Search Results (96)

This paper presents a hybrid diagnostic framework that integrates Knowledge Graphs (KGs) with Large Language Models (LLMs) to support fault diagnosis in complex, high-reliability systems such as nuclear power plants. The framework is based on the Dynamic Master Logic (DML) model, which organizes system functions, components, and dependencies into a hierarchical KG for logic-based reasoning. LLMs act as high-level facilitators by automating the extraction of DML logic from unstructured technical documentation, linking functional models with language-based reasoning, and interpreting user queries in natural language. For diagnostic queries, the LLM agent selects and invokes predefined tools that perform upward or downward propagation in the KG using DML logic, while explanatory queries retrieve and contextualize relevant KG segments to generate user-friendly interpretations. This ensures that reasoning remains transparent and grounded in the system structure. This approach reduces the manual effort needed to construct functional models and enables natural language queries to deliver diagnostic insights. In a case study on an auxiliary feedwater system used in the nuclear pressurized water reactors, the framework achieved over 90 percent accuracy in model element extraction and consistently interpreted both diagnostic and explanatory queries. The results validate the effectiveness of LLMs in automating model construction and delivering explainable AI-assisted health monitoring. Full article

The remarkable success of autoregressive Large Language Models (LLMs) is predicated on the causal attention mechanism, which enforces a static and rigid form of informational asymmetry by permitting each token to attend only to its predecessors. While effective for sequential generation, this hard-coded unidirectional constraint fails to capture the more complex, dynamic, and nonlinear dependencies inherent in sophisticated reasoning, logical inference, and discourse. In this paper, we challenge this paradigm by introducing Dynamic Asymmetric Attention (DAA), a novel mechanism that replaces the static causal mask with a learnable context-aware guidance module. DAA dynamically generates a continuous-valued attention bias for each query–key pair, effectively learning a “soft” information flow policy that guides rather than merely restricts the model’s focus. Trained end-to-end, our DAA-augmented models demonstrate significant performance gains on a suite of benchmarks, including improvements in perplexity on language modeling and notable accuracy boosts on complex reasoning tasks such as code generation (HumanEval) and mathematical problem-solving (GSM8k). Crucially, DAA provides a new lens for model interpretability. By visualizing the learned asymmetric attention patterns, it is possible to uncover the implicit information flow graphs that the model constructs during inference. These visualizations reveal how the model dynamically prioritizes evidence and forges directed logical links in chain-of-thought reasoning, making its decision-making process more transparent. Our work demonstrates that transitioning from a static hard-wired asymmetry to a learned and dynamic one not only enhances model performance but also paves the way for a new class of more capable and profoundly more explainable LLMs. Full article

Knowledge graph technology holds significant importance for efficient fault diagnosis in household appliances. However, the scarcity of public fault diagnosis data and the lack of automated knowledge extraction pose major challenges to knowledge graph construction. To address issues such as ambiguous entity boundaries, severe entity nesting, and poor entity extraction performance in fault diagnosis texts, this paper proposes a dual-strategy progressive knowledge extraction framework. First, to tackle the high complexity of fault diagnosis texts, an entity recognition model named RoBERTa-zh-BiLSTM-MUL-CRF is designed, improving the accuracy of nested entity extraction. Second, leveraging the semantic understanding capability of large language models, a progressive prompting strategy is adopted for ontology alignment and relation extraction, achieving automated knowledge extraction. Experimental results show that the proposed named entity recognition model outperforms traditional models, with improvements of 3.87%, 5.82%, and 2.05% in F1-score, recall, and precision, respectively. Additionally, the large language model demonstrates better performance in ontology alignment compared to traditional machine learning models. The constructed knowledge graph for household appliance fault diagnosis integrates structured fault diagnosis information. It effectively processes unstructured fault texts and supports visual queries and entity tracing. This framework can assist maintenance personnel in making rapid judgments, thereby improving fault diagnosis efficiency. Full article

Large language models (LLMs) often tend to hallucinate, especially in domain-specific tasks and tasks that require reasoning. Previously, we introduced SubGraph Retrieval Augmented Generation (SG-RAG) as a novel Graph RAG method for multi-hop question answering. SG-RAG leverages Cypher queries to search a given knowledge graph and retrieve the subgraph necessary to answer the question. The results from our previous work showed the higher performance of our method compared to the traditional Retrieval Augmented Generation (RAG). In this work, we further enhanced SG-RAG by proposing an additional step called Merging and Ordering Triplets (MOT). The new MOT step seeks to decrease the redundancy in the retrieved triplets by applying hierarchical merging to the retrieved subgraphs. Moreover, it provides an ordering among the triplets using the Breadth-First Search (BFS) traversal algorithm. We conducted experiments on the MetaQA benchmark, which was proposed for multi-hop question-answering in the movies domain. Our experiments showed that SG-RAG MOT provided more accurate answers than Chain-of-Thought and Graph Chain-of-Thought. We also found that merging (up to a certain point) highly overlapping subgraphs and defining an order among the triplets helped the LLM to generate more precise answers. Full article

