Texture Feature Representation Research Articles

Video flame recognition based on LGATP texture feature and sparse representation

Video flame recognition based on LGATP texture feature and sparse representation

CBIF-M: A content-based image filtering and retrieval scheme using multi-classifier filtering framework

Recently, the image retrieval process appears to be a challenging task to filter out a huge volume of objectionable images from retrieval. It is very easy for people of all ages to obtain such obscene images with just a few clicks on the internet, which could harm the adolescent. To prevent such serious social problems, this paper introduces a content-based image filtering (CBIF) framework in an image retrieval system. The proposed CBIF approach is split into two components. In the first component, the skin region of exposed human body parts is extracted based on a color segmentation method, and an efficient feature vector generation process is introduced based on frequent skin color pixels and scale-invariant structural features. A novel texture feature representation is presented in the retrieval and filtering process obtained by block-paired local binary pattern (BP-LBP). The final component introduces a novel image filtering framework for the retrieved images found from the first component based on a majority of voting method on three different machine learning classifiers backed by SVM, MLP, and CNN. The retrieval efficiency is analyzed with standard performance indicators such as precision, recall, and F-score on two large datasets: NPDI and NSFW.

Enhanced industrial text classification via hyper variational graph-guided global context integration.

Joint local context that is primarily processed by pre-trained models has emerged as a prevailing technique for text classification. Nevertheless, there are relatively few classification applications on small sample of industrial text datasets. In this study, an approach of employing global enhanced context representation of the pre-trained model to classify industrial domain text is proposed. To achieve the application of the proposed technique, we extract primary text representations and local context information as embeddings by leveraging the BERT pre-trained model. Moreover, we create a text information entropy matrix through statistical computation, which fuses features to construct the matrix. Subsequently, we adopt BERT embedding and hyper variational graph to guide the updating of the existing text information entropy matrix. This process is subjected to iteration three times. It produces a hypergraph primary text representation that includes global context information. Additionally, we feed the primary BERT text feature representation into capsule networks for purification and expansion as well. Finally, the above two representations are fused to obtain the final text representation and apply it to text classification through feature fusion module. The effectiveness of this method is validated through experiments on multiple datasets. Specifically, on the CHIP-CTC dataset, it achieves an accuracy of 86.82% and an F1 score of 82.87%. On the CLUEEmotion2020 dataset, the proposed model obtains an accuracy of 61.22% and an F1 score of 51.56%. On the N15News dataset, the accuracy and F1 score are 72.21% and 69.06% respectively. Furthermore, when applied to an industrial patent dataset, the model produced promising results with an accuracy of 91.84% and F1 score of 79.71%. All four datasets are significantly improved by using the proposed model compared to the baselines. The evaluation result of the four dataset indicates that our proposed model effectively solves the classification problem.

Integration and Innovation of Artificial Intelligence and Traditional English Translation Methods

Abstract In this study, we employ autoregressive and neural machine translation models to process natural language. Our exploration of AI translation workflows encompasses three key aspects: the encoder-decoder framework, text feature representation, and translation derivation methodologies. Subsequently, we identify the core evaluation indices essential for assessing intelligent translation systems. An innovative translation model is constructed by amalgamating AI-driven and traditional English translation techniques. This model is scrutinized for its effectiveness through various lenses, including the quality of English-to-Chinese poetry translations and manual evaluations. Comparative analysis of the proposed translation method with other extant methods across diverse datasets reveals superior performance metrics: the BLEU score consistently exceeds 4.5 across all three test sets, and the METEOR score ranges between 4.3 and 4.6, surpassing competing methods. Additionally, translation accuracy for sentences of varying lengths in the source language is maintained between 85% and 96%. The objective of this research is to conduct a comprehensive examination of the integration of AI English translation with conventional methodologies, aiming to foster innovative theoretical contributions and practical advancements in the field.

Multimodal integration for fake news detection on social media platforms

The widespread dissemination of fake news on social media platforms can cause serious social impact, making the detection of fake news on social media platforms an urgent problem to be solved. Up to now, scholars have proposed various methods ranging from traditional manual feature extraction to deep learning algorithms for detecting fake news. However, these methods still have some limitations and two difficult problems: (1) How to learn informative news feature representations without losing information as much as possible? (2) How to effectively fuse multi-modal information to obtain high-order complementary information about news and enhance fake news detection? To overcome these two difficulties, this article proposes a multi-modal fusion fake news detection model. Firstly, the model uses BERT and VGG-19 to obtain the text and image feature representations of news content, respectively, and then further fuses multi-modal information through a multi-modal attention mechanism module to obtain high-order complementary information between different modalities, thereby obtaining informative news feature representations for fake news detection. Experimental results on two real-world public datasets demonstrate the effectiveness of our model compared to mainstream detection methods.

