Knowledge Graph Research Articles

Data-Driven Strain Sensor Design Based on a Knowledge Graph Framework.

Wearable flexible strain sensors require different performance depending on the application scenario. However, developing strain sensors based solely on experiments is time-consuming and often produces suboptimal results. This study utilized sensor knowledge to reduce knowledge redundancy and explore designs. A framework combining knowledge graphs and graph representational learning methods was proposed to identify targeted performance, decipher hidden information, and discover new designs. Unlike process-parameter-based machine learning methods, it used the relationship as semantic features to improve prediction precision (up to 0.81). Based on the proposed framework, a strain sensor was designed and tested, demonstrating a wide strain range (300%) and closely matching predicted performance. This predicted sensor performance outperforms similar materials. Overall, the present work is favorable to design constraints and paves the way for the long-awaited implementation of text-mining-based knowledge management for sensor systems, which will facilitate the intelligent sensor design process.

A lightweight hierarchical graph convolutional model for knowledge graph representation learning

A lightweight hierarchical graph convolutional model for knowledge graph representation learning

Revealing association rules within intricate ecosystems: A spatial co-location mining method based on Geo-Eco knowledge graph

The analysis of association rules within ecosystems is crucial for monitoring, managing, and conserving natural resources. As widely adopted approaches for this task, geospatial methods involving spatial co-location pattern mining can reveal distribution rules and inherent associations among diverse geographical elements. Rooted in Tobler’s first law of geography, these methods focus on the impact of spatial proximity. However, apart from proximity, heterogeneity of environmental attributes such as elevation, temperature and precipitation are also essential for the formation of associations. For environmental co-location (Eco-location) pattern detection, we propose a method based on the Geo-Eco Knowledge Graph (GEKG) to mine multi-impact association rules. Firstly, we introduce the Adaptive Threshold (AT) to constrain the Delaunay triangular network, dynamically regulating adjacency relationships to generate geo-eco knowledge graph’s skeleton. For comprehensive ecosystem representation, various environmental attributes are integrated as semantic information into GEKG. In the reasoning of Eco-location patterns, we innovate beyond the traditional co-location paradigm by considering both spatial proximity and semantic similarity. Under the impact of various environmental information, sub-sets of geographically proximate entities are extracted to detect Eco-location patterns. For effective management and efficient computation, we utilize the Neo4j graph database to manage large-scale GEKG and mine Eco-location patterns with its graph search function. Experiments conducted on simulated and real-world ecological datasets show that, compared to existing techniques, our GEKG-based method can detect Eco-location patterns with greater accuracy and efficiency.

Graph Artificial Intelligence in Medicine.

In clinical artificial intelligence (AI), graph representation learning, mainly through graph neural networks and graph transformer architectures, stands out for its capability to capture intricate relationships and structures within clinical datasets. With diverse data-from patient records to imaging-graph AI models process data holistically by viewing modalities and entities within them as nodes interconnected by their relationships. Graph AI facilitates model transfer across clinical tasks, enabling models to generalize across patient populations without additional parameters and with minimal to no retraining. However, the importance of human-centered design and model interpretability in clinical decision-making cannot be overstated. Since graph AI models capture information through localized neural transformations defined on relational datasets, they offer both an opportunity and a challenge in elucidating model rationale. Knowledge graphs can enhance interpretability by aligning model-driven insights with medical knowledge. Emerging graph AI models integrate diverse data modalities through pretraining, facilitate interactive feedback loops, and foster human-AI collaboration, paving the way toward clinically meaningful predictions.

Ensemble neural models for ICD code prediction using unstructured and structured healthcare data