Catalysis research is complex and interdisciplinary, involving diverse physical effects and challenging data practices. Research data often captures only selected aspects, such as specific reactants and products, limiting its utility for machine learning and the implementation of FAIR (Findable, Accessible, Interoperable, Reusable) workflows. To improve this, semantic structuring through ontologies is essential. This work extends the established ontologies by refining logical relations and integrating semantic tools such as the Web Ontology Language or the Shape Constraint Language. It incorporates application programming interfaces from chemical databases, such as the Kyoto Encyclopedia of Genes and Genomes and the National Institute of Health’s PubChem database, and builds upon established ontologies. A key innovation lies in automatically decomposing chemical substances through database entries and chemical identifier representations to identify functional groups, enabling more generalized reaction classification. Using new semantic functionality, functional groups are flexibly addressed, improving the classification of reactions such as saponification and ester cleavage with simultaneous oxidation. A graphical interface (GUI) supports user interaction with the knowledge graph, enabling ontological reasoning and querying. This approach demonstrates improved specificity of the newly established ontology over its predecessors and offers a more user-friendly interface for engaging with structured chemical knowledge. Future work will focus on expanding ontology coverage to support a wider range of reactions in catalysis research. Full article

The evaluation of large language models (LLMs) increasingly relies on other LLMs acting as automated judges. While this approach offers scalability and efficiency, it raises serious concerns regarding evaluator reliability, positional bias, and ranking stability. This paper presents a scalable framework for diagnosing positional bias and instability in LLM-based evaluation by using controlled pairwise comparisons judged by multiple independent language models. The system supports mirrored comparisons with reversed response order, prompt injection, and surface-level perturbations (e.g., paraphrasing, lexical noise), enabling fine-grained analysis of evaluator consistency and verdict robustness. Over 3600 pairwise comparisons were conducted across five instruction-tuned open-weight models using ten open-ended prompts. The top-performing model (gemma:7b-instruct) achieved a 66.5% win rate. Evaluator agreement was uniformly high, with 100% consistency across judges, yet 48.4% of verdicts reversed under mirrored response order, indicating strong positional bias. Kendall’s Tau analysis further showed that local model rankings varied substantially across prompts, suggesting that semantic context influences evaluator judgment. All evaluation traces were stored in a graph database (Neo4j), enabling structured querying and longitudinal analysis. The proposed framework provides not only a diagnostic lens for benchmarking models but also a blueprint for fairer and more interpretable LLM-based evaluation. These findings underscore the need for structure-aware, perturbation-resilient evaluation pipelines when benchmarking LLMs. The proposed framework offers a reproducible path for diagnosing evaluator bias and ranking instability in open-ended language tasks. Future work will apply this methodology to educational assessment tasks, using rubric-based scoring and graph-based traceability to evaluate student responses in technical domains. Full article

Graph analytics has grown increasingly popular as a model for data analytics across a variety of domains. This has prompted an emergence of solutions for large-scale graph analytics, many of which integrate user-facing domain-specific languages (DSLs) to support graph processing operations. These DSLs fall into two categories: query-based DSLs for graph-pattern matching and graph algorithm DSLs. While graph query DSLs are now standardized, research on DSLs for algorithmic graph processing remains fragmented and lacks a cohesive framework. To address this gap, we conduct a systematic literature review of algorithmic graph processing DSLs aimed at large-scale graph analytics. Our findings reveal the prevalence of property graphs (with 60% of surveyed DSLs explicitly adopting this model), as well as notable similarities in syntax and features. This allows us to identify a common template that can serve as the foundation for a standardized graph algorithm model, improving portability and unifying design between different DSLs and graph analytics toolkits. We additionally find that, despite achieving remarkable performance and scalability, only 20% of surveyed DSLs see real-life adoption. Incidentally, all DSLs for which user documentation is available are developed as part of academia–industry collaborations or in fully industrial contexts. Based on these results, we provide a comprehensive overview of the current research landscape, along with a roadmap of recommendations and future directions to enhance reusability and interoperability in large-scale graph analytics across industry and academia. Full article

