Triple Store Research Articles

FIPs and Practice

FIPs and Practice

Performance benchmark on semantic web repositories for spatially explicit knowledge graph applications

Knowledge graph has become a cutting-edge technology for linking and integrating heterogeneous, cross-domain datasets to address critical scientific questions. As big data has become prevalent in today's scientific analysis, semantic data repositories that can store and manage large knowledge graph data have become critical in successfully deploying spatially explicit knowledge graph applications. This paper provides a comprehensive evaluation of the popular semantic data repositories and their computational performance in managing and providing semantic support for spatial queries. There are three types of semantic data repositories: (1) triple store solutions (RDF4j, Fuseki, GraphDB, Virtuoso), (2) property graph databases (Neo4j), and (3) an Ontology-Based Data Access (OBDA) approach (Ontop). Experiments were conducted to compare each repository's efficiency (e.g., query response time) in handling geometric, topological, and spatial-semantic related queries. The results show that Virtuoso achieves the overall best performance in both non-spatial and spatial-semantic queries. The OBDA solution, Ontop, has the second-best query performance in spatial and complex queries and the best storage efficiency, requiring the least data-to-RDF conversion efforts. Other triple store solutions suffer from various issues that cause performance bottlenecks in handling spatial queries, such as inefficient memory management and lack of proper query optimization.

The OpenBiodiv Knowledge Graph Rebuilt: A semantic hub on top of the ARPHA-published content and the Biodiversity Literature Repository

OpenBiodiv is a complex ecosystem of tools and services for RDF conversion of XML narratives of biodiversity articles including Darwin Core data into Linked Open Data (LOD), running on top of a graph database. OpenBiodiv provides four main types of services: Searching named entities (e.g., taxon names, taxon concepts, treatments, specimens, occurrences, gene sequences, bibliographic information, institutions, persons) in context, within and between articles. Answering questions based on the presence of certain named entities within specific article sections (e.g., titles, abstracts, introduction or other sections, taxon treatments). Identifying article sections for further text processing (NLP) and providing contextual information, stored in MongoDB. Federating the SPARQL endpoint with other triple stores to enrich the discovered knowledge. Searching named entities (e.g., taxon names, taxon concepts, treatments, specimens, occurrences, gene sequences, bibliographic information, institutions, persons) in context, within and between articles. Answering questions based on the presence of certain named entities within specific article sections (e.g., titles, abstracts, introduction or other sections, taxon treatments). Identifying article sections for further text processing (NLP) and providing contextual information, stored in MongoDB. Federating the SPARQL endpoint with other triple stores to enrich the discovered knowledge. Conversion of such data into RDF follows a general semantic model expressed in the OpenBiodiv-O ontology, an extension of the Treatment Ontology for knowledge representation of current and legacy biodiversity publications (Senderov et al. 2018) and uses two main sources, the full-text article XML published on the ARPHA Publishing Platform and the taxon treatments extracted by Plazi’s TreatmentBank from more than 100 biodiversity journals, stored in the Biodiversity Literature Repository at Zenodo. To ensure efficiency, quality control and fast tracking of all stages of the entire process of extraction, conversion to RDF and indexing of the content has been re-built on the Apache Kafka event streaming platform (Fig. 1). In this new format, OpenBiodiv provides not only a GraphDB SPARQL query endpoint but also indexes the named entities through Elasticsearch and additional provision of data to end users through a RESTful API and a number of user applications. OpenBiodiv is designed for a wide range of users who are interested in a deep-level bibliographic exploration, an ontology-linked search of various data elements (e.g., specimens, sequences, taxon concepts, persons), or co-existence of named entities (e.g., taxon names with a possible biotic relationships between them, or taxon names and potential habitats of occupation) in pre-defined sections of the articles. The SPARQL endpoint allows complex queries of various kinds (Dimitrova et al. 2021).

