarding imaging results and computation time.

The recognition of small unmanned aerial vehicle (UAV) has important application value. The very challenging problems in small UAV recognition are recognition accuracy of traditional methods and interpretability of deep learning-based methods. High-resolution radar sensors provide an attractive choice for this issue. In this paper, we propose a multi-feature fusion method for small UAV recognition based on high-resolution radar sensor, which can combine the structure features and micro-Doppler features of small UAV. A dataset contains two different type of small UAV targets, namely four-rotor UAV and two-rotor helicopter, is constructed for training and testing. The recognition accuracy of these two small UAVs based on the measured high-resolution radar data reached 98.5%. The experimental results demonstrate that the proposed multi-feature fusion method is effective for small UAV recognition.

Following the great success of deep learning in signal processing, Many models based on real-valued convolutional neural networks (CNNs) and recurrent neural networks (RNNs) have been proposed for automatic modulation classification (AMC). However, the modulation signal is not only temporally dependent but also complex-valued data. The real-valued deep learning models treat the real and imaginary parts of the complex-valued modulation signal as two independent real-valued inputs, which destroy the structure of the raw signal data and make the model more uninterpretable. Thus, this paper proposes a novel complex-valued parallel convolutional recurrent neural network (CPCRNN) specifically for AMC. CPCRNN combines parallel complex-valued CNN and RNN with redesigned complex-valued activation function and complex-valued max pooling. The model directly feeds the complex-valued raw signal to the complex-valued CNN to obtain the complex-valued feature maps, which are then transformed into amplitude and phase and fed to the RNN. Our model can first extract the complex-valued features of the modulation signal with complex-valued CNNs, and then extract the temporal features of the modulation signal with RNNs. CPCRNN achieved an overall accuracy of 62.29% and 69.93% on the benchmark datasets RadioML2016.10A and RadioML2018.01-simple, respectively, outperforming all the state-of-the-art algorithms.

Transcripts generated by automatic speech recognition (ASR) systems usually have poor readability caused by lacking of punctuation and containing a large portion of dis-fluency. Existing methods for automatic punctuation restoration have obtain great improvements by finetuning large pre-trained language models (LMs). However, large amount of well-formatted written language are used in pre-training and finetuning LMs, leaving a mismatch between the training and application. In this paper, we modify the ELECTRA model with disfluency generator and multi-task discriminator for automatic punctuation restoration. The generator dynamically inject disfluency into the input training data, and the discriminator is trained to distinguish disfluency from generator outputs and predict punctuation marks. Experimental results demonstrate that our proposed method significantly outperforms the baselines based on the English IWSLT dataset and our newly collected Chinese dataset.

With the use of deep learning, object detection has achieved great breakthroughs. However, existing object detection methods still can not cope with challenging environments, such as dense objects, small objects, and object scale variations. To address these issues, this paper proposes a novel keypoint-based detection framework, called CrossNet, which significantly improves detection performance with minimal costs. In our approach, an object is modeled as a cross that consists of a center keypoint and a specific size, which eliminates the need of hand-craft anchor design. The proposed CrossNet outputs three types of maps: the center map, size map, and offset map, where both center map and offset map are to predict the center keypoints of objects and the size map is to estimate the sizes (width and height) of objects. Specifically, we first design a cascaded center prediction method that introduces a coarse-to-fine idea to improve center prediction. Furthermore, since center prediction considered as a classification task is easier than size regression relatively, we design a center-attention size regression module that uses the detection results of centers to assist the size prediction. In addition, a slightly modified hourglass network is designed to enhance the quality of feature maps for center and size prediction. Extensive experiments are conducted to demonstrate the effectiveness of CrossNet on the challenging PASCAL VOC, COCO, KITTI, and WiderFace datasets. Empirical studies show that CrossNet achieves competitive results with top-ranked one-stage and two-stage detectors while being time-efficient.

Under the new normal of China's economic restructuring, corporate bond defaults occur frequently, so it is increasingly important to measure bond credit risk and strengthen risk management. This paper takes Chinese corporate bonds as the research object, and selects 35 default bonds and 861 non-default bonds as the research samples. The study conducted a descriptive statistical analysis of the survival time of the sample, and used the survival analysis method to summarize the types of companies, industries, or businesses that are more likely to have credit risks. This study is a good supplement to the existing research on bond default, and has certain reference significance for investors and risk management institutions.

