NeSy-Guidance: A Neuro-Symbolic Knowledge Graph for Academic Recommendations Combining Rule-Based Reasoning and Neural Inference

Artificial Intelligence Integration in Financial Systems and Enterprise Automation: Technical Architecture and Implementation Frameworks

Forest foods play a vital role in enhancing dietary diversity, human health, and the sustainable use of forest ecosystems. However, structured and machine-readable resources that systematically describe their taxonomic and nutritional attributes remain scarce. To fill this gap, we introduce ForestFoodKG, a comprehensive resource that integrates taxonomic hierarchy and nutritional composition of 1191 forest food items. The resource consists of two components—(i) the ForestFoodKG dataset, containing standardized taxonomic and nutritional records across seven biological levels, and (ii) the ForestFoodKG Knowledge Graph (ForestFoodKG-KG), which semantically links forest food entities using named entity recognition and relation extraction. The constructed graph comprises 4492 entities and 14,130 semantic relations, providing a structured foundation for intelligent querying, nutrition analytics, and ecological informatics. All data were manually verified and made publicly available in CSV format on GitHub. ForestFoodKG serves as the first structured knowledge base for forest foods, promoting data-driven research in nutrition science, sustainable forestry, and knowledge-based decision-making.

Recent advances in large language models (LLMs) and agent-based orchestration are transforming automated knowledge graph (KG) creation as well as robust question answering in complex domains. We present a modular, multi-agent system that extracts, integrates, and reasons over diverse NoSQL movie data, powered by state-of-the-art LLMs such as GPT-4.1. Our architecture converts unstructured plots, cast/crew metadata, and numeric attributes into high-fidelity KGs - enabling both natural language and programmatic queries. To maximize reliability and flexibility, the system unifies multiple retrieval strategies - keyword search, vector similarity, knowledge graph querying, and summarization - each deployed as an autonomous pipeline. Parallel orchestration via LangGraph supports adaptive engine selection, concurrent execution, and robust answer verification with LLM ensemble “jury” scoring. Critically, the framework features comprehensive observability, allowing detailed monitoring and analysis of agent decisions, pipeline performance, and query outcomes. By treating each retrieval method and LLM as a specialized agent, our approach delivers scalable, explainable, and highly accurate results (up to 97%), significantly surpassing monolithic solutions. This agentic, observable architecture paves the way for next-generation autonomous analytics, integration, and decision support across data-rich domains.

Introduction Nature finance involves complex, multi-dimensional challenges that require analytical frameworks to assess risks, impacts, dependencies, and systemic resilience. Existing financial systems lack structured tools to map dependencies between natural capital and financial assets. To address this, we introduce NatureKG, the first ontology and instantiated knowledge graph (KG) specifically tailored to nature finance, aiming to support financial institutions in assessing environmental risks, impacts, and dependencies systematically. Methods We designed a domain ontology grounded in ENCORE, the Science-Based Targets Network (SBTN), and peer-reviewed literature. This ontology defines entities such as Actions, Drivers of Nature Loss, Value Chains, Evidence, and Sources. The ontology was instantiated into NatureKG within Neo4j, consisting of 320 nodes and 540 relationships curated by domain experts. As a proof of concept, we constructed a Text2Cypher dataset and fine-tuned three open-source large language models (Phi-3, LLaMA-3.1-8B, and Mistral-7B) to translate natural language queries into Cypher graph queries. The models were trained and evaluated under different dataset split strategies (paraphrase, cypher-level, and generalization) using metrics such as BLEU, exact match, execution accuracy, and Macro F1 scores. Results Phi-3 achieved the highest execution accuracy (0.21) and Macro F1 score (0.56), demonstrating better structural and reasoning capability under paraphrase and schema generalization splits. LLaMA-3.1-8B exhibited balanced performance, while Mistral-7B lagged across most metrics. The results indicate that smaller, fine-tuned models can generalize effectively in low-resource, domain-specific settings, validating the feasibility of LLM-assisted querying for nature finance. Discussion Despite modest initial accuracy, this feasibility study establishes a baseline for integrating domain-specific ontologies with AI systems. NatureKG offers a reusable foundation for representing environmental risks, dependencies, and interventions, with potential to enhance transparency and scalability in sustainable finance decision support. Future work should expand dataset diversity, sectoral coverage beyond the built environment, and refine model reasoning through larger, domain-aligned data catalogues.

