Detecting ships from synthetic aperture radar (SAR) images is crucial in both military and civilian applications. In order to solve the problems of false and missed detection in complex scenes due to weak SAR image features, more small-size targets and large scale variations, we propose a context-guided and multi-scale fusion lightweight SAR ship detection model (CGMSF-SAR). First, the backbone network part incorporates the self-designed Multi-scale Group Convolutional Interaction Enhancement (MGCIE) module into the C2f module, which divides the input channels and assigns them to different sizes of convolutional kernels to capture various features across scales and reduce the computational redundancy. In the neck network part, an adaptive context-guided feature pyramid network (ACGFPN) is proposed, which captures the axial global context from horizontal and vertical directions through the pyramid context augmentation module and multiscale feature fusion module, explicitly models the rectangular key regions, and then uses dynamic interpolation fusion and multilayer recursive fusion to enhance the multiscale feature representation capability. Finally, a Dynamic Lightweight Detection Head (DLDH) is proposed to achieve efficient detection under low computation by GroupNorm to enhance the performance, shared convolution to reduce the parameters, and Scale layer to adapt to the target scale change. Experimental results on publicly available datasets show that the detection accuracy of CGMSF-SAR is significantly improved.

The core challenge in multimodal sentiment analysis is how to efficiently fuse features from text and images while ensuring inter-modal feature alignment and disparity preservation. However, existing approaches usually have limitations in (1) not being able to make full use of inter-modal semantic relationships for feature alignment and fusion, and (2) lacking effective mechanisms to decouple inter-modal shared and private information for good disparity preservation. To solve these problems, we propose a multimodal sentiment analysis based on feature alignment and differential decoupling model. Firstly, BERT and ViT are used for unimodal feature extraction for text and images respectively, and contrast learning mechanism is used to introduce contrast loss in the shared semantic space to optimise the directional consistency of modally aligned features. Secondly, we propose a bidirectional cross-modal attention mechanism to capture inter-modal affective associations and complementary information through figure-text and text-image interactive alignment features. Then, we propose a dynamic gated fusion mechanism to suppress the noise that may be introduced during the alignment process, and to assign weights and perform fusion between the original features and the aligned features, ensuring a balanced representation of modality-shared and modality-specific information. In addition, to alleviate the problem of modal features being diluted during the alignment and fusion process, we decompose modal features into shared and private features, which preserve modal-specific information while performing further alignment in the shared space. Finally, the contrast learning loss is utilized as a supervised signal to optimize the feature representation. Experimental results on public datasets show that the method proposed in this paper outperforms superior previous methods and improves the accuracy of emotion recognition.

Multimodal sentiment analysis (MSA) is primarily used to detect the sentiment contained in different modalities. Existing research has overlooked the multi-scale information of image within multimodal data, leading to insufficient extraction of image features by the network, which consequently reduces the overall performance of the model. Inspired by Mamba’s powerful sequence modeling capabilities, we propose a Multi-Scale Sequence Fusion Model for Multimodal Sentiment Analysis. The model fuses image and text features at each image scale, treating the multiple fusion results as a complete sequence, and utilizes Mamba to capture the potential sequential sentiment features within this sequence across different modalities. Specifically, we propose the Local Fusion Module (LFM) and the Global Sequence Sentiment Module (GSSM). LFM enhances the module’s understanding of context information and spatial relationships by integrating image features of different scales with text information. GSSM treats the fused information from multiple LFMs as sequential information, improving the model’s ability to capture global context. Additionally, to help the sequence interaction between multi-scale image and text, we design the Dual-modal Transformation Module to facilitate the structural transformation of image and text features. Experiments on publicly available datasets demonstrate that our proposed model outperforms previous approaches, indicating its feasibility and effectiveness in the multimodal sentiment analysis task.

