Search

• Bicomplex embeddings outperform real, complex, and quaternion spaces. • Circular and hyperbolic rotations improve temporal dynamics modeling. • Proven capability to infer complex relational patterns, both theoretically and experimentally. • Handles both time point and time span data effectively. • TBicomR shows performance gains on all benchmark datasets. Temporal knowledge graphs (TKGs) capture relationships and entities evolving over time, making event prediction a challenging task due to the complex temporal and relational dynamics. In this work, we propose BiCoTime, a novel model using bicomplex embeddings to represent entities, relations, and time. While quaternions capture asymmetric relations through non-commutativity, bicomplex numbers provide a commutative algebraic structure, ideal for modeling both symmetric and asymmetric relations. Unlike quaternions, bicomplex embeddings maintain interpretability in symmetric relations while preserving key algebraic properties like distributivity. Temporal rotations further enhance BiCoTime's ability to model the interaction between relations and time, capturing how entities and relationships evolve. This combination of bicomplex embeddings and temporal rotations ensures a more interpretable and accurate modeling of TKGs. Our experiments show that TBiComR achieved a 21% improvement in Mean Reciprocal Rank (MRR) on the ICEWS14 dataset, which emphasizes time points, and a 15% improvement on the YAGO11k dataset, which focuses on time spans. The choice of bicomplex numbers balances computational complexity and expressive power, offering efficient training and better predictive performance compared to models using quaternions or octonions. [ABSTRACT FROM AUTHOR]

Knowledge-aware recommender systems, which utilize knowledge graphs (KGs) to enrich item information, have been shown to improve the accuracy and explainability of recommendations. Besides, KGs are further explored to determine the intent of choosing items (i.e., the reason why users select items of interest). Conventional methods represent intents either as sets of relations in a KG or as KG entities. However, such approaches fail to fully leverage the combined information provided by both entities and relations. To address this issue, we propose a new KG-based user Intent Extraction Framework (KIEF) to capture user intents at a more fine-grained level for recommendation. Specifically, we propose a novel intent representation constructed with relation-aware entity representation, encouraging finer granularity for user intents. Furthermore, since a KG may contain noisy information that impairs the quality of user intent, it is compulsory to consider which factors in a KG are important to represent a user's intent. Thus, we introduce global intent which are comprehensive features for the entire interactions of all users and local intent, which are empirical features of individual users from personal history. By maximizing mutual information between global and local intents, KIEF captures user preference for items. Through extensive experiments on four real-world benchmark datasets, we prove the superior performance of KIEF over the state-of-the-art and analyze interpretable explanations for understanding user intents. [ABSTRACT FROM AUTHOR]

Due to the Knowledge Graph (KG) construction process, erroneous triples are virtually inevitable to be introduced into real-world KGs. Since these errors hinder the expressiveness and applicability of KGs, the development of knowledge graph error detection (KGED) methods is necessary. Despite the overall effectiveness of current KGED methods, their capacity to identify challenging errors is limited. In this work, we conduct empirical studies and find that previous works introduce structural and semantic bias, impeding the identification of erroneous triples, especially in challenging cases. To address this issue, we design a causal graph for the KGED task and propose a Dual De-confounded Causal Intervention (DuDCI) method for debiasing. Firstly, DuDCI utilizes the neighborhood and textual descriptions of triples to calculate their graph and text embeddings. Next, a Causal De-confounded Module is constructed to mitigate the impact of shortcuts caused by the bias through the front-door adjustment. Furthermore, we introduce Disentanglement Constraints to disentangle the information expressed by each embedding, thereby facilitating further bias mitigation. Experimental results on three widely used KGED datasets validate the effectiveness of DuDCI and demonstrate that DuDCI outperforms current KGED methods, with an improvement of at least 2.2%, especially in more challenging noise scenarios. [ABSTRACT FROM AUTHOR]

The process of developing new drugs is both time-consuming and costly, often taking over a decade and billions of dollars to obtain regulatory approval. Additionally, the complexity of patent protection for novel compounds presents challenges for pharmaceutical innovation. Drug repositioning offers an alternative strategy to uncover new therapeutic uses for existing medicines. Previous repositioning models have been limited by their reliance on homogeneous data sources, failing to leverage the rich information available in heterogeneous biomedical knowledge graphs. We propose HeteroKGRep, a novel drug repositioning model that utilizes heterogeneous graphs to address these limitations. HeteroKGRep is a multi-step framework that first generates a similarity graph from hierarchical concept relations. It then applies SMOTE over-sampling to address class imbalance before generating node sequences using a heterogeneous graph neural network. Drug and disease embeddings are extracted from the network and used for prediction. We evaluated HeteroKGRep on a graph containing biomedical concepts and relations from ontologies, pathways and literature. It achieved state-of-the-art performance with 99% accuracy, 95% AUC ROC and 94% average precision on predicting repurposing opportunities. Compared to existing homogeneous approaches, HeteroKGRep leverages diverse knowledge sources to enrich representation learning. Based on heterogeneous graphs, HeteroKGRep can discover new drug-disease associations, leveraging de novo drug development. This work establishes a promising new paradigm for knowledge-guided drug repositioning using multimodal biomedical data. [ABSTRACT FROM AUTHOR]

Zero-shot relation triplet extraction (ZeroRTE) endeavors to extract relation triplets from a test set using a model trained on a training set with disjoint relations from the test set. Current ZeroRTE approaches primarily rely on two strategies: 1) Combining pre-trained language models to generate additional training samples; 2) Adding a large number of parameters that require training from scratch on top of a pre-trained language model. However, the former approach does not ensure the quality of generated samples, and the latter often struggles to generalize to unseen relations in the test set, particularly when the training set is small. In this paper, we introduce a novel method, N ext S entence P rediction for R elation T riplet E xtraction (NSP-RTE), abstracting ZeroRTE as a higher-level next sentence prediction (NSP) task to enhance its generalization ability to unseen relation categories. NSP-RTE integrates modules for relation recognition, entity detection, and triplet classification, leveraging pre-trained BERT models with fewer parameters requiring training from scratch, while eliminating the need for additional sample generation. Our experiments on the FewRel and Wiki-ZSL datasets demonstrate that NSP-RTE, with its simple and efficient design, significantly outperforms previous methods. • This paper transforms the zero-shot relation triplet extraction task into a next-sentence prediction problem for enhanced generalization. • The proposed method directly addresses the zero-shot relation triplet extraction task based on contextual semantics, eliminating the need for additional sample synthesis. • Experimental results on publicly available datasets demonstrate that the proposed method significantly outperforms previous methods. [ABSTRACT FROM AUTHOR]

The massive spread of rumors on social networks has caused serious adverse effects on individuals and society, increasing the urgency of rumor detection. Existing detection methods based on deep learning have achieved fruitful results by virtue of their powerful semantic representation capabilities. However, the centralized training mode and the reliance on extensive training data containing user privacy pose significant risks of privacy abuse or leakage. Although federated learning with client-level differential privacy provides a potential solution, it results in a dramatic decline in model performance. To address these issues, we propose a Federated Graph Knowledge Distillation framework (FedGKD), which aims to effectively identify rumors while preserving user privacy. In this framework, we implement anonymization from both the feature and structure dimensions of graphs, and apply differential privacy only to sensitive features to prevent significant deviation in data statistics. Additionally, to improve model generalization performance in federated settings, we learn a lightweight generator at the server to extract global knowledge through knowledge distillation. This knowledge is then broadcast to clients as inductive experience to regulate their local training. Extensive experiments on four publicly available datasets demonstrate that FedGKD outperforms strong baselines and displays outstanding privacy-preserving capabilities. [ABSTRACT FROM AUTHOR]