Disease coding is the process of assigning one or more standardized diagnostic codes to clinical notes that are maintained by health practitioners (e.g. clinicians) to track patient condition. Such a coding process is often expensive and error-prone, as human medical coders primarily perform it. Automating diagnostic coding using Artificial Intelligence is seen as an essential solution in Hospital Information Management Systems and approaches built on Convolutional Neural Networks currently perform best. In this work, a neural model built on unstructured clinical text for enabling automatic diagnostic coding for given patient discharge summaries is proposed. We incorporate a structured self-attention mechanism designed to boost learning of label-specific vectors and the significant clinical text snippets associated with a certain label for this purpose. These vectors are then combined with a novel code description pipeline leveraging the descriptions provided for each standardized diagnostic code. The proposed model achieved best performance in terms of standard metrics over the MIMIC-III dataset, outperforming models based on Longformers and Knowledge graphs. Furthermore, to leverage structured clinical data to enhance the proposed model, and to enable improved diagnostic code prediction, model ensembling is considered. A neural model built on structured data by leveraging supervised machine learning algorithms such as random forest and boosting, is designed for multi-class code classification. Experimental results revealed that the proposed ensemble models show promising performance compared to traditional models that rely solely on unstructured or structured clinical data, emphasizing their suitability for real-world deployment.

Knowledge-guided data-driven design of ultra-high-performance geopolymer (UHPG)

Geopolymer has been identified as a promising family of sustainable construction materials alternative to cement-based materials. However, designing geopolymer utilizing solid wastes is a challenging task given the large variations of solid wastes in their physical and chemical properties. To overcome this challenge, this paper proposes a knowledge graph-guided data-driven approach to design geopolymer utilizing solid wastes, aimed at achieving high mechanical properties, low material cost, and low carbon emission, while largely improving material discovery efficiency. The proposed approach seamlessly integrates knowledge graph, machine learning, and multi-objective optimization, and has been utilized to design ultra-high performance geopolymer (UHPG). This approach has two main novelties: (1) The incorporation of knowledge graph imparts geopolymer domain knowledge, making the machine learning model interpretable and compliant with domain knowledge. (2) The consideration of physical and chemical properties of raw materials enables the utilization of various solid wastes. The results show that the proposed approach can reasonably predict geopolymer properties, interpret prediction results, and optimize UHPG design.

StaRS: Learning a Stable Representation Space for Continual Relation Classification.

Relation classification (RC) aims to detect the semantic relation between two annotated entities in a piece of sentence, serving as an essential task in automatic knowledge graph construction. Due to the emergence of new relations, there is a recent trend to train RC models in continual settings. To overcome the catastrophic forgetting problem in continual learning, existing research is devoted in a two-stage training paradigm, fast adaptation to novel relations, and memory replay for all historical relations. These memory-replay-based methods explore different techniques to mitigate the forgetting problem of continual RC (CRC) models during the memory replay stage. However, we find that the representation space undergoes distortion due to the incoming of fresh relations in the fast adaptation phase. To address this issue, we propose using a knowledge distillation strategy and designing a margin loss, aiming to maintain the stability of the RC model during adaptation to new relations. In addition, in the second stage, with a limited number of typical memory instances available, we introduce a self-contrastive learning objective to facilitate learning a balanced decision boundary for RC. Through training in two stages, our objective is to acquire a stable representation space to encode instances for CRC. We experimentally demonstrate the superiority of our model over competing methods in various settings, and the results suggest that our tailored designs can achieve better performance in CRC.

Knowledge graph network-driven process reasoning for laser metal additive manufacturing based on relation mining

Knowledge graph network-driven process reasoning for laser metal additive manufacturing based on relation mining

An ontology-based knowledge graph for representing interactions involving RNA molecules

The “RNA world” represents a novel frontier for the study of fundamental biological processes and human diseases and is paving the way for the development of new drugs tailored to each patient’s biomolecular characteristics. Although scientific data about coding and non-coding RNA molecules are constantly produced and available from public repositories, they are scattered across different databases and a centralized, uniform, and semantically consistent representation of the “RNA world” is still lacking. We propose RNA-KG, a knowledge graph (KG) encompassing biological knowledge about RNAs gathered from more than 60 public databases, integrating functional relationships with genes, proteins, and chemicals and ontologically grounded biomedical concepts. To develop RNA-KG, we first identified, pre-processed, and characterized each data source; next, we built a meta-graph that provides an ontological description of the KG by representing all the bio-molecular entities and medical concepts of interest in this domain, as well as the types of interactions connecting them. Finally, we leveraged an instance-based semantically abstracted knowledge model to specify the ontological alignment according to which RNA-KG was generated. RNA-KG can be downloaded in different formats and also queried by a SPARQL endpoint. A thorough topological analysis of the resulting heterogeneous graph provides further insights into the characteristics of the “RNA world”. RNA-KG can be both directly explored and visualized, and/or analyzed by applying computational methods to infer bio-medical knowledge from its heterogeneous nodes and edges. The resource can be easily updated with new experimental data, and specific views of the overall KG can be extracted according to the bio-medical problem to be studied.