Fill-in-the-blank question generation has become enormously popular and attracted lots of attention recently. However, most of the existing question generation datasets are developed for machine reading comprehension, which are not specifically designed for examination. To fill in the gap, in this paper, we propose a Chinese sentence fill-in-the-blank question generation dataset for examination (named CSFQGD), which will be released to the public11Resources are available at The dataset is composed of 20.5K questions from many real examinations in Chinese that cover a wide spectrum of learning subjects. Based on the proposed dataset, we test several well-known methods for fill-in-the-blank question generation and compare their performance. Our baseline study on this dataset shows that CSFQGD is a challenging test bed for further research.

Radar-based sensors provide an attractive choice for hand gesture recognition (HGR). The very challenging problems in radar-based HGR are radar echo data preprocessing and recognition accuracy. In this paper, we propose a convolutional neural network (CNN) for dynamic HGR based on a millimeter-wave Frequency Modulated Continuous Wave (FMCW) radar which operates at 77GHz. Six different dynamic hand gestures are designed and the time-frequency analysis of micro-Doppler signatures are adopted as the input to CNN. The measured data of the dynamic hand gestures are collected in different experimental scenarios. The recognition accuracy of the six gestures based on the measured data reached 95.2%. The experimental results demonstrate that the proposed method is effective in the measured data and the micro-Doppler signature is effective for dynamic HGR.

Detecting rotated ships is difficult in optical remote sensing images due to the challenges of complex scenes. Existing advanced rotated ship detectors are typically anchor-based algorithms that require plenty of predefined anchors. However, the use of anchors brings three critical problems: 1) a large number of anchors bring a huge amount of calculation; 2) the attributes (e.g., size and aspect ratios) of anchors are designed viaad hocheuristics; and 3) only a tiny fraction of anchors that overlap with ground-truth bounding boxes of ships tightly can be considered as positive samples, which causes an extreme imbalance between positive and negative samples. As a result, the detection accuracy will be influenced seriously when the design of anchors is not suitable. To address the above problems, this article proposes a novel anchor-free rotated ship detection framework, called SKNet, which detects rotated ships as keypoints in optical remote sensing images. In SKNet, a ship target is modeled as its center keypoint and morphological sizes, including the width, height, and rotation angle. Accordingly, we design two customized modules: orthogonal pooling and soft-rotate-nonmaximum suppression (NMS), where the former is to improve the prediction accuracy of the center keypoint and the morphological size, and the latter is to effectively remove redundant rotated ship detection results. Extensive experiments are conducted to demonstrate the effectiveness of SKNet on three optical remote sensing image data sets: HRSC2016, DOTA-ship, and HPDM-OSOD, which is collected by ourselves and published in this article. Empirical studies show that SKNet achieves state-of-the-art detection performance while being time-efficient. Overall, SKNet achieves the best speed–accuracy tradeoff.

Drug-drug interactions (DDIs) are crucial for drug research and pharmacovigilance. These interactions may cause adverse drug effects that threaten public health and patient safety. Therefore, the DDIs extraction from biomedical literature has been widely studied and emphasized in modern biomedical research. The previous rules-based and machine learning approaches rely on tedious feature engineering, which is labourious, time-consuming and unsatisfactory. With the development of deep learning technologies, this problem is alleviated by learning feature representations automatically. Here, we review the recent deep learning methods that have been applied to the extraction of DDIs from biomedical literature. We describe each method briefly and compare its performance in the DDI corpus systematically. Next, we summarize the advantages and disadvantages of these deep learning models for this task. Furthermore, we discuss some challenges and future perspectives of DDI extraction via deep learning methods. This review aims to serve as a useful guide for interested researchers to further advance bioinformatics algorithms for DDIs extraction from the literature.