Knowledge Graph Embedding (KGE) based deep neural networks contribute to recommender systems in diverse application scenarios. However, Catastrophic Forgetting (CForg) significantly degrades their performance. Although exemplar replay is commonly adopted as a possible remedy to alleviate the intensity of CForg, a trade-off between performance and complexity occurs in this process. Therefore, in this work, we introduce C ontinual K nowledge graph embedding enhancement for joint I nteraction-based N ext click recommendation (CKIN) to defy the CForg and assuage the complexity. Typically, we introduce the Semantic Relevance Estimation (SRE) technique to ensure information relevance by filtering out irrelevant-data and reducing the space complexity. We introduce the SRE-enhanced deep probabilistic technique to probably replay the most relevant exemplars to defy the CForg and reduce the time complexity. Moreover, we introduce the integration of locality-preserving loss into the KGE framework to optimize the loss. In substantial experiments on real-world datasets, CKIN outperforms the baseline methods by effectively meeting the highlighted challenges. • Introduce SRE to automatically store relevant exemplars. • Introduce SDP to probabilistically replay the exemplars. • Introduce integration of LPL as a distillation-loss into the BCE loss. • Design CGNN to learn representations and continually update KGEB. • CKIN outperforms the baseline methods. [ABSTRACT FROM AUTHOR]

Knowledge graph embedding (KGE), as a pivotal technology in artificial intelligence, plays a significant role in enhancing the logical reasoning and management efficiency of downstream tasks in knowledge graphs (KGs). It maps the intricate structure of a KG to a continuous vector space. Conventional KGE techniques primarily focus on representing static data within a KG. However, in the real world, facts frequently change over time, as exemplified by evolving social relationships and news events. The effective utilization of embedding technologies to represent KGs that integrate temporal data has gained significant scholarly interest. This paper comprehensively reviews the existing methods for learning KG representations that incorporate temporal data. It offers a highly intuitive perspective by categorizing temporal KGE (TKGE) methods into seven main classes based on dynamic evolution models and extensions of static KGE. The review covers various aspects of TKGE, including the background, problem definition, symbolic representation, training process, commonly used datasets, evaluation schemes, and relevant research. Furthermore, detailed descriptions of related embedding models are provided, followed by an introduction to typical downstream tasks in temporal KG scenarios. Finally, the paper concludes by summarizing the challenges faced in TKGE and outlining future research directions. [ABSTRACT FROM AUTHOR]

Recently, there has been an increasing interest in extracting and annotating tables on the Web. This activity allows the transformation of textual data into machine-readable formats to enable the execution of various artificial intelligence tasks, e. g., semantic search and dataset extension. Semantic Table Interpretation (STI) is the process of annotating elements in a table. The paper explores Semantic Table Interpretation, addressing the challenges of Entity Retrieval and Entity Disambiguation in the context of Knowledge Graphs (KGs). It introduces LamAPI , an Information Retrieval system with string/type-based filtering and s-elBat , an Entity Disambiguation technique that combines heuristic and ML-based approaches. By applying the acquired know-how in the field and extracting algorithms, techniques and components from our previous STI approaches and the state of the art, we have created a new platform capable of annotating any tabular data, ensuring a high level of quality. [ABSTRACT FROM AUTHOR]

To inhibit the spread of rumorous information and its severe impacts, fact checking aims at retrieving relevant evidence to verify the veracity of a given statement. Fact checking methods typically use knowledge graphs (KGs) as external repositories and develop reasoning mechanism to retrieve evidence for verifying the statement. Existing fact checking methods have focused on verifying the statement of a single claim expressed by a clause. However, as real-world rumorous information is usually complex and a textual statement is often composed of multiple clauses (i.e. represented as multiple claims instead of a single one), multi-claim fact checking is not only necessary but more important for practical applications. Multi-claim statements imply rich contextual information and modeling the interactions of multiple claims can facilitate better verification. In this paper, we propose a knowledge enhanced learning and semantic composition model for multi-claim fact checking. Our model consists of two modules, KG-based learning enhancement and multi-claim semantic composition. To fully utilize the contextual information implied in multiple claims, the KG-based learning enhancement module learns the dynamic context-specific representations via selectively aggregating relevant attributes of entities. To robustly verify multiple claims robustly, the multi-claim semantic composition module learns a unified representation for multiple claims by modeling inter-claim interactions, and then verify them as a whole on the basis of this. We conduct experimental studies to validate our proposed method, and the experimental results on three typically datasets confirmed the efficacy of our model for multi-claim fact checking. • Propose the first model to address the multi-claim fact checking problem. • KG-based module learns context-specific entity representations. • Graph-based module models inter-claim semantic interactions. • Experimental results on the benchmark KGs show the effectiveness of our model. [ABSTRACT FROM AUTHOR]

A taxonomy represents a hierarchically structured knowledge graph that forms the infrastructure for various downstream applications, including recommender systems, web search, and question answering. The exploration of automated induction from text corpora has yielded notable taxonomies such as CN-probase, CN-DBpedia, and Zhishi.schema. Despite these efforts, existing taxonomies still face two critical issues that result in sub-optimal hierarchical structures. On the one hand, commonly observed taxonomies exhibit a coarse-grained and "flat" structure, stemming from a noticeable lack of diversity in both nodes and edges. This limitation primarily originates from the biased and homogeneous data distribution. On the other hand, the semantic granularity among "siblings" within these taxonomies remains inconsistent, presenting a challenge in accurately and comprehensively identifying hierarchical relations. To address these issues, this study introduces a novel taxonomy induction framework composed of three meticulously designed components. Initially, we established a seed schema by leveraging statistical information from external data sources as distant supervision to append nodes and edges containing "generic semantics", thereby rectifying biased data distributions. Subsequently, a clustering algorithm is employed to group the nodes based on their similarities, followed by a refinement operation of the hierarchical relations among these nodes. Building on this seed schema, we propose a fine-grained contrastive learning method in the expansion module to strengthen the utilization of taxonomic structures, consequently boosting the precision of query-anchor matching. Finally, we meticulously scrutinized the hierarchical relations between each query and its siblings to ensure the integrity of the constructed taxonomy. Extensive experiments on real-world datasets validated the efficacy of our proposed framework for constructing well-structured taxonomies. • We propose a framework for producing well-structured and comprehensive taxonomies. • Distant supervision enhances seed taxonomy diversity, addressing data skewness. • Contrastive learning refines query-anchor matching and ensures semantic consistency. [ABSTRACT FROM AUTHOR]

Aiming to support Retrieval Augmented Generation (RAG) in the design process, we present a method to identify explicit, engineering design facts – { head entity:: relationship:: tail entity} from patented artefact descriptions. Given a sentence with a pair of entities (selected from noun phrases) marked in a unique manner, our method extracts their relationship that is explicitly communicated in the sentence. For this task, we create a dataset of 375,084 examples and fine-tune language models for relation identification (token classification task) and relation elicitation (sequence-to-sequence task). The token classification approach achieves up to 99.7% accuracy. Upon applying the method to a domain of 4,870 fan system patents, we populate a knowledge base of over 2.93 million facts. Using this knowledge base, we demonstrate how Large Language Models (LLMs) are guided by explicit facts to synthesise knowledge and generate technical and cohesive responses when sought out for knowledge retrieval tasks in the design process. [ABSTRACT FROM AUTHOR]

In this study, we focus on the task of selecting high-quality answers selection in knowledge-graph-based (KG-based) conversational question answering (ConvQA) system. Effectively exploring a user's intention and modeling historical interaction records are challenging. To address this challenge, we propose the D ialog gr A ph e N hanced promp T l E arning (DANTE) model, which simultaneously integrates sequential and structural information from questions and interactive logic. While the structural information was exploited in previous studies by simply converting it into linear strings in a "pre-train, predict" paradigm, DANTE comprises the use of a novel graph representation for jointly modeling the QA pairs, relevant KG paths, and dialog contexts. The dialog graph constructs in both the turn-level and dialog-level, where DANTE fuses the structural and sequential information deeply in a "pre-train, prompt, and predict" manner. The experimental results showed that DANTE improves the absolute points by 7.1% and 8.2% in terms of the P@1 and mean reciprocal rank metrics, respectively, on the ConvQuestions/ConvRef benchmark compared with state-of-the-art baselines. [ABSTRACT FROM AUTHOR]