Natural heritage digitization has evolved beyond simple 3D representation. Contemporary approaches require transparent documentation integrating biological, heritage, and digitization standards, yet existing frameworks operate in isolated domains without semantic interoperability. Current digitization frameworks fail to integrate biological standards (Darwin Core, ABCD), heritage standards (CIDOC-CRM), and digitization standards (CRMdig, PROV-O) into a unified semantic architecture, limiting transparent documentation of natural heritage data across its entire lifecycle—from physical observation through digital reconstruction to knowledge reasoning. This study proposes an integrated semantic framework comprising three components: (1) the E-DNH ontology, which adopts a triple-layer architecture (data–metadata–paradata) and a triple-module structure (nature–heritage–digital), bridging Darwin Core, CIDOC-CRM, CRMdig, and PROV-O; (2) the HR3D workflow, which establishes a standardized high-precision 3D data acquisition protocol that systematically documents paradata; and (3) the C-EDNH platform, which implements a Neo4j-based knowledge graph with semantic search capabilities, AI-driven quality assessment, and persistent identifiers (NSId/DOI). The framework was validated through digitization of 197 natural heritage specimens (68.5% avian, 24.9% insects, 5.1% mammals, 1.5% reptiles), demonstrating high geometric accuracy (RMS 0.18 ± 0.09 mm), visual fidelity (SSIM 0.92 ± 0.03), and color accuracy (ΔE00 2.1 ± 0.7). The resulting knowledge graph comprises 15,000+ nodes and 45,000+ semantic relationships, enabling cross-domain federated queries and reasoning. Unlike conventional approaches that treat digitization as mere data preservation, this framework positions digitization as an interpretive reconstruction process. By systematically documenting paradata, it establishes a foundation for knowledge discovery, reproducibility, and critical reassessment of digital natural heritage.

MGFN-WR: multi-modal knowledge graph completion with multi-layer graph fusion network based on weight regulation

Abstract Predicting tumor sensitivity to therapeutic agents is a central problem in precision oncology, yet developing models that can generalize to new, un-screened cancer types remains a significant challenge. Current precision oncology approaches benefit only a small fraction of cancer patients, partly due to the difficulty of computationally modeling the complex relationships among tumors, somatic mutations, and drug-gene pathways. To address this gap, we present PRELUDE, a heterogeneous graph neural network (GNN) framework designed to leverage these biological relationships to identify cancer cell-specific drug vulnerabilities. Our approach begins with the careful curation of a knowledge graph composed of: (1) drug-cell interactions from large-scale screening panels, (2) drug-gene relationships from curated inhibitory target databases, (3) cell-gene links derived from somatic loss-of-function mutation data, and (4) a comprehensive gene-gene interaction network We show that PRELUDE outperforms existing precision oncology baselines. Our curriculum learning approach forces the model to learn generalizable, biology-driven patterns, demonstrated by its ability to accurately predict responses for cell lines completely removed from the training graph, mimicking the challenge of predicting responses for new patients. Furthermore, our approach is interpretable, identifying effective drug target genes that interact with mutated genes in cancer cells. These findings highlight the potential of graph-based methods to enhance predictive modeling in precision oncology and support their broader adoption in data-driven cancer research. Citation Format: Luis E. Tafoya, Mikaela Dicome, Yue Hu, Macaulay Oladimeji, David Arredondo, Yanfu Zhang, Kushal Virupakshappa, Avinash Sahu. PRELUDE: A graph neural network for drug response prediction [abstract]. In: Proceedings of the AACR Special Conference in Cancer Research: Cancer Evolution: The Dynamics of Progression and Persistence; 2025 Dec 4-6; Albuquerque, NM. Philadelphia (PA): AACR; Cancer Res 2025;85(23_Suppl):Abstract nr A033.

In recent years, the development of AI-based systems has seen a drastic increase in popularity and investment. To assess and measure specific capabilities of AI-based systems, different benchmarks have been established. AI-driven approaches tend to outperform humans on most of these benchmarks, but no AI-based system was able to surpass average human performance on the Abstraction and Reasoning Corpus (ARC) benchmark. This paper presents an extensive PRISMA-guided literature review that assesses and classifies techniques and technologies utilized by solution approaches for the ARC benchmark. 538 manuscripts are screened, resulting in an inclusion of 65 publications in the final systematic literature review. As a result, a knowledge graph consisting of review protocols of manuscripts is created, that provides further insight into classification of solution approaches. Furthermore, an estimate of possible synergies and ensemble combinations between different approaches is provided by analyzing the task-level performance of solution approaches. The estimation is conducted based on the heat-maps created using the Szymkiewicz-Simpson coefficient and the Gain coefficient.