Zero-sample text classification algorithm based on BERT and graph convolutional neural network

Abstract In this study, we undertake a comprehensive examination of zero-shot text classification and its associated implications. We propose the adoption of the BERT model as a method for text feature representation. Subsequently, we utilize the Pointwise Mutual Information (PMI) metric to adjust the weight values within a graph convolutional neural network, thereby facilitating the construction of a text graph. Additionally, we incorporate an attention mechanism to transform this text graph, enabling it to represent the output labels of zero-shot text classification effectively. The experimental environment is set up, and the comparison and ablation experiments of the text classification model based on BERT and graph convolutional neural network with the baseline models are carried out in several different types of datasets, and the parameter settings of λ are adjusted according to the experimental results, and the convergence of the BERT model is compared to test the robustness of the model performance and the classification effect. When λ was set to 0.60, the model achieved the best results in each dataset. When the task is set to 5-way-5-shot, the convergence rate of the model for the Snippets dataset using the penultimate layer of features can reach 74%-80% of the training accuracy at the 5,000th step. The training accuracy gradually flattens out in the first 10,000 steps, and the model achieves classification accuracy in all four learning scenarios, with good stability.

Data governance and Gensini score automatic calculation for coronary angiography with deep-learning-based natural language extraction.

With the widespread adoption of electronic health records, the amount of stored medical data has been increasing. Clinical data, often in the form of semi-structured or unstructured electronic medical records (EMRs), contains rich patient information. However, due to the use of natural language by physicians when composing these records, the effectiveness of traditional methods such as dictionaries, rule matching, and machine learning in the extraction of information from these unstructured texts falls short of clinical standards. In this paper, a novel deep-learning-based natural language extraction method is proposed to overcome current shortcomings in data governance and Gensini score automatic calculation in coronary angiography. A pre-trained model called bidirectional encoder representation from transformers (BERT) with strong text feature representation capabilities is employed as the feature representation layer. It is combined with bidirectional long short-term memory (BiLSTM) and conditional random field (CRF) models to extract both global and local features from the text. The study included an evaluation of the model on a dataset from a hospital in China and it was compared with another model to validate its practical advantages. Hence, the BiLSTM-CRF model was employed to automatically extract relevant coronary angiogram information from EMR texts. The achieved F1 score was 91.19, which is approximately 0.87 higher than the BERT-BiLSTM-CRF model.

A Study of Changes in College English Online Phrases Supported by Clustering Algorithm

Abstract The hyperlinks on the Sina Weibo platform are tracked by a web crawler so as to obtain the textual resources of university English online phrases on the platform. To process microblog text data, this paper proposes the use of multiple linguistic expressions of translation machines to enhance the text feature representation and apply it to clustering to enhance the clustering results. Based on commonly used algorithms such as K-means, DBSCAN, and hierarchical clustering, we analyze the clustering effect of each algorithm according to internal and external evaluation indexes and finally determine the clustering algorithms to be used for the change of university English network terms. Using the English network word “hand” as an example, this paper observes a fluctuating trend in the heat of this word from January to June 2023, with peaks in heat generation in January and February, respectively, and a peak heat of about 10,500 in February. In terms of spatial variation, the heat level of the English word “hand” is high in the east and south regions and low in the west and north regions, but in February, due to the return of the population to their hometowns for the Lunar New Year, the north region exceeds the east and south regions with a heat level of 11086.77.

Deep feature extraction with tri-channel textual feature map for text classification

The complexity and diversity of texts make it difficult for shallow text classification models to capture deeper text features. Therefore, this paper takes advantage of the BiLSTM-CNN hybrid network based on the self-attention mechanism to extract text features, and constructs a new text feature representation similar to RGB-channel images, that is, tri-channel text feature maps. Drawing on the effectiveness of ResNet and SA-Net in deep networks, we designed a deep feature extraction network to capture deeper features in the text. We use the tri-channel text feature map as the input to the deep feature extraction network, and propose the Deeper Feature Text Classification (DFTC) end-to-end text classification model. Experiments prove that the DFTC model is competitive compared to the most advanced methods on five challenging text classification datasets.