Knowledge Graph (KG) completion through reasoning over KGs is one of the most concerned issue in research around knowledge graphs. Current work focuses more on studying individual KG, while a series of related multi-source KGs are more common in real applications. In order to effectively exert their complementary roles in completing each KG under the premise of ensuring data privacy, federated learning becomes the focus of attention. However, existing methods ignore the explicit complementarity among KGs in terms of structural information. And they generally align embeddings of overlapping instances across clients, which constrains flexibility of each client model. Based on this, we propose the first GNN-based federated embedding framework for multiple knowledge graph completion, GFedKG. It focuses on the complementary role among KGs in topology structures and adaptively generates different aggregated neighboring features for entities in distinct clients. Such treatment achieves some consistency among clients while retaining the flexibility of each client model to characterize respective KG. Moreover, entity embeddings that represent the features of entities are better privacy-preserved since neighboring embeddings are transmitted instead of them. And with the aim of realizing more refined and accurate aggregation of neighboring features from multiple clients, we innovatively propose a context-aware graph neural network, along with the context-enhanced evaluation function. Homophily neighbors are also taken into account to cope with the low-homophily aspect of KGs. Extensive experiments demonstrate the advantages of GFedKG. [ABSTRACT FROM AUTHOR]

In practical applications, knowledge graphs are typically associated with various types of multimodal data, including images. As representation learning methods based on unimodal knowledge graphs face performance challenges, there is a growing interest in multimodal knowledge graphs. However, the fusion of semantic information (the deeper meanings that lurk in context) and graph structure requires improvement. A few recent methods have combined image-semantic information only at the model level. Herein, we propose a model called cross-level multimodal semantic embedding learning (CLMSE). Initially, CLMSE leverages semantic similarities among images to enhance the data distribution of knowledge graphs. Subsequently, it utilizes the pseudoconnections produced to enhance its predictive performance. CLMSE extracts semantic data on the graph structure across varying neighborhood ranges using a higher-order and second-order decoder. The experiments show that CLMSE surpasses other multimodal knowledge graph embedding learning models in predictive performance, achieving state-of-the-art results. In a more in-depth analysis, we investigate the benefits of utilizing image-semantic information to optimize the knowledge graph distribution for various models, highlighting the effectiveness of the method. • Propose new modal fusion approach using dataset distribution and graph topology. • Introduce QGNN, improving embedding fusion over traditional GNNs. • Conduct tests on three benchmarks; show optimization enhances MKG learning. [ABSTRACT FROM AUTHOR]

Predicting future events through temporal knowledge graph reasoning has attracted widespread attention from researchers in recent years. Since future events are unknown, a comprehensive understanding of the semantic correlations and historical evolution patterns of entities and relations is a prerequisite for accurate reasoning. However, current models frequently overlook the semantic correlations of relations and fail to consider that the impact of historical repetitive facts on the query varies over time. To tackle these issues, a novel temporal knowledge graph reasoning model based on Relation Graphs and Time-Guided Attention mechanism (RG-TGA) within a local–global framework was proposed. Specifically, RG-TGA designed a relation graph construction and aggregation method in the local information encoder to capture the structural characteristics of relations. Additionally, an improved relation-aware graph attention network was performed to aggregate the local neighbors of entities. In the global information encoder, RG-TGA adopted a time-guided attention mechanism to allocate attention weights to historical repetitive facts based on their time spans. Finally, temporal knowledge graph reasoning was executed by integrating the local and global information. Experimental results on five public datasets show that RG-TGA outperforms the state-of-the-art model by 2.63%, 3.95%, 3.49%, and 2.71% in MRR, Hits@1, Hits@3, and Hits@10, respectively. [ABSTRACT FROM AUTHOR]

In temporal knowledge graphs (TKGs), the status of facts is intricately tied to the dynamic and precise nature of temporal factors. Existing research merely treats time as supplementary information, without considering the latent semantic changes caused by the positional changes of entities within specific relations in a temporal context. Furthermore, due to the coarse granularity of timestamps in existing TKGs, the number of multiple relations pattern among entities significantly increases, limiting model performance. This paper proposes a T ime- E volving Information and L atent S emantics model (TELS), which represents facts as dual quaternion embeddings to provide a compact and elegant representation. Specifically, we use timestamp dual quaternions, transforming the entity and relation into temporal entity and temporal relation through dual quaternion multiplication. Besides, we introduce semantic-aware dual quaternion to capture the latent semantics arising from the positional changes of entities within specific relations. Next, TELS consists of two parts: (a) We use semantic-aware dual quaternions to perform transformations on head entity and tail entity respectively through dual quaternion multiplication. Next, we utilize temporal relation to transform head entity to tail entity through dual quaternion multiplication. (b) We adopt an evolutionary hierarchical factor to encapsulate the differences in modulus distribution between the temporal head entity and temporal tail entity. In this way, TELS not only uses dual quaternions to handle key patterns and multiple relations pattern, but also handles evolutionary hierarchical patterns by capturing the modulus distribution differences between temporal entities. Meanwhile, TELS learns semantic-aware dual quaternion embeddings to capture the latent semantics endowed by relations to entities. Empirically, TELS can boost the performance over seven temporal knowledge graph benchmarks. [ABSTRACT FROM AUTHOR]

Knowledge Graphs, fundamental to intelligent applications, are increasingly critical in various domains, enhancing tasks like precise searching and personalized recommendation. Effectively representing entities and relationships in these graphs is key, especially as the Transformer model, despite its representational prowess, faces challenges in adapting to the graph's structure and complex relations. In this work, we present the Transformer with Relation-pattern Adaptive Contrastive Learning for Knowledge Graph Embedding (TracKGE). Specifically, TracKGE transforms the structural information of the knowledge graph into a sequence format that is more manageable for Transformers. In addition, we employ a relation-pattern adaptive contrastive learning module to capture a richer semantic and complex relationship pattern information of the knowledge graph. Lastly, by introducing a mask node model, it addresses the issue of incomplete information in the knowledge graph, further enhancing the model's capability to capture implicit relationships within it. To evaluate the performance of our model, we have chosen well-established models as baselines and executed link prediction tasks on four renowned datasets. Our experimental results reveal that our model excels in representing the semantics and intricate structures of Knowledge Graphs. It outperforms other advanced baseline models, showcasing its superior capability in handling complex data representations. [Display omitted] • TracKGE: Advanced KGE using Transformer for semantic insights. • Triple2Path: Transforming graphs to paths for better Transformer use. • New contrastive learning for complex relations in KGE. • Mask node mode: Enhancing hidden relationship detection. [ABSTRACT FROM AUTHOR]

Knowledge graph embedding (KGE) aims to capture rich semantic information about entities and relationships in KGs, which is essential for Knowledge Graph Completion (KGC) and various downstream tasks. Existing KGE models differentiate between entity and relationship embeddings by constructing indirect pretext tasks and scoring functions to discern different types of triplets. In contrast, this paper introduces a novel KGE method called Contrastive Predictive Embedding (CPE), which dispenses with the need for defining scoring functions or negative sampling. Specifically, CPE directly predicts embeddings for unknown entities based on the known entity and relationship embeddings in triplets and compares them with the true embeddings. Additionally, this paper proposes a special optimization approach to enhance the performance of various Translation-based models. Experimental results on four benchmark KGs demonstrate that CPE improves the performance of original KGE models while maintaining lower computational complexity. On the FB15k-237 dataset, CPE enhances the MRR and Hit @ k (k ∈ { 1 , 3 , 10 }) metrics of TransE by 1.55%, 3.37%, 4.58%, and 5.92%, respectively. [ABSTRACT FROM AUTHOR]