User-generated multimodal data can provide powerful sentiment clues for sentiment analysis task. Existing works have aligned common sentiment features in different modalities through various multimodal fusion methods. However, these works have certain limitations: (1) Previous research works only align common sentiment features between image and text, without fully exploring interactions among these features, leading to suboptimal analysis results. (2) Redundant noise in image and text increases the risk of feature misalignment during cross-modal alignment. To address these issues, we propose a Multimodal Semantic Fusion Network (MSFN) to deeply explore the semantic relationship between image and text for Multimodal Sentiment Analysis (MSA). Specifically, we align image region and text word features related to sentiment by using a gated attention mechanism. Subsequently, we employ graph convolutional networks to model the interactions among these features to obtain explicit sentiment semantics. The proposed gated attention mechanism corrects potential feature misalignment during cross-modal alignment using a gating mechanism. Moreover, considering not all image–text pairs have explicit corresponding sentiment features, we integrate implicit sentiment semantics to our model for enhancing reliability in analysis. Experimental results on benchmark datasets demonstrate the effectiveness of our proposed model compared to baselines.

Personality detection identifies personality traits in text. Current approaches often rely on deep learning networks for text representation but they overlook the significance of psychological language knowledge in connecting user language expression to psychological characteristics. Consequently, the accuracy of personality detection is compromised. To address this issue, this paper presents PS-GCN, a model integrating Psychological knowledge and Sentiment semantic features through Graph Convolution Networks. Firstly, the Bi-LSTM network captures local features of preprocessed sentences to accurately represent the output of sentence sentiment features. Secondly, GCNs map psycholinguistic knowledge, forming semantic networks of entities and relationships. P-GCN is designed to capture the dependency information between psycholinguistic features, while S-GCN utilises syntactic structure analysis to gather more abundant information features and enhance semantic understanding ability. Finally, attention calculation is employed to reinforce key features and weaken irrelevant information. Additionally, a sentence group model captures combined features of related sentences, effectively utilising the text structure to mine sentimental features. Experimental results on multiple datasets demonstrate that the proposed method significantly improves the classification accuracy in personality detection tasks.

To further enhance the accuracy of the Chinese patent classification, this paper proposes a model, based on the patent structure and takes the patent claim as subjects, with multi-instance multi-label learning as the main method. Firstly, the patent claims are divided into multiple independent texts using the sequence number as the splitting token. For each patent, multiple claims are regarded as multiple instances, and the corresponding IPCs serve as its multiple labels. Next, the concept of secondary_label is introduced following the composition rules of IPC, and the relationships between instances and multiple secondary_labels are mined through the construction of fully-connected layers. To capture more comprehensive semantic information of instances, BIGRU and self-attention are employed to enhance semantics and reduce information loss during the training process. Finally, the max-pooling operations are utilised to obtain the predicted categories of patents based on capturing the relationships between instances and different hierarchical labels. Experimental results on the '2017 Chinese patent dataset' demonstrate that the multi-instance multi-label approach can effectively mine deeper relationships between patents and labels in classification tasks. As a result, our model significantly improves the accuracy of patent text classification.

Patent Entity and Relation Extraction (PERE) aims to extract entities and entity-relation triples from unstructured patent texts. PERE is one of the fundamental tasks in patent text mining, providing crucial technical support for patent retrieval and technology opportunity discovery. Previous works struggle to capture the implicit semantic information hidden within overlapping triples, especially a large number of overlapping triples existing in patent texts. A Patent Entity and Relation Extraction model based on Context query and Axial attention is proposed, named PERE-CA. As for entity recognition, the text segment is regarded as candidate entity span and entity types are acquired by span classification. Subsequently, the semantic context related to an entity pair is calculated by a context query method. And the semantic context is integrated into entity pair representation. For relation extraction, axial attention is implemented to get the implicit semantic information among overlapping entity pairs. And then, the model outputs all valid entity-relation triples. Experimental results on the patent dataset TFH-2020 and the public dataset SciERC demonstrate that the implementation of context query and axial attention can effectively improve extraction performance.

Dynamic measurement via deep learning can be applied in many industrial fields significantly (e.g.electrical power load and fault diagnosis acquisition).Nowadays, accurate and continuous loading measurement is essential in coal mine production.The existing methods are weak in loading measurement because they ignore the symbol characteristics of loading and adjusting features.To address the problem, we propose a hybrid learning method (called ScTCN-LightGBM) to realize the loading measurement of industrial material effectively.First, we provide an abnormal data processing method to guarantee raw data accuracy.Second, we design a sided-composited temporal convolutional network that combines a novel transposed dimensionality-reduction convolution residual block with the conventional residual block.This module can extract symbol characteristics and values of loading and adjusting features well.Finally, we utilize the light-gradient boosting machine to measure loading capacity.Experimental results show that the ScTCN-LightGBM outperforms existing measurement models with high metrics, especially the stability coefficient R 2 is 0.923.Compared to the conventional loading measurement method, the measurement performance via ScTCN-LigthGBM improves by 40.2% and the continuous measurement time is 11.28s.This study indicates that the proposed model can achieve the loading measurement of industrial material effectively.

