The goal of community search in heterogeneous information networks (HINs) is to identify a set of closely related target nodes that includes a query target node. In practice, a size constraint is often imposed due to limited resources, which has been overlooked by most existing HIN community search works. In this paper, we introduce the size-bounded community search problem to HIN data. Specifically, we propose a refined (k, P)-truss model to measure community cohesiveness, aiming to identify the most cohesive community of size s that contains the query node. We prove that this problem is NP-hard. To solve this problem, we develop a novel B\&B framework that efficiently generates target node sets of size s. We then tailor novel bounding, branching, total ordering, and candidate reduction optimisations, which enable the framework to efficiently lead to an optimum result. We also design a heuristic algorithm leveraging structural properties of HINs to efficiently obtain a high-quality initial solution, which serves as a global lower bound to further enhance the above optimisations. Building upon these, we propose two exact algorithms that enumerate combinations of edges and nodes, respectively. Extensive experiments on real-world datasets demonstrate the effectiveness and efficiency of the proposed methods.

Multi-modal knowledge graph completion (MMKGC) leverages visual, textual, and structural information to predict missing entities in knowledge graphs. However, existing approaches predominantly rely on early fusion strategies that project heterogeneous modalities into unified representations, diminishing modality-specific discriminative capabilities and failing to fully exploit inter-modal complementarity. Unlike prior fusion-based models that merge all modalities into a single latent space, our approach treats each modality as an independent reasoning expert and integrates their complementary outputs through an interpretable ensemble mechanism. To address these limitations, we propose the Multi-Branch Ensemble with Cross-Modal Fusion (MBE-CF) framework, which preserves modality-specific characteristics through independent reasoning pathways while capturing cross-modal interactions via dedicated fusion mechanisms. Our approach introduces three key innovations: (1) a multi-modal attention fusion (MMAF) module that employs multi-head self-attention mechanisms to capture cross-modal interactions and obtain robust fused multi-modal representations, (2) a context triple encoder (CTE) that processes head-relation-tail sequences through modality-specific Transformer encoders to inject relational context into entity representations, and (3) an Ensemble Inference (EI) approach that uses independent Tucker decoders for structural, visual, textual, and multi-modal branches, aggregates per-branch scores at inference, and optimizes during training a Decision Fusion objective that sums branch-specific head and tail losses. The framework maintains architectural symmetry across modalities while facilitating comprehensive integration of heterogeneous reasoning patterns to leverage the complementary strengths of different information sources. Extensive experiments on three benchmark datasets (DB15K, MKG-W, and MKG-Y) demonstrate that MBE-CF achieves state-of-the-art performance, outperforming 9 baseline methods across multiple evaluation metrics.

Multi-modal Knowledge Graph Completion (MKGC) aims at leveraging multi-modal information to infer missing objective facts in incomplete multi-modal knowledge graphs, thereby significantly enhancing their expressive capabilities. The segmentation of semantic data, including image segmentation and word-level descriptions, often contain implicit relationships between entities that are frequently overlooked by existing methodologies, thus limiting the effectiveness of reasoning tasks. Therefore, we propose a novel completion inference method based on fine-grained semantic segmentation, which enhances reasoning capability by utilizing implicit relationships between entities. Primarily, we introduce the concept of Semantic Related Entity (SRE) and a novel SRE selection algorithm, which captures the semantic neighboring relationships of entities based on segmentation semantic similarity to fully exploit the semantic association information. Subsequently, we propose a Multi-modal Related Entity Fusion Transformer (M-REFT) model to effectively utilize SREs from semantic modalities and neighbors from structural modality for completion inference. The M-REFT employs a hierarchical Transformer architecture to encode the fusion modality representation between each entity and its SREs, and then decode the triplet representation with the neighbor information to identify missing entities in incomplete triplets. We conducted extensive comparative experiments with several state-of-the-art models on three datasets, demonstrating the significant performance advantages of M-REFT. A series of ablation experiments and case studies further validate the rationality and necessity of the SRE concept and the SRE selection algorithm.

Graph data has been widely applied due to its powerful expressive capabilities. The release of raw graph data without preprocessing may lead to privacy information leakage. Thus, generating privacy-protected graphs is necessary for data analysis. Current privacy protection methods in graphs focus on securing attributes like degree distribution, triangle counts, and node information, but they often overlook the need to protect user group relationships. Additionally, some privacy-preserving graph publishing methods introduce significant noise due to the chosen graph generation techniques and the points at which noise is added. This paper aims to propose an effective graph synthesis algorithm by using differential privacy named DPCAG (Differentially Private Community Affiliation Graph Generation Model) for protecting user group relationships. Firstly, it is observed that there are numerous small probabilities in the adjacency matrix D generated by the affiliation matrix F, directly utilizing it to construct graph G would result in the generation of a substantial number of redundant edges. Therefore, we introduce a generating threshold theta to filter out unnecessary edges. Secondly, to achieve a better balance between data availability and the level of privacy protection, two budget allocation schemes are designed based on the introduction of k-truss to describe the tightness of group relationships. Lastly, we demonstrate the proposed model satisfies differential privacy mathematically and the effectiveness of DPCAG is validated using four real graph datasets.