Recent advancements in knowledge graph embedding have enabled the representation of entities and relations in continuous vector spaces. Performing link prediction on incomplete knowledge graphs using these embeddings has emerged as a challenging task, drawing considerable research interest. Knowledge graphs in the real world typically exhibit a natural hierarchical structure. Hyperbolic embedding methods have shown promising results in modeling hierarchical data, especially in low-dimensional representations. However, existing hyperbolic models for knowledge graph embedding are generally shallow, resulting in limited expressiveness. Additionally, some deep hyperbolic models exist but are often constrained to fixed curvature spaces, which may not optimally capture varying hierarchical structures. In this work, we propose DeER, a novel multi-layer hyperbolic model that leverages a trainable curvature space, marking an advancement in the depth and adaptability of hyperbolic modeling for knowledge graph embeddings. We specifically design a hyperbolic convolutional neural network to learn the heterogeneous interactions between entities and relations within this adaptable hyperbolic space. Additionally, we introduce a hyperbolic feedforward neural network to transform hyperbolic features across multiple layers, enhancing the model's ability to capture complex hierarchical relationships. Experimental results on three benchmark datasets demonstrate that DeER significantly enhances expressiveness and hierarchical modeling performance when compared to both shallow and other deep baseline approaches. [ABSTRACT FROM AUTHOR]

Temporal knowledge graph (TKG) multi-hop reasoning is a dominant approach that infers the target entity by walking along the path connecting entities and relations. However, in most TKGs, there are multiple relations related to an identical entity and multiple tail entities for an identical pair of head entity and relation. This characteristic leads to a significant amount of redundant data in the binary action space which is a collection of binary tuples composed of relations and entities, thereby impeding model inference, i.e., action space explosion. We propose a method to address this issue that dismantles the reasoning process into a relation level for relation reasoning and an entity level for entity reasoning. In addition, hybrid time encoding is proposed to enhance the utilization of the timestamp information in the reasoning process. Moreover, K-means-based reward shaping is proposed to alleviate the issue of sparse reward matrices by comparing the pre-clustered labels of both the predicted entity and the target entity. A text transformer module is used to deal with the limitation of a single information modality by integrating the text information of TKG into model reasoning. Our method is evaluated using four benchmark datasets, and the results verify its superiority over state-of-the-art baselines. • Reasoning is dismantled into a relation level for relation reasoning and an entity level for entity reasoning. • Hybrid time encoding enhances the utilization of timestamp information in reasoning. • K-means-based reward shaping alleviates the issue of sparse reward matrices. • Text transformer is applied to deal with the limitation of a single information modality. [ABSTRACT FROM AUTHOR]

The design and development of text-based knowledge graph completion (KGC) methods leveraging textual entity descriptions are at the forefront of research. These methods involve advanced optimization techniques such as soft prompts and contrastive learning to enhance KGC models. The effectiveness of text-based methods largely hinges on the quality and richness of the training data. Large language models (LLMs) can utilize straightforward prompts to alter text data, thereby enabling data augmentation for KGC. Nevertheless, LLMs typically demand substantial computational resources. To address these issues, we introduce a framework termed constrained prompts for KGC (CP-KGC). This CP-KGC framework designs prompts that adapt to different datasets to enhance semantic richness. Additionally, CP-KGC employs a context constraint strategy to effectively identify polysemous entities within KGC datasets. Through extensive experimentation, we have verified the effectiveness of this framework. Even after quantization, the LLM (Qwen-7B-Chat-int4) still enhances the performance of text-based KGC methods. 1 1 Code and datasets are available at https://github.com/sjlmg/CP-KGC. This study extends the performance limits of existing models and promotes further integration of KGC with LLMs. [ABSTRACT FROM AUTHOR]

Conversational recommendation systems (CRS) aim to proactively elicit user preferences through multi-turn conversations for item recommendations. However, most existing works focus solely on user's current conversation information, which fails to capture user implicit preferences comprehensively. Moreover, these approaches primarily center around pairwise relations among data in CRS to enhance item representations, while largely overlooking the complicated relationships in CRS. To address these limitations, we propose a hypergraph-based knowledge-enhanced CRS model namely Multi-aspect Knowledge-enhanced Hypergraph Attention Network for Conversational Recommendation Systems (MKHCR). We construct three hypergraphs based on multiple aspects knowledge to mine high-order relations among data for enhancing user implicit preference representations. Specifically, we build a session hypergraph to capture high-order complicated relations in the historical conversations to explore user implicit preferences. To mitigate the data scarcity issue, we incorporate knowledge graphs and items review information, modeling them within hypergraph structure to learn complicated semantic relationships, thereby enhancing item representations. Moreover, a hypergraph attention network with a dual attention mechanism is proposed to flexibly aggregate important high-order features from these hypergraphs, which contributes to enhance user preference representations for both the recommendation and conversation generation tasks. Extensive experiments on two publicly available CRS datasets validate the effectiveness of our proposed MKHCR model, which exhibits significant improvements across key evaluation metrics, including HR@50, MRR@50, and NDCG@50, achieving enhancements of 6.76%, 9.16%, and 7.92%, respectively. [ABSTRACT FROM AUTHOR]

Recommender systems enhanced by a knowledge graph (KG) have attained widespread popularity and attention in recent years. However, traditional KG-based recommender systems encounter the challenge of gradient explosion as the network depth increases. Additionally, the abundance of unreliable paths in a KG has a detrimental impact on feature representation learning. In this article, we propose a KG-enhanced recommender system based on residual network and attention mechanism, which can capture high-order connectivity and long-range dependencies of the KG. Specifically, a resource allocation approach is employed to calculate the resource amount, which is subsequently utilized to evaluate the path reliability of the KG. After completing path extraction, we employ an attention mechanism to capture semantic correlations and structural information. To leverage the KG for enhancing recommender systems, we design a deep residual network with shortcut connections, effectively amalgamating advanced and abstract features using deep neural networks. The introduction of shortcut connections not only facilitates the fitting of residual mappings but also mitigates potential issues such as gradient explosion and convergence difficulties due to excessive network depth. Extensive experiments conducted on three standard datasets over baseline methods have demonstrated the superiority of our proposed recommender system. [ABSTRACT FROM AUTHOR]

Few-shot object detection (FSOD) can effectively improve object detection performance with limited training data, attracting increasing interest from researchers. Due to the limited sample size, there is often a large variance in the learned features, resulting in deviating from the class center and being confused with other classes. People are often able to accurately identify few-shot objects based on discriminative information. Motivated by it, we propose a discriminative prototype with dual decoupled contrast learning (DP-DDCL) method to address the issue of limited training data in FSOD. By introducing domain knowledge comprising the CLIP and attribute knowledge graph to obtain explicit and implicit semantic and visual information, we construct a novel discriminative prototype that enhances the representation of different classes (especially few-shot samples). Simultaneously, dual decoupled contrast learning consists of the decoupled of positive and negative samples, and the decoupled of discriminative prototype and instance contrast learning is proposed. It makes samples of the same class cluster around their respective class prototype while maintaining a clear semantic boundary between different classes. Extensive experiments demonstrate the efficacy of our method, surpassing the current state-of-the-art in any-shot and all-data splits, with maximum performance gains of up to +9.15% on the standard PASCAL VOC benchmark and +2.5% on the challenging COCO benchmark. [ABSTRACT FROM AUTHOR]