Process variants usually exhibit a high degree of internal heterogeneity, in the sense that the executions of the process differ widely from each other due to contextual factors, human factors, or deliberate business decisions. Understanding differences among process variants helps analysts and managers to make informed decisions as to how to standardise or otherwise improve a business process. Existing process variant mining approaches typically fall short in full supporting semantic process variability mining, especially rarely taking activity behaviour relationships and trace context semantic into consideration. Here, we propose a semantic process variant discovery method, aimed at solving the difficulty of distinguishing similar-but-different behaviours directly from event logs. More specifically, we adapt concepts of benchmark logs and trace context tree to formalise context semantic of event log, to classify benchmark logs into several parts, thereby the clustered trace cohorts are mapped to discover the configurable process variants. In the experimental part, some performance metrics of the proposed method are evaluated and calculated by real-world event logs, supporting the usefulness of the proposed method. The experimental results show that the proposed method is able to distinguish similar-but-different behaviours and is superior to the characteristic trace clustering method using conventional neural networks.

Mapping-based cross-domain recommendation (CDR) can effectively tackle the cold-start problem in traditional recommender systems.However, existing mapping-based CDR methods ignore datasparse users in the source domain, which may impact the transfer efficiency of their preferences.To this end, this paper proposes a novel method named Neighbor Interaction-based Personalized Transfer for Cross-Domain Recommendation (NIPT-CDR).This proposed method mainly contains two modules: (i) an intra-domain item supplementing module and (ii) a personalised feature transfer module.The first module introduces neighbour interactions to supplement the potential missing preferences for each source domain user, particularly for those with limited observed interactions.This approach comprehensively captures the preferences of all users.The second module develops an attention mechanism to guide the knowledge transfer process selectively.Moreover, a meta-network based on users' transferable features is trained to construct personalised mapping functions for each user.The experimental results on two real-world datasets show that the proposed NIPT-CDR method achieves significant performance improvements compared to seven baseline models.The proposed model can provide more accurate and personalised recommendation services for cold-start users.

As a foundation task of natural language processing, text classification is widely used in information retrieval, public opinion analysis, and other related tasks.Facing the problem of sparse features of Chinese short texts, which affects the classification accuracy of Chinese short texts, this paper proposes a Chinese short text classification method based on the Character Frequency Sub-word Enhancement (CFSE), which can effectively improve the classification accuracy of Chinese short texts.First, the initial Chinese-character sequence is mapped to the corresponding Character Frequency Sub-word (CFS) sequence based on the global character 1 frequency information.Second, the relationship features among data are extracted based on BiLSTM-Att processing CFS sequence, and the semantic features of the initial Chinese-character sequence are obtained through ERNIE.Finally, these two kinds of features are fused and input into the text classifier to obtain the classification results.Experimental results show that the proposed method can improve the classification accuracy of Chinese short texts.

Aspect-based sentiment analysis (ABSA) aims to automatically identify the sentiment polarity of specific aspect words in a given sentence or document. Existing studies have recognised the value of interactive learning in ABSA and have developed various methods to precisely model aspect words and their contexts through interactive learning. However, these methods mostly take a shallow interactive way to model aspect words and their contexts, which may lead to the lack of complex sentiment information. To solve this issue, we propose a Lightweight Multilayer Interactive Attention Network (LMIAN) for ABSA. Specifically, we first employ a pre-trained language model to initialise word embedding vectors. Second, an interactive computational layer is designed to build correlations between aspect words and their contexts. Such correlation degree is calculated by multiple computational layers with neural attention models. Third, we use a parameter-sharing strategy among the computational layers. This allows the model to learn complex sentiment features with lower memory costs. Finally, LMIAN conducts instance validation on six publicly available sentiment analysis datasets. Extensive experiments show that LMIAN performs better than other advanced methods with relatively low memory consumption.