This paper proposes a robust obstacle detection and recognition method for driver assistance systems. Unlike existing methods, our method aims to detect and recognize obstacles on the road rather than all the obstacles in the view. The proposed method involves two stages aiming at an increased quality of the results. The first stage is to locate the positions of obstacles on the road. In order to accurately locate the on-road obstacles, we propose an obstacle detection method based on the U-V disparity map generated from a stereo vision system. The proposed U-V disparity algorithm makes use of the V-disparity map that provides a good representation of the geometric content of the road region to extract the road features, and then detects the on-road obstacles using our proposed realistic U-disparity map that eliminates the foreshortening effects caused by the perspective projection of pinhole imaging. The proposed realistic U-disparity map greatly improves the detection accuracy of the distant obstacles compared with the conventional U-disparity map. Second, the detection results of our proposed U-V disparity algorithm are put into a context-aware Faster-RCNN that combines the interior and contextual features to improve the recognition accuracy of small and occluded obstacles. Specifically, we propose a context-aware module and apply it into the architecture of Faster-RCNN. The experimental results on two public datasets show that our proposed method achieves state-of-the-art performance under various driving conditions.

The recognition of cloud maps plays an important role in the field of meteorological prediction. The mainstream recognition of cloud maps mainly focuses on satellite cloud maps, while research on ground-to-air cloud maps is less. In this paper, we propose an analysis framework - YUN - to quickly locate and identify clouds in these maps. A Quick Cloud region Proposal(QCP) was designed to segment cloud maps to the greatest extent. Then we use the convolutional neural network to extract features of the cloud maps and use fully connected network to predict the type of cloud. In addition, a voting mechanism with a variety of methods has achieved good prediction results.

Detection and classification of X-ray baggage security imagery are of great significance not only in daily life. However, up to now, it is still difficult for the traditional image processing methods to detect and distinguish objects which we want to detect in X-Ray baggage security imagery. Moreover, it is even more challenging to classify different types of objects. Although the current state-of-the-art detectors have achieved impressive performance on various public datasets with visible images, these detectors fail to deal with objects in X-ray images. This paper proposes an object detection algorithm for X-Ray baggage security screening images. Firstly, to outline the detected object from X-Ray baggage security imagery, we propose a fore-ground-background segmentation method which based on color information. Then, to classify and outline different type of object in X-Ray im-age, a deep convolutional neural networks (DCNNs) based object detection framework Faster R-CNN are proposed. In this stage we also use transfer learning method to speed up Faster R-CNN. The proposed method is proved to achieve 77% mAP by in a real-word dataset which contains 32,253 sub-way X-Ray baggage security screening images.

Recommendation system is able to recommend items that are likely to be preferred by the user. Hybrid recommender systems combine the advantages of the collaborative filtering and content-based filtering for improved recommendation. Hybrid recommendation methods use as many significant factors as possible to generate recommendation, which is practically very functional in real scenarios. However, such method has not been applied to book recommendation yet. Thus, in this paper, we propose a set of novel features which can be categorized into three types: latent features, derived features and content features. These features can be combined to form a new hybrid feature vector containing rating information and content information. Then, we adopted learning-to-rank to use the proposed feature vector as the input for book recommendation. Collaborative Ranking (CR) and Probabilistic Matrix Factorization (PMF) are compared with our proposed method. The experimental results show that the proposed method outperforms CR and PMF. It shows that, on [email protected], PMF achieves 0.713818, CR achieves 0.690072 vs. our method achieves 0.742689 which is 4.04% over PMF and 7.62% over CR.

A good concept drifting stream classifier should have the following two characteristics: 1) sensitive to the new concept when concept drifts; 2) have stable high accuracy when concept is stable. Most published methods and algorithms may succeed in one aspect while neglecting the other. In this paper, we proposed an adaptive ensemble classifier for concept drifting stream classification which focuses on both the above two aspects called AEC. Our AEC includes two stages: the online stage and the offline stage, and three phases: classifier updating, ensemble classifiers reconstruction and component classifier subset selection for final decision. We take a new online bagging classification model that is based on incremental learner such as Naive Bayes classifier and can keep enough history information and create good diversity between different component classifiers. Then we take an offline scheme to do the ensemble reconstruction and ensemble subset selection: that is to drop certain number of classifiers periodically and use only a portion of the whole ensemble, the combination of which may yield better accuracy to make classification. Experiment justifies the superiority of our model in both accuracy and sensitivity in the concept drifting environment.