The electrocardiogram (ECG) signal comprising P-, Q-, R-, S-, and T-waves is an indispensable noninvasive diagnostic tool for analyzing physiological conditions of the heart. In general, traditional ECG intelligent diagnosis methods gradually extract features of the signal from input data until they can classify the ECG signal. However, the decision-making process of ECG intelligence models is implicit to clinicians. Clinical experts rely on clear and specific features extracted from ECG data to diagnose cardiac diseases effectively. Inspired by this clinical diagnosis mechanism, we propose an ECG knowledge graph (ECG-KG) framework primarily to improve ECG classification by presenting knowledge of ECG clinical diagnosis. In particular, the ECG-KG framework contains an ECG semantic feature extraction module, a knowledge graph construction module, and an ECG classification module. First, the ECG semantic feature extraction module locates the key points using the difference value method and further calculates the ECG attribute features. Further, the knowledge graph construction module utilizes attribute features to design entities and relationships for constructing abnormal ECG triples. The triples vectorize ECG abnormalities through the strategy of knowledge graph embedding strategy. Finally, the ECG classification module combines the ECG knowledge graph with the graph convolutional network model and adequately integrates expert knowledge to identify ECG abnormalities. Experiments conducted on the benchmark QT, the CPSC-2018, and the ZZU-ECG datasets show that the ECG-KG framework considerably outperforms other ECG diagnosis models, indicating the effectiveness of the ECG-KG framework for ECG abnormality diagnosis. [ABSTRACT FROM AUTHOR]

As the basis of various downstream applications such as automated question answering and knowledge reasoning, Temporal Knowledge Graph (TKG) embedding learning has attracted many interests. Recent works have explored several methods to learn dynamic embedding for TKGs and led to success in many applications, however, they are difficult to be optimized. In order to accurately catch semantic information propagation of entity-to-entity and entity-to-relation, as well as track dynamics of entities and relations in continuous time, we proposed a novel TKG embedding model ODETKGE. Instead of using recurrent neural network to model dynamics of semantic representations in continuous time, we leverage neural ordinary differential equations to model the dynamic evolution, which enable us to capture long-range dependencies between embeddings at different timestamps, track the dynamics of TKGs with irregular intervals and remain computationally efficient. Furthermore, to effectively model semantics of knowledge graph structure, we combine entity-relation collaboratively updated graph convolutional networks and neural ordinary differential equations. Additionally, to acquire important semantics such as creation and disappearance of relations, we utilize a graph transition layer. Comparing with RNN based models, the proposed ODETKGE can use fewer parameters to achieve deeper layers, and make the transformation of representations between two timestamps smoother. Extensive experiments have been executed on real-world datasets and the results verify that ODETKGE is superior than the state-of-the-art models in link forecasting, long-interval link forecasting, irregular-interval link forecasting, ablation study and training time. [ABSTRACT FROM AUTHOR]

The extraction of relational triplets in threat intelligence is a critical aspect of constructing a knowledge graph. However, the field encounters challenges like high semantic similarity among entities, limited relevance of entities, and a heavy reliance on experts, leading to low extraction efficiency. Currently, there is a lack of research on extracting threat intelligence relational triplets, which necessitates the development of an efficient extraction model. This study proposes a Span-based Multivariate Information-aware Embedding Network (SMIEN) for the joint extraction of threat intelligence relation triplets. SMIEN introduces aware embedding modules to capture fine-grained features of multivariate information, including semantic, temporal order, dependency, spans, span pairs, entity labels, and relational labels, and to enhance their interaction. Designed a Type-Aware Graph Convolutional Network (TA-GCN) to enhance the representation of key information between less relevant tokens in threat intelligence text sentences within the Dependency-Aware Embedding layer. The Entity-Aware Embedding module is designed to enhance fine-grained interaction between span information and other multivariate information, while the Relationship-Aware Embedding module is designed to determine correlations between span pairs and enhance interaction with relationship labels. Experimental results on the HACKER, RE-DNRTI, and RE-IVTI datasets demonstrate micro F1 scores of 44.73%, 81.74%, and 44.15%, respectively, highlighting the method's effectiveness in extracting cyber threat intelligence and Internet of Vehicle (IoV) threat intelligence information. In addition, we have constructed a threat intelligence ontology (CVTIO) on which we have built a preliminary threat intelligence knowledge graph (CVTIKG) with 23,636 relational triplets. The CVTIKG can be found at https://github.com/wangxtz/CVTIKG. [ABSTRACT FROM AUTHOR]

Learning disentangled entity representations has garnered significant attention in the field of knowledge graph completion (KGC). However, the existing methods inherently overlook the indicative role of relations and the correlation between latent factors and relations, leading to suboptimal entity representations for KGC tasks. In the current study, we introduce the Disentangled Relational Graph Neural Network with Contrastive Learning (DRGCL) method, designed to acquire disentangled entity representations guided by relations. In particular, we first devise the factor-aware relational message aggregation approach to learn entity representations under each semantic subspace and obtain latent factor representations by attention mechanisms. Subsequently, we propose a discrimination objective for factor-subspace pairs using a contrastive learning approach, which compels the factor representations to distinctly capture the information associated with different latent factors and promote the consistency between factor representations and semantic subspaces. Through disentanglement, our model can generate relation-aware scores tailored to the provided scenario. Extensive experiments have been conducted on three benchmark datasets and the results demonstrate the superiority of our method compared with strong baseline models. [ABSTRACT FROM AUTHOR]

In recent years, knowledge graphs (KGs) have gained considerable traction across various domains, especially in the realm of recommender systems, where their integration has garnered significant interest. These integrations aim to enhance the accuracy of recommendations by leveraging user–item interaction data and item attributes. However, existing methods encounter several challenges, including excessive smoothing, sparse data, redundancy, inadequate consideration of auxiliary information, and limitations in constructing deep networks. To address these challenges and enhance knowledge-graph-based recommendation methods that ignore auxiliary information and framework redundancy in neighborhood information aggregation, this study proposes a novel graph neural network recommendation model based on interactive embedding. This model capitalizes on both the KG and user–item interaction matrix to extract valuable information, refine aggregation methods, and optimize the overall performance. Specifically, user–relation interactive embedding is formed by establishing connections between users and relations using the user–item interaction matrix and KG as the foundation. This interactive embedding merges with the user through convolutional neural networks (CNNs), independently participating in the aggregation of neighborhood information to provide more contextual information for recommendation. The set of users who interact with items is extracted employing the user–item interaction matrix and creating item–user interactive embedding. It is then merged with the item using a CNN. Lastly, the final representations of the users and items for prediction are obtained. Experimental evaluations conducted on six real recommendation datasets demonstrate that our proposed model outperforms existing baselines. [ABSTRACT FROM AUTHOR]

Relation extraction has become a crucial step for the automatic construction of Knowledge Graph (KG). Recently, researchers leverage Sequence-to-Sequence (Seq2Seq) models for Joint Entity and Relation Extraction (JERE). Nevertheless, traditional decoding methods entail the generation of the target sequence incrementally from left to right by the decoder, without the ability to revise earlier predictions when errors occur. This limitation becomes evident when decoding errors manifest prior to the current decoding step. Furthermore, the interrelations among triplets originating from the same sentence exhibit a robust correlation, which has been overlooked. In this paper, we propose B idirectional D ecoding with Co -graph re presentation (BDCore) to address the issues mentioned above. Specifically, we first introduce a backward decoder to decode the target sequence in a reverse order. Then, the forward decoder introduces two attention mechanisms to simultaneously considering the hidden states of the encoder and the backward decoder. Thus, the backward decoding information helps to alleviate the negative impact of the forward decoding errors. Besides, we construct a relation co-occurrence graph (Co-graph) and exploit Graph Convolutional Network (GCN) to capture the relation correlation. The extensive experiments demonstrate the benefits of the proposed bidirectional decoding and co-graph representation for relation extraction. Compared to the previous methods, our approach significantly outperforms the baselines on the NYT benchmark. • The Seq2Seq approaches disregard forward decoding errors and relation co-occurrence. • GCN enables to efficiently learn node representation of relation co-occurrence graph. • Bidirectional decoding relieves the negative impact of the forward decoding errors. • Two attention mechanisms simultaneously consider backward and forward information. [ABSTRACT FROM AUTHOR]