Manifold-based sparse representation for opinion mining

What the consumer thinks about an organization's products, services, and events is a crucial performance indicator for businesses. The brief opinion pieces were quickly published on websites and social media platforms and have been analyzed by machine learning methods. The classical text feature representation methods suffer from high dimensionality, sparsity, noisy, irrelevant and redundant information. This paper focuses on how to enhance feature representation for opinion mining. Some nonlinear feature selection methods based on manifold assumption have been exploited to resolve these problems. The inherent manifold configuration was commonly ascertained through a nearest neighbor graph, whereby the neighbors in the current techniques may exhibit diverse polarities. To alleviate this burden, it is proposed to exploit both manifold assumption and sparse property as prior knowledge for opinion representation to learn intrinsic structure from data. First, the graph representation of user reviews based on the mentioned prior knowledge is learned. Then, the spectral properties of the learned graph are exploited to present data in a new feature space. The proposed algorithm is applied to four various common input features on two benchmark datasets, the Internet Movie Database (IMDB) and the Amazon review dataset. Our experiments reveal that the proposed algorithm yields considerable enhancements in terms of F-measure, accuracy, and other standard performance measures compared to the combination of state-of-the-art features with various classifiers. The highest classification accuracies of 99.15 and 91.97 are obtained in the proposed method on IMDB and Amazon using a linear SVM classifier, respectively. The impact of the parameters of the proposed algorithm is also investigated in this paper. The incorporation of a sparse manifold-based representation has led to noteworthy advancements beyond the baseline, and this success serves to validate the underlying assumptions.

Named Entity Recognition for Few-Shot Power Dispatch Based on Multi-Task

In view of the fact that entity nested and professional terms are difficult to identify in the field of power dispatch, a multi-task-based few-shot named entity recognition model (FSPD-NER) for power dispatch is proposed. The model consists of four modules: feature enhancement, seed, expansion, and implication. Firstly, the masking strategy of the encoder is improved by adopting whole-word masking, using a RoBERTa (Robustly Optimized BERT Pretraining Approach) encoder as the embedding layer to obtain the text feature representation, and an IDCNN (Iterated Dilated CNN) module to enhance the feature. Then the text is cut into one Chinese character and two Chinese characters as a seed set, the score for each seed is calculated, and if the score is greater than the threshold value ω, they are passed to the expansion module as candidate seeds; next, the candidate seeds need to be expanded left and right according to offset γ to obtain the candidate entities; finally, to construct text implication pairs, the input text is used as a premise sentence, the candidate entity is connected with predefined label templates as hypothesis sentences, and the implication pairs are passed to the RoBERTa encoder for the classification task. The focus loss function is used to alleviate label imbalance during training. The experimental results of the model on the power dispatch dataset show that the precision, recall, and F1 scores of the recognition results in 20-shot samples are 63.39%, 61.97%, and 62.67%, respectively, which is a significant performance improvement compared to existing methods.

Text Classification Based on Neural Network Fusion

The goal of text classification is to identify the category to which the text belongs. Text categorization is widely used in email detection, sentiment analysis, topic marking and other fields. However, good text representation is the point to improve the capability of NLP tasks. Traditional text representation adopts bag-of-words model or vector space model, which loses the context information of the text and faces the problems of high latitude and high sparsity,. In recent years, with the increase of data and the improvement of computing performance, the use of deep learning technology to represent and classify texts has attracted great attention. Convolutional Neural Network (CNN), Recurrent Neural Network (RNN) and RNN with attention mechanism are used to represent the text, and then to classify the text and other NLP tasks, all of which have better performance than the traditional methods. In this paper, we design two sentence-level models based on the deep network and the details are as follows: (1) Text representation and classification model based on bidirectional RNN and CNN (BRCNN). BRCNN’s input is the word vector corresponding to each word in the sentence; after using RNN to extract word order information in sentences, CNN is used to extract higher-level features of sentences. After convolution, the maximum pool operation is used to obtain sentence vectors. At last, softmax classifier is used for classification. RNN can capture the word order information in sentences, while CNN can extract useful features. Experiments on eight text classification tasks show that BRCNN model can get better text feature representation, and the classification accuracy rate is equal to or higher than that of the prior art. (2) Attention mechanism and CNN (ACNN) model uses the RNN with attention mechanism to obtain the context vector; Then CNN is used to extract more advanced feature information. The maximum pool operation is adopted to obtain a sentence vector; At last, the softmax classifier is used to classify the text. Experiments on eight text classification benchmark data sets show that ACNN improves the stability of model convergence, and can converge to an optimal or local optimal solution better than BRCNN.

STNet: shape and texture joint learning through two-stream network for knowledge-guided image recognition.