Fake review detection has the characteristics of huge stream data processing scale, unlimited data increment, dynamic change, and so on. However, the existing fake review detection methods mainly target limited and static review data. In addition, deceptive fake reviews have always been a difficult point in fake review detection due to their hidden and diverse characteristics. To solve the above problems, this article proposes a fake review detection model based on sentiment intensity and PU learning (SIPUL), which can continuously learn the prediction model from the constantly arriving streaming data. First, when the streaming data arrive, the sentiment intensity is introduced to divide the reviews into different subsets (i.e., strong sentiment set and weak sentiment set). Then, the initial positive and negative samples are extracted from the subset using the marking mechanism of selection completely at random (SCAR) and Spy technology. Second, building a semi-supervised positive-unlabeled (PU) learning detector based on the initial sample to detect fake reviews in the data stream iteratively. According to the detection results, the data of initial samples and the PU learning detector are continuously updated. Finally, the old data are continually deleted according to the historical record points, so that the training sample data are within a manageable size and prevent overfitting. Experimental results show that the model can effectively detect fake reviews, especially deceptive reviews.

Abstract Due to the low adjustment accuracy of manual prediction, conventional programmable logic controller systems can easily lead to inaccurate and unpredictable load problems. The existing multi-agent systems based on various deep learning models has weak ability for advanced multi-parameter prediction while mainly focusing on the underlying communication consensus. To solve this problem, we propose a hybrid model based on a temporal convolutional network with the feature crossover method and light gradient boosting decision trees (called TCN-LightGBDT). First, we select the initial dataset according to the loading parameters' tolerance range and supply supplementing method for the deviated data. Second, we use the temporal convolutional network to extract the hidden data features in virtual loading areas. Further, a two-dimensional feature matrix is reconstructed through the feature crossover method. Third, we combine these features with basic historical features as the input of the light gradient boosting decision trees to predict the adjustment values of different combinations. Finaly, we compare the proposed model with other related deep learning models, and the experimental results show that our model can accurately predict parameter values.

Conventional manual-programmable logic controller systems have confronted the problems of the unbalance load and the unreasonable bins allocation in industrial loading field. Furthermore, various optimisation models with multi-agent systems have been proposed for the single-layer scheduling and communicating, which results in either a high time cost or a difficult multi-target regression. In this paper, we propose a hierarchical-fuzzy bins allocation method and a multi-parameter adjustment values prediction model in the multi-agent collaborative control system. The method intuitively achieves topgallant and hierarchical bins allocation by different fuzzy rule bases. The multi-parameter adjustment values prediction model utilising parallel-multi LSTM(PM-LSTM) is located on the accurate multi-parameter prediction. First, new loading reference standards and an abnormal data procession method are adopted for the dataset collection. Second, the LSTM-1 is used to extract the time-series features in the loading process. Third, a two-dimensional and reconstructed matrix integrates comprehensive features with the feature crossover method. The matrix will be used as inputs to predict the adjustment value of multi parameters by the LSTM-2. Finally, the relationship model among multi parameter values is built and fitted. Experiment results show better effects for the reasonable bins allocation and balanced industrial loading.

The classification performance of a k-nearest neighbour (KNN) method is dependent on the choice of the k neighbours of a query. However, it is difficult to optimise the performance of KNN by choosing appropriate neighbours and an appropriate value of k. Moreover, the performance of KNN suffers from the use of a simple majority voting method. To address these three issues, we propose a new locally adaptive k-nearest centroid neighbour classification based on the average distance (AD-LAKNCN) in this paper. First, the k neighbours of the query based on the nearest centroid neighbour (NCN) are found, and the discrimination classes with different k values are derived from the number and distribution of each class of neighbours considered in the query. Then, based on the distribution information in the discrimination class for each k, the adaptive k and the final classification result are obtained. The experimental results based on 24 real-world datasets show that the new method achieves better classification performance than nine other state-of-the-art KNN algorithms.