Inductive relation prediction, which handles unseen entities at the reasoning stage, has the potential to complete continuously expanding knowledge graphs. Existing inductive approaches predominantly focus on providing alternatives predicted by the subgraph structure of the target. However, many approaches neglect the correlation between semantic relationships, which does not mine additional implicit information. In this paper, we propose MUIF, a meta-learning framework with an updated information flow that addresses these challenges. Further, MUIF aims to enhance the role of relational semantics in inductive prediction, while emphasizing the importance of distinguishing relational paths. We propose an updated information flow mechanism for strengthening the relational pass, and we use a contrastive strategy to represent strongly associated paths. Further, meta-learning is applied to transfer the updated total embedding, thereby enhancing prediction accuracy. Further, the Hits@10 of MUIF can reach up to 97.03% in the link prediction experiments, thereby demonstrating the excellent performance of the proposed approach. [ABSTRACT FROM AUTHOR]

• Methodology to represent dynamic knowledge and explore emerging communities from corpora. • Analysis of communities and centrality over time is conducted, along with a proposed method for generating high-impact relations. • A dynamic knowledge graph is generated over the last nine years using the KDD papers corpus. • Emerging publications in KDD favor the creation of novel relations instead of entities. • Cohesion of the generated communities over time is constant, and the average community size grew over time. Knowledge graphs represent relationships between entities. These graphs can take dynamic forms to trace changes along time through text models and further used by reasoning systems with the intent to answer queries. In this research we explore their applicability for extracting temporal patterns of knowledge in the form of communities. To this end, we propose a method for generating knowledge relationships over unconnected components of a knowledge graph, allowing for a targeted exploration of emerging contents in corpora. This analysis is applied to the corpora of the Conference on Knowledge Discovery and Data Mining (KDD) publications over the last decade. We find the key knowledge communities over time and rank the underlying concepts. Results show that the publication efforts increasingly focus on graph research and the creation of relationships instead of new concepts. The acquired results confirm the validity of the proposed knowledge discovery methodology for community-centered analysis of emerging changes in dynamic knowledge graphs. [ABSTRACT FROM AUTHOR]

In the age of massive data, the construction of knowledge graph has increasingly become a forceful support for the downstream applications of artificial intelligence. However, the information of entities in knowledge graphs is usually incomplete, so it is urgent to supplement the relations of entities through entity alignment task. Frustratingly, the current entity alignment models are facing serious challenges. First, some models only focus on structural features and other auxiliary information (e.g., attributes, images and descriptions), but ignore the features of the entity itself can be scaled resulting in over-smoothing issue. Second, most models utilize higher-order networks to aggregate neighborhood information by stacking layers, but the training cost of these models are drastically higher. Third, most models are supervised or semi-supervised, but there are few pre-aligned seeds for alignment, which greatly limits the improvement of model performance. Hence, to address the above three issues, we propose a Higher-Order Graph Neural Network with Local Inflation for entity alignment, named HOLI-GNN. Specifically, we introduce a local inflation mechanism, which enlarges the each feature of entities to mitigate the impact of over-smoothing caused by neighborhood aggregation. Additionally, we propose a novel higher-order encoder to capture higher-order information. Furthermore, our model also employ currently popular iteration strategy to increase labeled entity pairs, which can markedly promote the performance of align task. Finally, we perform comprehensive experiments to validate the effectiveness of our model on benchmark datasets. The results strongly indicate that our model exhibits better performance than the state-of-the-art models. [ABSTRACT FROM AUTHOR]

Deep neural networks have shown promise in various fields including image processing, natural language processing and speech processing. Despite their proven usefulness in both theory and empirical studies, significant difficulties in their application still persist. More specifically, enabling deep neural networks to reason over data for knowledge acquisition is arguably challenging. Neural networks have the ability to learn patterns and generalize over data but often lack model interpretability, i.e., it is not always clear how they reason over data. Enabling neural networks to symbolically reason is one way to bring explainability and improve knowledge acquisition. In this paper, we introduce a Deduction-Preserved Embedding (DPE) model designed to learn a distributed representation that integrates symbolic deductive reasoning. Our approach effectively incorporates symbolic rules, resulting in enhanced link prediction performance for knowledge graph completion. Our experimental results highlight that the proposed model consistently outperforms baseline methods, demonstrating superior overall performance despite having a relatively small number of parameters. Notably, the version of the model that incorporates symbolic rules surpasses the majority of compared models, particularly in scenarios characterized by data sparsity. [ABSTRACT FROM AUTHOR]

Knowledge Graph (KG) provides a structured representation of domain knowledge by formally defining entities and their relationships. However, distinct communities tend to employ different terminologies and granularity levels to describe the same entity, leading to the KG heterogeneity issue that hampers their communications. KG matching can identify semantically similar entities in two KGs, which is an effective solution to this problem. Similarity Measures (SMs) are the foundation of the KG matching technique, and due to the complexity of entity heterogeneity, it is necessary to construct a high-level SM by selecting and combining the basic SMs. However, the large number of SMs and their intricate relationships make SM construction an open challenge. Inspired by the success of Evolutionary Algorithms (EA) in addressing the entity matching problem, this work further proposes a novel Self-adaptive Designed Genetic Programming (SDGP) to automatically construct the SM for KG matching. To overcome the drawbacks of the classic EA-based matching methods, a new individual representation and a novel fitness function are proposed to enable SDGP automatically explore the SM selection and combination. Then, a new Adaptive Automatic Design (AAD) method is introduced to adaptively trade off SDGP's exploration and exploitation, which can determine the timing of AAD and efficiently determine the suitable breeding operators and control parameters for SDGP. The experiment uses the Ontology Alignment Evaluation Initiative's Knowledge Graph (KG) data set to test the performance of SDGP. The experimental results show that SDGP can effectively determine high-quality KG alignments, which significantly outperform state-of-the-art KG matching methods. • A new individual representation and a novel fitness function are presented. • A new Automated Algorithm Design method is proposed to improve SDGP's performance. • The effectiveness of SDGP is validated on OAEI's Knowledge Graph data set. [ABSTRACT FROM AUTHOR]

Introducing prior auxiliary information from the knowledge graph (KG) to assist the user–item graph can improve the comprehensive performance of the recommender system. Many recent studies have shown that the ensemble properties of hyperbolic spaces fit the scale-free and hierarchical characteristics exhibited in the above two types of graphs well. Therefore, hyperbolic-based recommender systems have achieved a series of outstanding performances. However, in existing hyperbolic methods, equivariance is not considered. Thus, they cannot generalize symmetric features under given transformations, which seriously limits the capability of the model. Moreover, they cannot balance preserving the heterogeneity and mining the high-order entity information to users across two graphs. To fill these gaps, we propose a rigorous Lorentz group equivariant knowledge-enhanced collaborative filtering (LECF) model. Innovatively, we jointly update the attribute embeddings (containing the high-order entity signals from the KG) and hyperbolic embeddings (the distance between hyperbolic embeddings reveals the recommendation tendency) via the LECF layer with Lorentz Equivariant Transformation. Moreover, we propose Hyperbolic Sparse Attention Mechanism to sample the most informative neighboring nodes. Lorentz equivariance is strictly maintained throughout the entire model, and enforcing equivariance is proven necessary experimentally. Extensive experiments on three real-world datasets demonstrate that the proposed LECF remarkably outperforms state-of-the-art methods. [ABSTRACT FROM AUTHOR]