The human brain processes shape and texture information separately through different neurons in the visual system. In intelligent computer-aided imaging diagnosis, pre-trained feature extractors are commonly used in various medical image recognition methods, common pre-training datasets such as ImageNet tend to improve the texture representation of the model but make it ignore many shape features. Weak shape feature representation is disadvantageous for some tasks that focus on shape features in medical image analysis. Inspired by the function of neurons in the human brain, in this paper, we proposed a shape-and-texture-biased two-stream network to enhance the shape feature representation in knowledge-guided medical image analysis. First, the two-stream network shape-biased stream and a texture-biased stream are constructed through classification and segmentation multi-task joint learning. Second, we propose pyramid-grouped convolution to enhance the texture feature representation and introduce deformable convolution to enhance the shape feature extraction. Third, we used a channel-attention-based feature selection module in shape and texture feature fusion to focus on the key features and eliminate information redundancy caused by feature fusion. Finally, aiming at the problem of model optimization difficulty caused by the imbalance in the number of benign and malignant samples in medical images, an asymmetric loss function was introduced to improve the robustness of the model. We applied our method to the melanoma recognition task on ISIC-2019 and XJTU-MM datasets, which focus on both the texture and shape of the lesions. The experimental results on dermoscopic image recognition and pathological image recognition datasets show the proposed method outperforms the compared algorithms and prove the effectiveness of our method.

Multimodal emotion recognition based on audio and text by using hybrid attention networks

Multimodal Emotion Recognition (MER) has recently become a popular and challenging topic. The most key challenge in MER is how to effectively fuse multimodal information. Most of prior works may not fully consider the inter-modal and intra-modal attention mechanism to jointly learn intra-modal and inter-modal emotional salient information for further improving the performance of MER. To address this problem, this paper proposes a new MER framework based on audio and text by using Hybrid Attention Networks (MER-HAN). The proposed MER-HAN combines three different attention mechanisms such as the local intra-modal attention, the cross-modal attention, and the global inter-modal attention to effectively learn intra-modal and inter-modal emotional salient features for MER. Specifically, an Audio and Text Encoder (ATE) block equipped with deep learning techniques with the local intra-modal attention mechanism is initially designed to learn high-level audio and text feature representations from the corresponding audio and text sequences, respectively. Then, a Cross-Modal Attention (CMA) block is presented to jointly capture high-level shared feature representations across audio and text modalities. Finally, a Multimodal Emotion Classification (MEC) block with the global inter-modal attention mechanism is provided to obtain final MER results. Extensive experiments conducted on two public multimodal emotional datasets, i.e., IEMOCAP and MELD datasets, show the advantage of the proposed MER-HAN on MER tasks.

BDNet: A BERT-based dual-path network for text-to-image cross-modal person re-identification

Text-to-image person re-identification (TI-ReID) aims to provide a descriptive sentence to find a specific person in the gallery. The task is very challenging due to the huge feature differences between both image and text descriptions. Currently, most approaches use the idea of combining global and local features to get more fine-grained features. However, these methods usually acquire local features with the help of human pose or segmentation models, which makes it difficult to use in realistic scenarios due to the introduction of additional models or complex training evaluation strategies. To facilitate practical applications, we propose a BERT-based framework for dual-path TI-ReID. Without the help of additional models, our approach directly employs visual attention in the global feature extraction network to allow the network to adaptively learn to focus on salient local features in image and text descriptions, which enhances the network’s attention to local information through a visual attention mechanism, thus strengthening the global feature representation and effectively improving the global feature representation. In addition, to learn text and image modality invariant feature representations, we propose a convolutional shared network (CSN) to learn image and text features together. To optimize cross-modal feature distances more effectively, we propose a global hybrid modal triplet global metric loss. In addition to combining local metric learning and global metric learning, we also introduce the CMPM loss and CMPC loss to jointly optimize the proposed model. Extensive experiments on the CUHK-PEDES dataset show that the proposed method performs significantly better than the current research results, achieving a Rank-1/mAP accuracy of 66.27%/ 57.04%.

Modified GAN with Proposed Feature Set for Text-to-Image Synthesis

Automated synthesis of practical images from the text could be useful and interesting; however, present AI systems are yet far from this objective. Nevertheless, in current years, powerful and generic Recurrent Neural Network (RNN) structures were introduced to train discriminative text feature representation. In the meantime, Deep Convolutional GANs have started producing highly convincing images of specified categories, like room interiors, album covers, and faces. In this research work, we plan to develop a new model for text-to-image synthesis, which contains three important phases: (i) feature extraction, (ii) text encoding, and (iii) optimal image synthesis. Initially, the text features such as improved TF–IDF, bag of words, and N-gram are extracted from the text and they are trained by Bi-LSTM. During the encoding of an image from text, cross-modal feature grouping is performed. Further, the image is synthesized using modified GAN (MGAN) with a new loss function. Here, for precise synthesis of images, the weights of GAN are optimized using Self-improved Social Ski-Driver (SI-SSD) optimization algorithm. Eventually, the superiority of the suggested model is examined via an assessment over existing schemes.