Enhancing Patient Empowerment and Health Literacy: Integrating Knowledge Graphs with Language Models for Personalized Health Content Delivery.

Health literacy empowers people to access, understand and apply health information to effectively manage their own health and to be an active participant in healthcare decisions. In this paper we propose a conceptual model for cognitive factors affecting health literacy and related socioeconomic aspects. Then we develop the HEALIE Knowledge Graph to represent the model, drawing from various medical ontologies, resources, and insights from domain experts. Finally, we combine the Knowledge Graph with a Large Language Model to generate personalised medical content and showcase the results through an example.

Harnessing the Core Propagation Phenomenon Ontology to Develop a Knowledge Graph for Tracking Health-Related Phenomena.

Biomedical data analysis and visualization often demand data experts for each unique health event. There is a clear lack of automatic tools for semantic visualization of the spread of health risks through biomedical data. Illnesses such as coronavirus disease (COVID-19) and Monkeypox spread rampantly around the world before governments could make decisions based on the analysis of such data. We propose the design of a knowledge graph (KG) for spatio-temporal tracking of public health event propagation. To achieve this, we propose the specialization of the Core Propagation Phenomenon Ontology (PropaPhen) into a health-related propagation phenomenon domain ontology. Data from the UMLS and OpenStreetMaps are suggested for instantiating the proposed knowledge graph. Finally, the results of a use case on COVID-19 data from the World Health Organization are analyzed to evaluate the possibilities of our approach.

MeDaX: A Knowledge Graph on FHIR.

In Germany, the standard format for exchange of clinical care data for research is HL7 FHIR. Graph databases (GDBs), well suited for integrating complex and heterogeneous data from diverse sources, are currently gaining traction in the medical field. They provide a versatile framework for data analysis which is generally challenging for raw FHIR-formatted data. For generation of a knowledge graph (KG) for clinical research data, we tested different extract-transform-load (ETL) approaches to convert FHIR into graph format. We designed a generalised ETL process and implemented a prototypic pipeline for automated KG creation and ontological structuring. The MeDaX-KG prototype is built from synthetic patient data and currently serves internal testing purposes. The presented approach is easy to customise to expand to other data types and formats.

Machine Learning upon RDF Knowledge Graphs for Drug Safety: A Case Study on Reactome Data.

Artificial Intelligence (AI), particularly Machine Learning (ML), has gained attention for its potential in various domains. However, approaches integrating symbolic AI with ML on Knowledge Graphs have not gained significant focus yet. We argue that exploiting RDF/OWL semantics while conducting ML could provide useful insights. We present a use case using signaling pathways from the Reactome database to explore drug safety. Promising outcomes suggest the need for further investigation and collaboration with domain experts.

From Clinical Information Systems to Personalized Health Knowledge Graphs.

This paper presents a versatile solution to formally represent the contents of electronic health records. It is based on the knowledge graph paradigm, and semantic web standards RDF and OWL. It employs the established semantic standards SNOMED CT and FHIR, which warrant international interoperability. A graph-based form is not only useful to feed different target visualizations, but it can also be subject to AI-powered services such as (fuzzy) retrieval and summarization.

Evaluating the Predictive Features of Person-Centric Knowledge Graph Embeddings: Unfolding Ablation Studies.

Developing novel predictive models with complex biomedical information is challenging due to various idiosyncrasies related to heterogeneity, standardization or sparseness of the data. We previously introduced a person-centric ontology to organize information about individual patients, and a representation learning framework to extract person-centric knowledge graphs (PKGs) and to train Graph Neural Networks (GNNs). In this paper, we propose a systematic approach to examine the results of GNN models trained with both structured and unstructured information from the MIMIC-III dataset. Through ablation studies on different clinical, demographic, and social data, we show the robustness of this approach in identifying predictive features in PKGs for the task of readmission prediction.

Construction of Knowledge Graphs: Current State and Challenges

Construction of Knowledge Graphs: Current State and Challenges

SIMpat: A Synthetic Benchmark for Similarity Metrics on Patient Representations.