Modern scientific discovery faces challenges in integrating the rapidly expanding and diverse knowledge required for exploring novel knowledge in biology. While traditional hypothesis-driven research has proven effective, it is constrained by human cognitive limitations, knowledge complexity, and the high costs of trial-and-error experimentation. Deep learning models, particularly graph neural networks (GNNs), have accelerated scientific progress. However, the vast predictions generated make manual selection for experimental validation impractical. Attempts to leverage large language models (LLMs) for filtering predictions and generating novel hypotheses have been impeded by issues such as hallucinations and the lack of structured knowledge grounding, which undermine their reliability. To address these challenges, we propose HypoChainer, a collaborative visualization framework that integrates human expertise, LLM-driven reasoning, and knowledge graphs (KGs) to enhance scientific discovery visually. HypoChainer operates through three key stages: (1) Contextual Exploration: Domain experts employ retrieval-augmented LLMs (RAGs) and visualizations to extract insights and research focuses from vast GNN predictions, supplemented by interactive explanations for in-depth understanding; (2) Hypothesis Construction: Experts iteratively explore the KG information relevant to the predictions and hypothesis-aligned entities, gaining knowledge and insights while refining the hypothesis through suggestions from LLMs; and (3) Validation Selection: Predictions are prioritized based on the refined hypothesis chains and KG-supported evidence, identifying high-priority candidates for validation. The hypothesis chains are further optimized through visual analytics of the retrieval results. We evaluated the effectiveness of HypoChainer in hypothesis construction and scientific discovery through a case study and expert interviews.

Bacterial antimicrobial resistance is one of the most pressing global health challenges. Infections with resistant pathogens increase patient morbidity and mortality due to limited treatment options. Rapid and reliable identification of resistance is therefore crucial. However, conventional culture-based diagnostics are slow, typically requiring at least 48 hours from patient sample arrival to result. In contrast, matrix-assisted laser desorption/ionization time-of-flight (MALDI-TOF) mass spectrometry, routinely used for species identification, can provide data 24 hours earlier. Repurposing this technique for antimicrobial resistance prediction has shown promise, but limited predictive performance and a lack of statistically grounded uncertainty estimates have hindered clinical integration. To address these issues, we propose an antimicrobial resistance detection framework using a knowledge-graph-enhanced conformal predictor. Conformal prediction outputs sets of likely effective antibiotics with statistical guarantees, ensuring that resistance detection meets a predefined error rate. Our approach improves upon standard conformal prediction by integrating domain knowledge through a drug- and species-specific knowledge graph that captures interdependencies between antibiotics, such as inferable resistance patterns between broad- and narrow-spectrum agents, as well as co-resistance patterns within antibiotic classes. This predictor is layered on top of a novel classifier that surpasses state-of-the-art models and overcomes key technical limitations of earlier approaches. We evaluated our framework on two clinically relevant species, Klebsiella pneumoniae and Escherichia coli, using the DRIAMS dataset. Our results demonstrate that our conformal predictor consistently achieved the expected coverage guarantees and that the knowledge-graph enhancement significantly reduced false discovery rates compared to standard conformal approaches. By adding statistically grounded uncertainty estimates and improving predictive performance, our framework strengthens the reliability of early antimicrobial resistance predictions from MALDI-TOF data. This could support the clinical integration of such rapid diagnostics by increasing trust in their results and enabling better-informed early treatment decisions.1.

Representation learning for entity alignment (EA) aims to identify entities in different knowledge graphs (KGs) that refer to the same real-world object by comparing their embedding similarity. Although many EA models perform well on synthetic benchmark datasets, this performance does not always transfer to real-world, incomplete, and domain-specific data. A systematic comparison between synthetic benchmarks and original heterogeneous datasets is still limited. Many EA models also restrict the alignment search space to validation entities, limiting coverage of real KG content. Within this setting, our results show that embedding-based EA models continue to face generalization challenges in realistic large-scale KG search spaces. We evaluate several competitive EA models-commonly tested on benchmarks such as DBP15K-on multiple real-world heterogeneous datasets. The experiments reveal a performance decrease when moving beyond synthetic benchmarks, indicating that current models do not fully capture the characteristics of real data. We also analyze semantic similarity and profiling features of the datasets to help explain these differences. This study outlines practical limitations of embedding-based EA methods and provides insights for developing approaches that better handle the variability and complexity found in real-world KG alignment tasks.