A Deep Neural Network-Based Approach to Media Hotspot Discovery

In recent years, with the rapid development of social network media, it has become a valuable research direction to quickly analyze these texts and find out the current hotspots from them in real time. To address this problem, this paper proposes a method to discover current hotspots by combining deep neural networks with text data. First, the text data features are extracted based on the graphical convolutional neural network, and the temporal correlation of numerical information is modeled using gated recurrent units, and the numerical feature vectors are fused with the text feature vectors. Then, the K-means algorithm is optimized for the initial point selection problem, and a clustering algorithm based on the maximum density selection method in the moving range is proposed. Finally, the text feature representation method based on graph convolutional neural network is combined with the clustering algorithm based on the moving range density maximum selection method to build a deep learning-based media hotspot discovery framework. The accuracy of the proposed media hotspot discovery method and the comprehensive evaluation of the computing time have been verified experimentally.

Learning Pixel Affinity Pyramid for Arbitrary-Shaped Text Detection

Arbitrary-shaped text detection in natural images is a challenging task due to the complexity of the background and the diversity of text properties. The difficulty lies in two aspects: accurate separation of adjacent texts and sufficient text feature representation. To handle these problems, we consider text detection as instance segmentation and propose a novel text detection framework, which jointly learns semantic segmentation and a pixel affinity pyramid in a unified fully convolutional network. Specifically, the pixel affinity pyramid is proposed to encode multi-scale instance affiliation relationships of pixels, which is not only robust to varying shapes of text but also provides an accurate boundary description for separating closely located texts. In the inference phase, a simple but effective post-processing is presented to reconstruct text instances from the semantic segmentation results under the guidance of the learned pixel affinity pyramid, achieving good accuracy and efficiency. Furthermore, to enhance the representation of text features in the neural network, two modules — the Region Enhancement Module (REM) and Attentional Fusion Module (AFM) — are proposed. The REM models the semantic correlations of regional features to enhance the features from the text area, which effectively suppresses false-positive detection. The AFM adaptively fuses multi-scale textual information through an attention mechanism to obtain abundant text semantic features, which benefits multi-sized text detection. Extensive ablation experiments are conducted demonstrating the effectiveness of the REM and AFM. Evaluation results on standard benchmarks, including Total-Text, ICDAR2015, SCUT-CTW1500, and MSRA-TD500, show that our method surpasses most existing text detectors and achieves state-of-the-art performance, denoting its superior capability in detecting arbitrary-shaped texts.

Rainstorm-Induced Emergency Recognition from Citizens’ Communications Based on Spatial Feature Extraction and Transfer Learning

The rapid recognition of rainstorm-induced emergencies (e.g., building and road inundation, facility damage, and trapped citizens) is vital to timely disaster response. One big challenge that limits the performance of emergency recognition is the data imbalance between different emergency domains. The present study aims to develop an effective cross-domain transfer learning framework for rainstorm-induced emergency recognition based on the text reports provided by citizens. The critical component of the framework is the use of joint distribution adaption (JDA) analysis embedded in a discriminative feature mapping procedure, which transfers rich knowledge learned from large-scale datasets to the learning task from small emergence data. Considering the feature incompleteness that is caused by short text length, a basic probability assignment function is constructed and applied to extract important spatial features for rainstorm emergency recognition, with an improved marginal Fisher analysis being adopted to optimize cross-domain text feature representation. The proposed scheme is validated using the empirical data of ten emergency classes from Wuhan City, China. Our experimental results show that the proposed method could significantly address data imbalance and thus help achieve high recognition performance through domain knowledge complementation. Meanwhile, the use of various spatial features is proved to be effective in tackling missing features. This scheme can be further developed into smart systems for rainstorm disaster response with reasonable performance and imbalanced sample sizes.

Evaluation on Fine-grained Cross Modal Image and Text Retrieval Method Based on Deep Adversarial Hashing Technology

Fine-grained cross modal image and text retrieval is one of the hot topics in recent years, with the goal of achieving efficient and accurate image and text retrieval in different modal datasets. In this paper, a new type of cross-modal graphic retrieval method based on deep adversarial hashing is used to learn the feature representation of images and text through deep network, and realize the feature hash representation through adversarial training. The adversary network is introduced into the network structure. By maximizing the mutual information of hash coding and minimizing the anti classification loss of hash coding, the feature representation of image and text is highly consistent and separable in the hash coding space. Experiments have proved that compared with the existing algorithm, this algorithm can be retrieved better.