Sentiment classification can provide the decision support of social applications such as trend judgment, public opinion monitoring, etc. However, the accuracy of sentiment classification for Chinese Weibo is still not satisfactory due to the complexity of Chinese. In addition, affected by the different organisational structure levels, the sentiment tendency of fewer Weibo Comments may be judged to be the opposite. To solve the problem above, this paper presents a Chinese sentiment classification model based on extended sentiment dictionary and organisational structure of comments. First, the sentiment dictionary can be extended by using seven dictionaries, which include the base sentiment dictionary and six additional dictionaries. Then, the sets of rules are constructed, which include inter-sentence rules and organisational structure rules. Finally, comments on three hot topics are crawled and used to make the data sets for sentiment calculation. Accordingly, based on the result of sentiment calculation, sentiment classification is completed. The effectiveness of the proposed model is verified through comparison experiments, and the experimental results are also discussed.

Attribute-level sentiment element extraction aims to obtain the word pair < opinion target, opinion word > from texts, which mainly obtain fine-grained evaluation information in the attribute level. Due to the information fragmentation and semantic sparseness of product reviews, it is difficult to capture more comprehensive local information from unstructured texts, which leads to the incorrect extraction of some word pairs. Aimed at the problem, this paper proposes a framework for Attribute-Level Sentiment Element Extraction (ALSEE) towards product reviews. Firstly, a small amount of sample data is selected by random sampling, and multiple features (including part of speech, word distance, dependency relationship and semantic role) which are labelled. The labelled data are used as the training set. Then, the Condition Random Field (CRF) model is applied to extract opinion targets (OT) and opinion words (OW). The self-training strategy is used to achieve the semi-supervised learning of CRF model through iterative training. Finally, target-opinion word pairs with modifying relationship are obtained by dependency parsing. Compared with the existing methods, the proposed framework can effectively extract attribute-level sentiment elements though experimental results.

The existing Emotion-Cause Pair Extraction (ECPE) has made some achievements, and it is applied in many tasks, such as criminal investigations. Previous approaches realised extraction by constructing different networks, but they did not fully exploit the original information of the data, which led to low extraction precision. Moreover, the extraction precision will also be decreased when the model is attacked by adversarial samples. To address the above problems, a new model CL-ECPE is proposed in this article to improve the extraction precision through contrastive learning. First, contrastive sets are constructed by adversarial samples. The contrastive sets are used as the raw data of adversarial training and the test data of the pilot experiment. Then, adversarial training is used to get contrastive features according to the training target. The acquisition of contrastive features can improve extraction precision. Experimental results on the benchmark emotion cause corpus show our method outperforms the state-of-the-art method by over 12.49%, as well as demonstrates the strong robustness of CL-ECPE.

Chinese short comment texts have the characteristics of feature sparseness, interlacing, irregularity, etc., which makes it difficult to fully grasp the overall emotional tendency of users. In response to such problem, the text proposes a new method based on ELECTRA and hybrid neural network. This method can more accurately capture the emotional features of the text, improve the classification effect, enhance the evaluation feedback mechanism, and facilitate user decision-making. First, in the embedding layer, ELECTRA model is used to replace BERT model, which can avoid the inconsistency of the mask training and fine-tuning process of the traditional pre-training model. Then, in the training layer, the self-attention mechanism and the BiLSTM are selected to obtain the fine-grained semantic representation information of the review text more comprehensively. Finally, in the output layer, the softmax classifier classifies the input corpus according to the sentiment characteristics of the Chinese short text. The experimental results show that the proposed model has an efficiently improvement in accuracy and there are some discoveries about the training effect of the pre-training model on text sentiment analysis tasks.

It is challenging to build domain-specific emotional lexicon for film reviews, due to its unique characteristics, such as massive data, endless new login words, and others. To improve the accuracy of film reviews classification, this article proposes a method for extending emotional lexicon based on word distance and point mutual information. First, using the improved K-means++ algorithm to cluster and select seed words with obvious emotional tendencies. Next, the Distance of Word and Point Mutual Information (DW-PMI) algorithm is presented to determine the emotional polarity of emotional words in the domain of film reviews. Four types of vocabulary, including degree adverb, negation, emoticon and emotion dictionary in the film reviews domain are added to the basic emotion dictionary to extend the film reviews emotional lexicon. From the experimental results, the expanded emotional lexicon of the Chinese film reviews can improve the accuracy and preciseness of the film reviews emotion analysis.