Introduction: This paper explores the potential of Linked Data to enhance library catalog records, using the internationally recognized song "The Girl from Ipanema". Method: The famous song was used as a case study to illustrate the possibilities of the usage of a knowledge graph and metadata (authority and bibliographic data) enrichment. Results: By applying Linked Data principles, the study demonstrates how metadata enrichment can improve access and discoverability in library systems. This is achieved by including International Resource Identifiers (IRIs), the utilization of external sources such as the Virtual International Authority File (VIAF), and the Last.fm application programming interface (API). In addition, a knowledge graph was introduced as a tool for providing structured relationships with bibliographic entities within a catalog. Conclusions: The Last.fm open API is used to enrich metadata, offer track listings, and add album art to search results, thus enhancing user experience and information retrieval. The case study includes a knowledge graph in the online library catalog.

Leveraging personalized diversity level for recommendations with knowledge graph

Abstract — We present VAKIL, a domain-focused virtual assistant for Indian law created by fine-tuning Microsoft’s Phi-3 Mini (4K Instruct) model with Low-Rank Adaptation (LoRA). The training corpus comprised curated Supreme Court judgments, constitutional provisions, statutory legislation, and other authoritative legal texts. Fine-tuning was performed on a RunPod RTX A6000 instance for 18 hours across two epochs, yielding a marked decrease in training loss and improved domain alignment. To ensure responses are factually grounded, VAKIL integrates a Retrieval-Augmented Generation (RAG) pipeline: documents are chunked, tokenized, embedded, and indexed using a FAISS semantic vector store to enable high-precision retrieval. For relational and precedent-style reasoning, we construct a Neo4j AuraDB knowledge graph representing cases, statutes, and legal doctrines, and we apply a Graph Neural Network (GNN) over this graph to capture cross-document relationships and citation structure. We evaluate VAKIL through intrinsic measures (loss curves, perplexity, and reductions in hallucination) and extrinsic tasks (legal question answering, statutory interpretation, and judgment summarization). Compared to the unadapted Phi-3 base, our fine-tuned model shows improved accuracy, stronger domain specificity, and greater interpretability. The final model is released on the Hugging Face Hub and served via a serverless vLLM runtime on RunPod to support low-latency API access. A chat-based interface was implemented to support learners, researchers, and legal practitioners. VAKIL offers a reproducible workflow for building regionally focused legal assistants that balance retrieval, graph-based reasoning, and model fine-tuning. It is designed as an educational and research aid—not as a substitute for professional legal advice—and provides a transparent foundation for future enhancements. Key Words: Legal AI, RAG, LoRa, Phi-3, Knowledge Graph, GNN, Indian Law, FAISS, Legal QA, Judgment Summarization.

Boosting Knowledge Graph with Diverse-Aware Intent Inference for recommendations.

Circular Economy offers a promising approach to achieve sustainability goals by circulating resources and closing resource loops. Industrial Symbiosis adopts similar concept in industrial systems that reduces raw material consumption and waste production through collaborative waste-to-resource exchanges. While waste-to-resource databases provide valuable knowledge for IS opportunity identification, existing databases are mainly constructed manually and are restricted by their sizes and scalability. In this work, we propose an automated framework to construct a Waste-to-Resource Knowledge Graph (W2RKG) from pertinent research papers using Large Language Models, which enhances coverage, scalability, and standardisation of the resulting database. The framework comprises a Retrieving Module, an Extraction Module, and a Fusion Module, that collectively transform unstructured text into a well-organised knowledge graph. The final constructed database contains 3518 waste entities, 4471 resource entities and 33,679 waste-to-resource relationships. Extensive experiments and evaluation results demonstrate the efficacy of the proposed method and the overall high quality of the constructed database. The study, thereby, contributes an automatic framework for waste-to-resource database construction and provides a readily accessible W2RKG to support Industrial Symbiosis practitioners in identification applications.

Construction of Q&A methods based on knowledge graphs and large language models-improving the accuracy of landscape pest and disease Q&A

A novel memory interaction neural network for multi-label drug-drug interaction prediction with neighbor importance sampling.

An adaptive causal path reasoning model for marine diesel engine fault diagnosis using knowledge graph