Construction projects still face persistent barriers to adopting whole life costing (WLC), such as fragmented data, a lack of standardization, and inadequate tools. This study addresses these limitations by proposing a core ontology for WLC, developed using an ontology design science research methodology. The ontology formalizes WLC knowledge based on ISO 15686-5 and incorporates professional insights from surveys and expert focus groups. Implemented in web ontology language (OWL), it models cost categories, temporal aspects, and discounting logic in a machine-interpretable format. The ontology’s interoperability and extensibility are validated through its integration with the building topology ontology (BOT). Results show that the ontology effectively supports cost breakdown, time-based projections, and calculation of discounted values, offering a reusable structure for different project contexts. Practical validation was conducted using SQWRL queries and Python scripts for cost computation. The solution enables structured data integration and can support decision-making throughout the building life cycle. This work lays the foundation for future semantic web applications such as knowledge graphs, bridging the current technological gap and facilitating more informed and collaborative use of WLC in construction. Full article

Large language models (LLMs) have driven transformative progress in artificial intelligence, yet critical challenges persist in data management and privacy protection during model deployment and training. The approximate nearest neighbor (ANN) search, a core operation in LLMs, faces inherent trade-offs between efficiency and security when implemented through conventional locality-sensitive hashing (LSH)-based secure ANN (SANN) methods, which often compromise either query accuracy due to false positives. To address these limitations, this paper proposes a novel secure ANN scheme based on nearest neighbor graph (NNG-SANN), which is designed to ensure the security of approximate k-nearest neighbor queries for vector data commonly used in LLMs. Specifically, a secure indexing structure and subset partitioning method are proposed based on LSH and NNG. The approach utilizes neighborhood information stored in the NNG to supplement subset data, significantly reducing the impact of false positive points generated by LSH on query results, thereby effectively improving query accuracy. To ensure data privacy, we incorporate a symmetric encryption algorithm that encrypts the data subsets obtained through greedy partitioning before storing them on the server, providing robust security guarantees. Furthermore, we construct a secure index table that enables complete candidate set retrieval through a single query, ensuring our solution completes the search process in one interaction while minimizing communication costs. Comprehensive experiments conducted on two datasets of different scales demonstrate that our proposed method outperforms existing state-of-the-art algorithms in terms of both query accuracy and security, effectively meeting the precision and security requirements for nearest neighbor queries in LLMs. Full article

The rapid growth in scientific knowledge has created a critical need for advanced systems capable of managing mathematical knowledge at scale. This study presents a novel approach that integrates ontology-based knowledge representation with large language models (LLMs) to automate the extraction, organization, and reasoning of mathematical knowledge from LaTeX documents. The proposed system enhances Mathematical Knowledge Management (MKM) by enabling structured storage, semantic querying, and logical validation of mathematical statements. The key innovations include a lightweight ontology for modeling hypotheses, conclusions, and proofs, and algorithms for optimizing assumptions and generating pseudo-demonstrations. A user-friendly web interface supports visualization and interaction with the knowledge graph, facilitating tasks such as curriculum validation and intelligent tutoring. The results demonstrate high accuracy in mathematical statement extraction and ontology population, with potential scalability for handling large datasets. This work bridges the gap between symbolic knowledge and data-driven reasoning, offering a robust solution for scalable, interpretable, and precise MKM. Full article

Preserving cultural heritage can be efficiently supported by structured and semantic representation of historical artifacts. Bookbinding, a critical aspect of book history, provides valuable insights into past craftsmanship, material use, and conservation practices. However, existing bibliographic records often lack the depth needed to analyze bookbinding techniques, provenance, and preservation status. This paper presents a proof-of-concept system that explores how Large Language Models (LLMs) can support knowledge graph engineering within the context of 19th-century Greek bookbinding (1830–1900), and as a result, generate a domain-specific ontology and a knowledge graph. Our ontology encapsulates materials, binding techniques, artistic styles, and conservation history, integrating metadata standards like MARC and Dublin Core to ensure interoperability with existing library and archival systems. To validate its effectiveness, we construct a Neo4j knowledge graph, based on the generated ontology and utilize Cypher Queries—including LLM-generated queries—to extract insights about bookbinding practices and trends. This study also explores how semantic reasoning over the knowledge graph can identify historical binding patterns, assess book conservation needs, and infer relationships between bookbinding workshops. Unlike previous bibliographic ontologies, our approach provides a comprehensive, semantically rich representation of bookbinding history, methods and techniques, supporting scholars, conservators, and cultural heritage institutions. By demonstrating how LLMs can assist in ontology/KG creation and query generation, we introduce and evaluate a semi-automated pipeline as a methodological demonstration for studying historical bookbinding, contributing to digital humanities, book conservation, and cultural informatics. Finally, the proposed approach can be used in other domains, thus, being generally applicable in knowledge engineering. Full article