Similarity and clustering tasks based on data extracted from electronic health records on the patient level suffer from the curse of dimensionality and the lack of inter-patient data comparability. Indeed, for many health institutions, there are many more variables, and ways of expressing those variables to represent patients than patients sharing the same set of data. To lower redundancy and increase interoperability one strategy is to map data to semantic-driven representations through medical knowledge graphs such as SNOMED-CT. However, patient similarity metrics based on this knowledge-graph information lack quantitative evaluation and comparisons with pure data-driven methods. The reasons are twofold, firstly, it is hard to conceptually assess and formalize a gold-standard similarity between patients resulting in poor inter-annotator agreement in qualitative evaluations. Secondly, the community has been lacking a clear benchmark to compare existing metrics developed by scientific communities coming from various fields such as ontology, data science, and medical informatics. This study proposes to leverage the known challenges of evaluating patient similarities by proposing SIMpat, a synthetic benchmark to quantitatively evaluate available metrics, based on controlled cohorts, which could later be used to assess their sensibility regarding aspects such as the sparsity of variables or specificities of patient disease patterns.

A hierarchical and interlamination graph self-attention mechanism-based knowledge graph reasoning architecture

Knowledge Graph (KG) is an essential research field in graph theory, but its inherent incompleteness and sparsity influence its performance in several fields. Knowledge Graph Reasoning (KGR) aims to ameliorate those problems by mining new knowledge from subsistent knowledge. As one of the downstream tasks of KGR, link prediction is of great significance for improving the quality of KG. Recently, the Graph Neural Network (GNN)-based method became the most effective way to achieve the link prediction task. However, it still suffers from problems such as incomplete neighbor and relation-level information aggregation and unstable learning of the entity's features. To improve those issues, a Hierarchical and Interlamination Graph Self-attention Mechanism-based (HIGSM) plug-and-play architecture is proposed for KGR in this paper. It is composed of three-level layers: feature extractor, encoder, and decoder. The feature extractor makes our architecture more effective and stable for the retrieval of new features. The encoder is equipped with a two-stage encoding mechanism accompanied by two mixture-of-expert strategies, which enables our architecture to capture more practical reasoning information to improve prediction accuracy and generalization of the model. The decoder can use existing KGR models and compute the scores of triples in KG. The extensive experimental results and ablation studies on four KGs unambiguously demonstrate the state-of-the-art prediction performance of the proposed HIGSM architecture compared to current GNN-based methods.

Coherence mode: Characterizing local graph structural information for temporal knowledge graph

A Temporal Knowledge Graph (TKG) is designed for the effective modelling of the temporal relationships and dynamics of entities, events and concepts. Owing to its temporal attributes, a TKG offers greater benefits for reasoning than a static knowledge graph (KG). However, existing approaches for TKG reasoning do not consider coherent relationships between numerous facts, a term borrowed from specific parlance reflecting the interaction between objectives, such as observation and removal agents. This characteristic suggests that the model can obtain more insights from simultaneous coherent relationships. To address this problem, we develop a label-based process to construct a TKG from temporal event data in these domains. Based on the process, we build a specific TKG called Simulated Agent Interaction Knowledge Graph (SAIKG). In addition, we propose a novel TKG reasoning mechanism, termed the Coherence Mode. It is premised on an event coherence manner, enabling the prediction of unknown facts. Extensive experimental studies on different datasets demonstrate the effectiveness of the Coherence Mode integrated with typical models.

Querying large-scale knowledge graphs using Qualitative Spatial Reasoning

In this paper we consider how Qualitative Spatial Reasoning (QSR) can be used to answer queries over large-scale knowledge graphs such as YAGO and DBPedia. We describe the challenges associated with spatially querying knowledge graphs such as point based representations, sparsity of qualitative relations, and scale. We address these challenges and present a query engine, Parallel Qualitative Reasoner-Query Engine (ParQR-QE), that uses a novel distributed qualitative spatial reasoning algorithm to provide answers to GeoSPARQL queries. An experimental evaluation using a range of different query types and the YAGO knowledge graph shows the advantages of QSR techniques in comparison to purely quantitative approaches.