Association link network (ALN) can organize massive Web information to provide many intelligent services in our big data society. Effective semantic layered technologies not only can provide theoretical support for knowledge discovery in Web resources, but also can improve the searching efficiency of related information systems such as Web information system and industrial information system. How to realize the layer division of association semantic by the hierarchy analysis of ALN is an important research topic. To solve this problem, this paper proposes a hierarchy-cutting model of association semantic. First, experiments of four types of keywords with different linking roles are conducted to discover the possible distribution law. Experimental results show that these keywords with association role reveal previous power-law distribution. Then, based on the discovered power-law distribution, up-cutting and down-cutting points are presented to divide the association semantic into three layers. At the same time, theories of the hierarchy-cutting model are presented. Finally, examples of current core topic and permanent topics belonging to a domain are given. The experiments show that hierarchy-cutting points have high accuracy. The multilayer theory of association semantic can provide a theoretical support for knowledge recommendation with different particle sizes on ALNs.

Web information contain plentiful, significant knowledge which is eager to be explored by users.Effective semantic layered technology not only can provide theoretical support for knowledge discovery in Web resources, but also can improve the searching efficiency of the related information system.This paper builds the multi-layer theory of association semantic based on the power-law distribution of linking keywords.First, some experiments of four types of keywords with different linking role are done to discover the possible distribution law.Experiment results show that four types of keywords are all reveal power-law distribution.Then, based on the discovered power-law distribution, the multi-layer theory of association semantic is built.The multi-layer theory of association semantic can provide a theoretical support for knowledge recommendation with different particle size on Association Link Network (ALN).

Summary Association Link Network (ALN) can organize massive news data to support many intelligent Web applications. The degree estimating of nodes in ALN, including out‐degree and in‐degree, is an important and significant research. It can provide effective support for some applications such as the control of network structure, the rapid positioning of Web resources in ALN. This paper proposes a model for estimating the out‐degree of any one node in ALN from semantic feature view, which can greatly reduce the searching scope for the rapid positioning of Web resources stored in large‐scale database. First, we explore the main factors of forming the out‐degree of any one node from semantic feature view by qualitative analysis. Then, based on the result of qualitative analysis, we propose the model for estimating the out‐degree of any one node in ALN, including the model framework, the first estimating theory and its further optimization method. Experimental results show that the proposed estimating model as well as the optimization method have a high precision. Copyright © 2016 John Wiley & Sons, Ltd.

With the increase of information scale of web events on the time, it is extremely difficult and challenging to grasp the semantics of web events artificially, because of the limitation of the time and energy of human beings. Herein, we propose a method to map the web event to keyword level association link network (KALN) for deep analysis of the semantics of web events, such as the evolution semantics of web events. Firstly, the original KALN is constructed at a given time by traditional data mining technologies. Then, the hierarchical KALN, consisted of Theme Layer Network, Backbone Layer Network and Tidbit Layer Network, is built based on the original KALN by information entropy to identify the different semantic levels of the web event, including stable semantics, sub-stable semantics and unstable semantics. With the semantic analysis of hierarchical KALN, human could easily gain a thorough understanding of the web event. Finally, experiments show that our method can effectively capture the different level semantics of web events.

Abstract The semantically associated network on the Web is a Semantic Link Network built by mining the associated relation between Web pages. The associated link from page A to page B indicates that users who have browsed page A is likely to also browse page B. This paper explores the statistical properties of the associated network on the Web. Web pages of a specific domain are automatically downloaded by a Web crawler to build an associated network. We analyze the associated network at different domain thresholds and classify the topology into three states, that is, the original state, the kernel state and the final state. A mathematical model is built to study the in‐degree distribution, the out‐degree distribution and the total‐degree distribution for both the kernel state and the final state. By tuning the model parameters to reasonable values, we obtain the distinct power‐law forms for the three degree distributions with exponents that agree well with the statistical data. The proposed model can not only describe the evolving processes of the associated network on the Web, but also provides theory basis for complex applications such as semantic community discovery, intelligent browsing and recommendation. Copyright © 2009 John Wiley & Sons, Ltd.