Retrieval-Augmented Generation (RAG) has become a central approach to enhance the factual consistency and domain specificity of large language models (LLMs) by incorporating external context at inference time. However, most existing RAG systems rely on dense vector-based similarity, which fails to capture complex semantic structures, relational dependencies, and multimodal content. In this paper, we introduce Graph Codes—a matrix-based encoding of Multimedia Feature Graphs—as an alternative retrieval paradigm. Graph Codes preserve semantic topology by explicitly encoding entities and their typed relationships from multimodal documents, enabling structure-aware and interpretable retrieval. We evaluate our system in two domains: multimodal scene understanding (200 annotated image-question pairs) and clinical question answering (150 real-world medical queries with 10,000 structured knowledge snippets). Results show that our method outperforms dense retrieval baselines in precision (+9–15%), reduces hallucination rates by over 30%, and yields higher expert-rated answer quality. Theoretically, this work demonstrates that symbolic similarity over typed semantic graphs provides a more faithful alignment mechanism than latent embeddings. Practically, it enables interpretable, modality-agnostic retrieval pipelines deployable in high-stakes domains such as medicine or law. We conclude that Graph Code-based RAG bridges the gap between structured knowledge representation and neural generation, offering a robust and explainable alternative to existing approaches. Full article

The emergence of large language models (LLMs) has introduced new possibilities for government-oriented question-answering (QA) systems. Nonetheless, limitations in retrieval accuracy and response quality assessment remain pressing challenges. This study presents ARGUS (Answer Retrieval and Governance Understanding System), a fine-tuned LLM built on a domain-adapted framework that incorporates hybrid retrieval strategies using LlamaIndex. ARGUS improves factual consistency and contextual relevance in generated answers by incorporating both graph-based entity retrieval and associated text retrieval. A comprehensive evaluation protocol combining classical metrics and RAGAS indicators is employed to assess answer quality. The experimental results show that ARGUS achieved a ROUGE-1 score of 0.68 and a semantic relevance score of 0.81. To validate the effectiveness of individual system components, a chain-of-thought mechanism inspired by human reasoning was employed to enhance interpretability. Ablation results revealed improvements in ROUGE-1 to 68.5% and S-BERT to 74.9%, over 20 percentage points higher than the baseline. Additionally, the hybrid retrieval method outperformed pure vector (0.73) and pure graph-based (0.71) strategies, achieving an F1 score of 0.75. The main contributions of this study are twofold: first, it proposes a hybrid retrieval-augmented QA framework tailored for government scenarios; second, it demonstrates the system’s reliability and practicality in addressing complex government-related queries through the integration of human-aligned metrics and traditional evaluation methods. ARGUS offers a novel paradigm for providing trustworthy, intelligent government QA systems. Full article

Multi-Hop Knowledge Graph Question Answering (multi-hop KGQA) aims to obtain answers by analyzing the semantics of natural language questions and performing multi-step reasoning across multiple entities and relations in knowledge graphs. Traditional embedding-based methods map natural language questions and knowledge graphs into vector spaces for answer matching through vector operations. While these approaches have improved model performance, they face two critical challenges: the lack of clear interpretability caused by implicit reasoning mechanisms, and the semantic gap between natural language queries and structured knowledge representations. This study proposes the DRKG (Decomposed Reasoning over Knowledge Graph), a constrained multi-hop reasoning framework based on large language models (LLMs) that introduces explicit reasoning plans as logical boundary controllers. The innovation of the DRKG lies in two key aspects: First, the DRKG generates hop-constrained reasoning plans through semantic parsing based on LLMs, explicitly defining the traversal path length and entity-retrieval logic in knowledge graphs. Second, the DRKG conducts selective retrieval during knowledge graph traversal based on these reasoning plans, ensuring faithfulness to structured knowledge. We evaluate the DRKG on four datasets, and the experimental results demonstrate that the DRKG achieves 1%–5% accuracy improvements over the best baseline models. Additional ablation studies verify the effectiveness of explicit reasoning plans in enhancing interpretability while constraining path divergence. A reliability analysis further examines the impact of different parameters combinations on the DRKG’s performance. Full article

Question answering over domain-specific knowledge graphs implies several challenges. It requires sufficient knowledge of the world and the domain to understand what is being asked, familiarity with the knowledge graph’s structure to build a correct query, and knowledge of the query language. However, mastering all of these is a time-consuming task. This work proposes a prompt-based approach that enables natural language to generate SPARQL queries. By leveraging the advanced language capabilities of large language models (LLMs), we constructed prompts that include a natural-language question, relevant contextual information from the domain-specific knowledge graph, and several examples of how the task should be executed. To evaluate our method, we applied it to an aviation knowledge graph containing accident report data. Our approach improved the results of the original work—in which the aviation knowledge graph was first introduced—by 6%, demonstrating its potential for enhancing SPARQL query generation for domain-specific knowledge graphs. Full article