Open-domain multi-modal dialogue system heavily relies on visual information to generate contextually relevant responses. The existing open-domain multi-modal dialog generation methods ignore the complementary relationship between multiple modalities, and are difficult to integrate with LLMs. To tackle these challenges, we introduce AutoGraph, an innovative method for constructing visual context graphs automatically. We aim to structure complex information and seamlessly integrate it with large language models (LLMs), aligning information from multiple modalities at both semantic and structural levels. Specifically, we fully connect the text graphs and scene graphs, and then trim unnecessary edges via LLMs to automatically construct a visual context graph. Next, we design several graph sampling grammar for the first time to convert graph structures into sequence which is suitable for LLMs. Finally, we propose a two-stage fine-tuning strategy to allow LLMs to understand graph sampling grammar and generate responses. We validate our proposed method on text-based LLMs, and visual-based LLMs, respectively. Experimental results show that our proposed method achieves state-of-the-art performance on multiple public datasets.

Abstract Hierarchical Text Classification (HTC) is a formidable task which involves classifying textual descriptions into a taxonomic hierarchy. Existing methods, however, have difficulty in adequately modeling the hierarchical label structures, because they tend to focus on employing graph embedding methods to encode the hierarchical structure while disregarding the fact that the HTC labels are rooted in a tree structure. This is significant because, unlike a graph, the tree structure inherently has a directive that ordains information flow from one node to another—a critical factor when applying graph embedding to the HTC task. But in the graph structure, message-passing is undirected, which will lead to the imbalance of message transmission between nodes when applied to HTC. To this end, we propose a unidirectional message-passing multi-label generation model for HTC, referred to as UMP-MG. Instead of viewing HTC as a classification problem as previous methods have done, this novel approach conceptualizes it as a sequence generation task, introducing prior hierarchical information during the decoding process. This further enables the blocking of information flow in one direction to ensure that the graph embedding method is better suited for the HTC task and thus resulted in the enhanced tree structure representation. Results obtained through experimentation on both the public WOS dataset and an E-commerce user intent classification dataset demonstrate that our proposed model can achieve superlative results.

Abstract Automatic License Plate detection and Recognition (ALPR) is a key problem in intelligent transportation systems with wide applications in traffic monitoring, electronic toll collection (ETC), intelligent parking lots (IPLs), and elsewhere. Although numerous methods have been proposed in the literature, it is still challenging to strike a good balance between the accuracy and efficiency of ALPR. In this paper, a novel end‐to‐end CNN‐based model is proposed, called Fast and Accurate Network with Feature Enhancement (FAFEnet), to jointly detect the license plates and recognize the characters with high accuracy and efficiency. Specifically, the FAFEnet model seamlessly integrates two CNN‐based models, namely the detection and recognition modules, into a unified framework to reduce accumulated errors and computational overheads in two separate steps. The detection module is a lightweight model with only seven convolutional layers yet achieves over 99.8% accuracy rates for license plate detection across all datasets. The recognition module utilizes two feature enhancement blocks to compensate and enhance the shallow character features extracted by the detection module. Furthermore, the joint optimization of detection and recognition modules exploits the feature association in two modules, and thus improves the prediction accuracy while reducing the execution time. Finally, extensive experimental results on several real‐world datasets demonstrate that FAFEnet outperforms all the competitors in terms of both accuracy and efficiency.

Subgraph matching is one fundamental issue in the research area of graph analysis, which has a wild range of applications, including question answering, semantic search and community detection. Recent studies have designed some near-optimal matching orders and data indexes to reduce the unpromising redundant calculations. However, the influences of recalculation on some positive candidate vertices were ignored in the iterative process of subgraph matching. In this paper, the novel concepts of an anchored node pair and anchored relationship are proposed as the theoretical basis to solve recalculation on subgraph matching. Thus, two key aspects are considered to reduce the redundant and recalculated node pairs. The first aspect involves using the dominating relationship of the anchored node and its follower to prune the negative node pairs, and an index of matching-driven flow graph is built to minimize the positive candidate vertices by using a heuristic algorithm. The second aspect involves exploring the anchored relationship to analyze the recalculated region of a matching stream, and two novel strategies are designed to manipulate the intermediate results of the partial subgraph isomorphism to avoid revalidation in the subgraph matching process. Extensive empirical studies on real and synthetic datasets demonstrate that our techniques outperform the state-of-the-art algorithms.

Complex Event Process (CEP) is very essential in Semantic Web of Things (SWoT) that deploy a large number of sensor devices, like smart traffic and smart city. CEP mainly solves heterogenous problems of stream data processing, where streaming data is connected to internet by a mass of wireless sensor devices. The core work of CEP is rule updating. Existing researches of rule updating are designed for static environments, and it is quite laborious to transplant those rules for dynamic environments. To enhance the portability of event rules, a method of automatic rule updating based on machine learning is proposed to learn the rules of a dynamic environment. Experimental results reveal that the proposed methods are effective and efficient.

In order to solve the challenges brought by the operation and maintenance of power system in the era of big data, APM (Application Performance Management) system is introduced, which can monitor the operation of software and hardware system, show the health of system operation, and find the performance bottleneck. On the Hadoop platform, a big data deep mining and analysis platform based on map / reduce mode is built, integrating regression analysis, association analysis, data classification, data clustering, text mining, web mining and other data mining algorithms. It can complete 100TB level data retrieval in 30s, and then analyze; the system monitoring server can run stably in a cluster of 256 nodes. The use of APM system can prevent performance bottlenecks, greatly reduce the response time of performance problem processing, and quickly locate the location of performance problems, so as to ensure higher availability and stability of information system.