In recent years, E-learning has gained immense popularity as a prominent mode of education. However, accurately recommending learning resources from a vast amount of data remains a significant challenge. This study addresses two primary challenges impacting recommendation performance. Firstly, an imbalance exists between the abundance of available learning resources and the limited interaction behavior of learners. Secondly, the existing algorithms often overlook dynamic preference information, focusing primarily on learners' short-term, static preferences only by learning interactive behavior but disregarding the multi-correlation between learning resources and learners. To tackle these challenges, we propose MkEGC (Multi-knowledge Enhanced Graph Convolution), a novel framework for learning resource recommendation. We approach the recommendation process as a Markov decision process. Initially, we construct a dual knowledge graph convolutional network, operating in learning resource-knowledge and learner-knowledge domains. This network facilitates the extraction of vector features from learning resources, enhances the learner vector representation, and captures higher-order preferences. Subsequently, we design hierarchical and attention weighting strategies to effectively extract latent hierarchical information from the knowledge graph. Finally, we integrate the learning resource state, the learning interaction state, and the sequence state to represent a multi-dimensional learner-state within the Markov decision framework, enabling prise learning resource recommendations. To validate the effectiveness of MkEGC, we conduct extensive experiments, comparing multiple sets of metrics with six state-of-the-art recommendation algorithms, utilizing two real-world datasets. • Exploring learners' long-term preference changes. • Optimizing the extraction ability of learners' behavioral latent characteristics. • Devising a novel multi-dimensional learner state representation method. • Employing hierarchical weighted and attention-weighted strategies. [ABSTRACT FROM AUTHOR]

Most knowledge graphs in the real world suffer from incompleteness which can be addressed through knowledge graph representation learning (KGRL) techniques that use known facts to infer missing links. In this paper, a novel multi-scale adaptive selection network for KGRL, namely SelectE, is developed to learn richer multi-scale interactive features and automatically select important features, thereby achieving promising KGRL performance. Specifically, first, the input feature matrix is redesigned to better cooperate with multi-scale convolution to improve the interaction of entities and relations. Second, a multi-scale learning module is designed to learn richer multi-scale features from the input matrix using multiple branches with different kernel sizes. Third, to automatically strengthen the important features and suppress the useless features, a multi-scale adaptive selection mechanism is proposed to dynamically allocate the weights of the obtained features based on their contained information. The core of SelectE is to maximize interactions while also considering how to better utilize features. Finally, the outstanding performance of SelectE is validated by a series of comparison experiments on seven benchmark datasets (FB15k-237, WN18RR, FB15k, WN18, YAGO3–10, KINSHIP, UMLS). The experimental results show that SelectE outperforms other state-of-the-art models, demonstrating its remarkable performance and generalization. [ABSTRACT FROM AUTHOR]

Graph neural network (GNN)-based methods have demonstrated remarkable performance in various knowledge graph (KG) tasks. However, most existing approaches rely on observing all entities during training, posing a challenge in real-world knowledge graphs where new entities emerge frequently. To address this limitation, we introduce Decentralized Attention Network (DAN). DAN leverages neighbor context as the query vector to score the neighbors of an entity, thereby distributing the entity semantics only among its neighbor embeddings. To effectively train a DAN, we introduce self-distillation, a technique that guides the network in generating desired representations. Theoretical analysis validates the effectiveness of our approach. We implement an end-to-end framework and conduct extensive experiments to evaluate our method, showcasing competitive performance on conventional entity alignment and entity prediction tasks. Furthermore, our method significantly outperforms existing methods in open-world settings. [ABSTRACT FROM AUTHOR]

A temporal knowledge graph (TKG) is a set of facts associated with different timestamps. TKG completion (TKGC) is the task of inferring unknown facts based on known facts, where mining and understanding the expressional characterization of timestamps from the fact sets are key. Expressional characterizations are complex because of two aspects: type-diversity and dependence-variability. Existing approaches do not comprehensively consider time points and time intervals; hence, facts with different time forms cannot be managed simultaneously. Additionally, these approaches overlook the interdependence and variability of timestamp positions in different types of facts, such as the relative order (e.g., whether the facts belong to the past, present, or future) and relative distance (e.g., the largest distance at which the facts occur), thus hindering the adequate perception of position. Therefore, we propose a novel TKGC model named C omplex E xpressional C haracterizations with B lock D ecomposition (CEC-BD). This model is based on the block decomposition model and exploits sophisticated interactions for different types of facts by unifying the treatment of both time points and time intervals. In addition, we introduce a time-sensitive approach that injects timestamps with embedded information regarding fact occurrences into the CEC-BD model to capture the relative order and distance. To cluster and order time embeddings more effectively, we add a bias component that is randomly initialized and learned during training and further propose a temporal smoothness scheme. In particular, the model achieved a mean reciprocal rank of 3.05% and 10 times the speedup of state-of-the-art baseline models. • We address the problems of type-diversity and dependence-variability. • For type-diversity, we unify the processing in encoding the factor matrix. • For dependence-variability, we present the temporal-position method. • Developing a new temporal smoothing function for temporal knowledge graph. [ABSTRACT FROM AUTHOR]

Recently, heterogeneous graph attention network (HGAT) has been widely applied to various machine learning tasks and achieved remarkable results with sufficient labeled data. However, it is noteworthy that in many tasks, labeled data is scarce and the data labeling process is expensive. To that end, this paper presents a novel framework for learning from data with limited labels by organically integrating active learning (AL) and semi-supervised learning (SSL) into heterogeneous graph network. Our framework consists of three components: heterogeneous information network (HIN), HGAT and active semi-supervised learning strategy (ASSL). The adjustable HIN is first constructed by fusing multi-modal features. Then the HGAT is used to encode HIN and the double-layer attention mechanism can mitigate noises during information fusion. The ASSL is further designed to enrich the high-quality labeled training data for model train. Samples selected by uncertainty-aware AL and samples labeled by pseudo-label selection based SSL are finally mixed to iteratively train the ASSL-HGAT framework. Experimental results prove that ASSL-HGAT surpasses the compared state-of-the-art methods on five representative semi-supervised benchmark datasets under the same experimental settings. • To alleviate the influence of limited labels on the performance of HGAT, we utilize active semi-supervised learning to empower heterogeneous graph attention network. • Under the uncertainty-aware mechanism, we smoothly fuse uncertainty sampling-based AL with pseudo-label-based SSL to further improve the quality of training data and the model's generalization ability. • Knowledge graphs are utilized to introduce external information to alleviate the short text sparsity problem. • Extensive experimental results show that the proposed ASSL-HGAT significantly outperforms state-of the-art methods on five benchmark datasets. [ABSTRACT FROM AUTHOR]

Multiobjective Multitasking Optimization (MO-MTO) has become a hot research spot in the field of evolutionary computing. The fundamental problem of MO-MTO is to inhibit the negative transfer phenomenon. Mining the relationship among multiple optimization tasks and identifying the effective transferred knowledge have been proven a feasible way for the inhibition of negative transfer. In this paper, the solutions of different tasks are regarded as the vertices of two separate sets. Through constructing the bipartite graph of the vertices from these two sets, the relationship of different populations can be expressed and the valuable knowledge can be identified to transfer. Furthermore, a historical knowledge correction strategy is designed to deal with some special cases when identifying knowledge by the bipartite graph. A series of experiments are conducted on two MO-MTO test suits, and the results have demonstrated the efficacy of the proposed algorithm. [ABSTRACT FROM AUTHOR]

Knowledge Hypergraph (KHG) is a higher-order extension of the Knowledge Graph (KG), and its relation prediction is based on known data to predict unknown higher-order relations, thereby providing useful knowledge services. However, the existing KHG relation algorithms still have some limitations: (i) most studies only consider the influence of the direct neighbors, and (ii) they ignore the complex interactions existing inside higher-order facts. Based on this, we propose a KHG relation prediction model HoGCNF 2 based on higher-order hypergraph convolutional network and feature fusion. Dual-channel hypergraph convolutional network considers the significant and higher-order information propagation of entities. Feature fusion strategy considers different types of higher-order structures. Besides, attention mechanism adaptively assigns weights to the learned embeddings. Extensive experiments demonstrate the superiority of HoGCNF 2 on different datasets. Specifically, the M R R result improves by 2.6% on the unfixed dataset FB-AUTO, and improves by 9.7% on the fixed dataset WikiPeople-4. Our implementations are publicly available at: https://doi.org/10.24433/CO.5584354.v1. [ABSTRACT FROM AUTHOR]

