2026· Poster Volume 0008 The 2026 Twenty-Second International Conference on Intelligent Computing July 23-26, 2026 Toronto, Canada· 0 citations
TL;DR
A pre-extractor model based on a hybrid architecture of rules and neural networks is introduced, which is used to identify long tail entities in the dataset and generate several candidate tail entities through relationships to improve the inference performance of the model.
Abstract
Knowledge graph completion is one of the core tasks in the field of knowledge graphs, which predicts missing links through inference of existing facts. With the advancement of deep learning technology, utilizing end-to-end deep learning models for knowledge graph completion has become a cutting-edge research direction. However, the performance of current knowledge graph completion models is still limited by text quality and incomplete structure. To address this issue, this paper proposes a method of using large models for data augmentation to improve the inference performance of the model. Specifically, we first introduced a pre-extractor model based on a hybrid architecture of rules and neural networks, which is used to identify long tail entities in the dataset and generate several candidate tail entities through relationships. Then, we use this data to have the Large Language Model infer the most factual triplet. Finally, we use the enhanced dataset for predictive inference. SAGE achieves better results on three standard KGC datasets. For instance, on the FB15K-237 dataset, compared to the SimKGC baseline model, SAGE improves Hits@1 by 1%, Hits@3 by 0.9%, and Hits@10 by 1.6%.
: Knowledge graphs organize real-world facts as structured triples and have become a fundamental resource for search engines, question answering, recommender systems, and knowledge-enhanced large language models. However, real-world knowledge graphs remain highly incomplete, which limits their downstream reasoning ability. Existing pre-trained language model-based knowledge graph completion methods provide strong textual semantic representations, but they usually model graph structure only as shallow auxiliary features and remain weak in distinguishing structurally similar entities and topology-near negative samples. To address this limitation, this paper proposes a Dual-Level Structural Context Collaborative Framework (DSC 2 F) for knowledge graph completion. At the instance level, the framework introduces Structural Neighborhood Context (SNC) to inject local neighborhood evidence into the language model input and Relation-Aware Attention (RAA) to condition structural aggregation on the current relation. At the batch level, it constructs topology-aware training batches with biased random walk with restart, so that in-batch negatives are locally related to positive samples and impose stronger structural discrimination pressure. Experiments on WN18RR, FB15k-237, and Wikidata5M show that DSC 2 F achieves the best mean reciprocal rank and Hits@1 on all three datasets, consistently outperforming strong embedding-based and pre-trained language model-based baselines. Ablation studies and structural configuration analyses further verify that SNC, RAA, and Batch-Level Structural Context provide complementary benefits. These results demonstrate that collaborative modeling of instance-level and batch-level structural context can effectively enhance structure-aware entity representation and improve fine-grained entity prediction.
This framework performs LLM knowledge elicitation to extract factual knowledge from the model’s internal representations and transforms sentence-level representations into entity-level representations and aligns them within a unified space.
Deyu Chen, Qiyuan Li, Jinguang Gu et al.· Proceedings of the Thirty-Fi...· 0 citations
A Hypergraph-enhanced graph contrastive learning framework for Graph Out-Of-Distribution detection (termed HGOOD), which constructs two branches to hierarchically mine graph compact semantics in a comprehensive manner and introduces a cross-branch prototype contrast that aligns the captured graph patterns with their cross-branch clustering prototypes to enhance the semantic manifold of the in-distribution graph.
Xuan-Ting Fan, Chen-Yu Wang, Yue-Yue Gao et al.· Proceedings of the Thirty-Fi...· 0 citations
With the rapid development of information technology in military and complex system evaluation domains, the issue of "information overload" regarding evaluation data has become increasingly prominent. Traditional recommendation algorithms rely heavily on simple historical interactions and lack the capacity to capture semantic characteristics of evaluation texts, nor can they fully exploit high-order structural information across heterogeneous networks. To address these challenges, this paper proposes an evaluation data recommendation method based on a Deep Semantic-Aligned Heterogeneous Graph Neural Network (DeSe-HGNN). The core innovation lies in the bidirectional coupling of deep semantic alignment and dynamic topology optimization. Unlike traditional models that treat text and graphs as isolated inputs, DeSe-HGNN introduces an implicit similarity infiltration mechanism. First, a dual-tower model incorporating BERT and a Convolutional Neural Network (CNN) is developed to extract deep semantic-level features from evaluation text data, thereby constructing initial data-data affinity graphs. Second, a continuous graph learner is employed to dynamically optimize the topology of both user-user and data-data subgraphs, effectively mitigating noise and resolving data sparsity in the initial graph. Finally, a multi-layer Heterogeneous Graph Neural Network (HGNN) is architecture-designed to aggregate high-order cross-semantic representations along multiple meta-paths, combined with an attention mechanism for rating prediction. Experimental rationales and theoretical analyses demonstrate that the proposed model provides superior interpretability and enhanced recommendation accuracy in professional evaluation domains.
Quan-Dong Wang, Peng-Fei Yang, Qian Huang et al.· 2026 12th International Conf...· 0 citations
The Uncertainty-aware graph-based deep learning to rank (UGD-LTR) framework tackles two critical gaps in learning to rank, and can train independently in distributed environments, and graph feature extraction can exploit distributed graph algorithms, providing a clear roadmap for future high-performance computing (HPC) deployment.
M. Piroozmand, A. Moeini, M. Mazoochi et al.· Journal of Supercomputing· 0 citations
Knowledge graph (KG) contains rich factual and item-related information, and improves recommendation accuracy and explainability significantly. However, existing KG-based methods mainly focus on utilizing ID-based structural information, which has shown insurmountable performance bottlenecks. Actually, there is also a large amount of textual information in KG. It becomes a great challenge on how to utilize textual information for higher performance. Recently, large language models (LLMs) have been proved to have powerful semantic understanding capabilities and provide a novel approach for enhancing KG. To this end, this paper proposes an LLMs-augmented knowledge graph (LAKG) model for semantic-aware recommendation, which introduces LLMs to extract semantic features from associated texts to enhance the semantic-aware ability. First, semantic features of users and items are extracted from associated texts by the LLMs. Second, high-order structural features, which are modeled with fine-grained user intents, are further extracted from structural information by a graph attention network. Third, the semantic features and structural features are deeply fused to generate high-quality feature representations of users and items. The augmented feature information is great valuable for recommendation systems to predict the purchasing behavior of users. Experimental results show that the proposed model outperforms existing state-of-the-art baseline models, and the average accuracy improves by 4.46% on three public benchmarks.
Yuanming Zhang, Yongbiao Lou, Wenbo Zheng et al.· ACM Transactions on Knowledg...· 0 citations
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.