Knowledge graph embedding (KGE) and knowledge graph reasoning (KGR) aim to automatic completion of knowledge graph (KG). The difference is that most KGE models learn the embedded representation at a triple level. In contrast, KGR models focus more on optimizing decision-making and enhancing the interpretability of reasoning processes with multi-hop paths. As a result, KGE models are better at learning triplet embeddings, whereas KGR models can capture the multihop information between entity pairs. However, KGE and KGR models only focus on one aspect that affects the completion performance. This paper proposes a plug-and-play collaborative learning framework (CLF) for jointly enhancing knowledge graph embedding and reasoning, which can accommodate existing KGR and KGE models. The two models exchange training experiences in this framework to realize mutual learning through a collaborative learning module. In this module, a new distance function is designed to maintain the independence of candidate entities' probabilities and avoid information loss. Furthermore, a knowledge augmentation module is designed to identify missing key triples to assist in the further iterative training of the framework. Extensive experiments on the benchmark datasets demonstrate that our framework significantly improves the performance of existing models. [ABSTRACT FROM AUTHOR]

Plausible reasoning is an interesting and viable approach for semantic data analytics as it provides a non-deterministic and exploratory approach to inferring new knowledge from large datasets. Plausible reasoning identifies meaningful associations between data elements by analyzing the underlying semantics of the data to identify knowledge that can assist in complex decision-making, especially when dealing with incomplete and noisy knowledge. In this work, we present a plausible reasoning approach that computerizes formal plausible patterns using semantic web methods for knowledge representation and reasoning over knowledge graphs to discover new knowledge from large datasets. We implemented six plausible patterns, representing three different types of semantic relationships, by developing PL-OWL as a plausible extension to the Web Ontology Language (OWL). We incorporated the plausible patterns within our plausible reasoning framework—SeDan (SEmantics-based Data Analytics) —to provide plausible reasoning-based query answering over knowledge graphs to generate new knowledge and provide answers to complex queries. We evaluated our plausible reasoning-based query answering approach by answering medical questions from BioASQ challenges using SeDan which incorporates the Semantic MEDLINE database, DrugBank, and Disease Ontology for supplementary semantic associations. Our experimental results show that our plausible reasoning method was able to expand the query answering coverage of the Semantic MEDLINE database by 37% to generate plausible answers to previously unanswered queries—importantly 88% of the plausible answers generated were clinically reasonable and verified by a domain expert. [ABSTRACT FROM AUTHOR]

Successful software development requires a cohesive team with the right mix of technical skills and the ability to collaborate effectively. However, forming a software team that can execute tasks with precision and efficiency requires a deep understanding of each member's competence, experience, and cooperation history. Previously, automated software team selection has evaluated technical skills, cohesion, and cooperation history. However, the previous method had some limitations. Particularly, local features directly calculated from team members were subjective to the researchers' views, and the method ignored the temporal aspect of open-source software development. To overcome these limitations, this paper proposes a knowledge-graph software team recommendation framework called TeReKG. This framework encapsulates temporal collaboration patterns and uses a temporal knowledge graph to encode software collaboration history, technical abilities, task dependencies, and project structure. TeReKG was against state-of-the-art team recommendation algorithms using three popular open-source software projects: Moodle, Apache, and Atlassian. The evaluation results show that TeReKG outperforms the state-of-the-art baselines in both single-role and team recommendation tasks. These findings demonstrate that knowledge graph embedding can be effectively utilized in automated recommendation tasks in software engineering. Additionally, this highlights the potential for knowledge graphs to capture global information that can benefit various software development applications, including impact prediction of software repositories, code clone detection, and source code retrieval. [ABSTRACT FROM AUTHOR]

The relation completion for knowledge graph requires expanding and enriching a knowledge graph by predicting the missing relation in a given triple which has known head and tail entities. In this paper, we propose a unified embedding-based relation completion framework which mainly includes two contributions. Firstly, based on embedding of triples generated by any embedding model, we utilize deep neural networks to learn feature representations of relations from head and tail entities. This allows us to propose a multi-dimensional feature prediction model for missing relations of triples. Based on the predictive features of missing relations, we match the best relation within the candidate relation set for relation completion. Secondly, to reduce the impact of noisy features and further improve the effectiveness of relation completion, we consider the extraction of key features as a submodular optimization problem by establishing a normalized, non-decreasing submodular function. Finally, testing on multiple public knowledge graph datasets, the results demonstrate that our proposed relation completion framework can significantly improve existing relation completion approaches. [ABSTRACT FROM AUTHOR]

• We highlight the importance of cognitive psychology to improve click-through-rate (CTR) in recommender systems (RS) and develop a Cognitive-based Knowledge Learning Framework (CKLF) for recommendation. To our best knowledge, CKLF is the first framework which applies cognitive psychology into KG-based RS. • We propose Spreading Activation Network (SAN) and Sequence-sensitive Attention Mechanism (SAM) in CKLF. SAN employs spreading activation theory to achieve the high-order connectivity on the Knowledge Graph (KG). SAM employs the Ebbinghaus forgetting curve to track interest evolution of users. Both mechanisms make it possible to obtain diverse information associated with interactions. • We conduct extensive experiments using real-world datasets and KGs, which demonstrates the superiority of CKLF compared to state-of-art baselines and its enhanced performance in diversity and serendipity. We performed ablation and sparsity experiments to demonstrate the effectiveness of both SAN and SAM. In addition, we designed a case study to provide deeper insight on the relation between accuracy and beyond-accuracy. Recommender systems (RS) have been widely used in Web applications such as search engines, social media platforms and e-commerce portals. Accuracy-related metrics have always been the primary goals for RS. Most RS tend to recommend items that are most relevant or similar to users' historical records, which may narrow users' cognitive scope and lead to information cocoons. This paper proposes a Cognitive-based Knowledge Learning Framework (CKLF), which addresses this challenge by integrating cognitive psychology to the construction of the neural network. CKLF consists of two key parts: a Spreading Activation Network (SAN) and a Sequence-sensitive Attention Mechanism (SAM). The SAN uses spreading activation theory to activate diversified and high-ordered entities in an auxiliary knowledge graph, and to refine embeddings of users and items. The SAM incorporates the Ebbinghaus forgetting theory into the attention mechanism to capture evolution features of users' sequential interactions. Extensive experiments on two public benchmarks show, compared with start-of-the-art baselines, CKLF not only improves accuracy but more importantly achieves superior performance in the beyond-accuracy objectives such as diversity and serendipity. Therefore, CKLF can effectively alleviate information cocoons and improve the overall quality of RS. [ABSTRACT FROM AUTHOR]

Pretrained language models (PLMs) have been developed rapidly which establish impressive performance on many open-domain downstream tasks. However, conducting these pretrained models directly without additional network architectures on special domain tasks like multi-label disease diagnosis cannot perform well. Recently, prompt learning has been a new paradigm in PLM field which is more convenient and well-performed than the traditional fine-tuning approach for different domain tasks. However, prompt engineering is challenging because it takes time and experience. In this paper, we propose a new prompt learning method named Knowledge-based Dynamic PrompT (KBDPT) to deal with these problems. Firstly, we import medical knowledge into PLMs by prompt templates which make results of the disease diagnosis more reasonable and qualified. Compared with the fine-tuning approach, this method needs fewer trainable parameters and less training data but achieve better performance. Secondly, unlike most existing pre-defined prompt methods, KBDPT dynamically generates prompts based on personal medical information and a large-scale medical knowledge graph, which can provide more valuable guidance information for disease diagnosis. Lastly, the proposed model also ensembles multiple prompts from all possible diseases to introduce more knowledge and obtain differential diagnosis results. Experiments of multi-label disease diagnosis are conduct on three real-world EMR datasets. Results demonstrate that our model can be used in various pretrained models and outperform both classical deep learning methods and fine-tuning PLMs. The source code of our proposed model has been released at: https://github.com/loxs123/KBDPT. [ABSTRACT FROM AUTHOR]