Bootstrapping a Biodiversity Knowledge Graph

The "biodiversity knowledge graph" is a nice metaphor for connecting biodiversity data sources, but can we actually build it? Do we have sufficient linked data available? Given that a knowledge graph is an aggregation of data from multiple sources, how do we give those sources credit for that data, and how do we handle changes to that data? Given that the classic interface to a knowledge graph is an intimidatingly empty SPARQL query box, how do we make the knowledge within a graph more accessible? This talk discusses an attempt to build a knowledge graph with an eye on how to maintain the graph in the future. It adopts a model similar to Global Biodiversity Information Facility (GBIF) and CheckListBank where individual data providers make datasets available as independently citable units with Digital Object Identifiers (DOIs). Each dataset comprises linked data in the form of N-triples. To create a knowledge graph we simply download one or more such datasets and add them to a triple store. Each data source is assigned to its own named graph, such that we have provenance for each dataset, and we can update any dataset independently. Furthermore, anyone can build their own knowledge graph by mixing and matching the set of data (people, publications, taxa, etc.) most appropriate to their interests. To bootstrap this approach, exemplar datasets are created based on data harvested from ORCID, Zenodo, and taxonomic name databases. Each demonstration dataset could be replaced in the future by data published directly by those providers. In some cases there are sufficient shared identifiers (such as DOIs and ORCIDs) to form a graph, but taxonomic data typically forms isolated islands. To help the knowledge graph coalesce we need "glue" in the form of datasets that link pairs of different identifiers, such as Life Science Identifiers (LSIDs) for names to DOIs for publications. With the addition of those cross links we can start to generate bibliographies for taxa, discover communities of taxonomic expertise, and more. This model of building a knowledge graph also opens opportunities for smaller, focussed datasets to be added to the graph using the same approach (as set of N-triples archived in an online repository). In order to be useful, a knowledge graph needs to be easy to query and visualise. Simply providing a SPARQL endpoint is unlikely to be enough. As part of this project, I developed a GraphQL interface to the knowledge graph to provide a set of standard queries that can support a simple web interface to the graph. This provides a way to explore the graph as it is being developed, which in turn can highlight gaps in connectivity and coverage that need to be addressed.

Jumping Evaluation of Nested Regular Path Queries

Nested regular path queries are used for querying graph databases and RDF triple stores. We propose a new algorithm for evaluating nested regular path queries on a graph from a set of start nodes in combined linear time. We show that this complexity upper bound can be reduced by making it dependent on the size of the query's top-down needed subgraph, a notion that we introduce. For many queries in practice, the top-down needed subgraph is way smaller than the whole graph. Our algorithm is based on a novel compilation schema from nested regular path queries to monadic datalog queries. Its complexity upper bound follows from known properties of top-down datalog evaluation. As an application, we show that our algorithm permits to reformulate in simple terms a variant of a very efficient automata-based algorithm proposed by Maneth and Nguyen that evaluates navigational path queries in datatrees based on indexes and jumping. Moreover, it overcomes some limitations of Maneth and Nguyen's: it is not bound to trees and applies to graphs; it is not limited to forward navigational XPath but can treat any nested regular path query and it can be implemented efficiently without any dedicated techniques, by using any efficient datalog evaluator such as LogicBlox.

Design of Knowledge Graph Retrieval System for Legal and Regulatory Framework of Multilevel Latent Semantic Indexing.

Latent semantic analysis (LSA) is a natural language statistical model, which is considered as a method to acquire, generalize, and represent knowledge. Compared with other retrieval models based on concept dictionaries or concept networks, the retrieval model based on LSA has the advantages of strong computability and less human participation. LSA establishes a latent semantic space through truncated singular value decomposition. Words and documents in the latent semantic space are projected onto the dimension representing the latent concept, and then the semantic relationship between words can be extracted to present the semantic structure in natural language. This paper designs the system architecture of the public prosecutorial knowledge graph. Combining the graph data storage technology and the characteristics of the public domain ontology, a knowledge graph storage method is designed. By building a prototype system, the functions of knowledge management, knowledge query, and knowledge push are realized. A named entity recognition method based on bidirectional long-short-term memory (bi-LSTM) combined with conditional random field (CRF) is proposed. Bi-LSTM-CRF performs named entity recognition based on character-level features. CRF can use the transition matrix to further obtain the relationship between each position label, so that bi-LSTM-CRF not only retains the context information but also considers the influence between the current position and the previous position. The experimental results show that the LSTM-entity-context method proposed in this paper improves the representation ability of text semantics compared with other algorithms. However, this method only introduces relevant entity information to supplement the semantic representation of the text. The order in the case is often ignored, especially when it comes to the time series of the case characteristics, and the “order problem” may eventually affect the final prediction result. The knowledge graph of legal documents of theft cases based on ontology can be updated and maintained in real time. The knowledge graph can conceptualize, share, and perpetuate knowledge related to procuratorial organs and can also reasonably utilize and mine many useful experiences and knowledge to assist in decision-making.

Wukong+G: Fast and Concurrent RDF Query Processing Using RDMA-Assisted GPU Graph Exploration

RDF graph has been increasingly used to store and represent information shared over the Web, including social graphs and knowledge bases. With the increasing scale of RDF graphs and the concurrency level of SPARQL queries, current RDF systems are confronted with inefficient concurrent query processing on massive data parallelism. The situation becomes more severe in the face of data-intensive queries (aka heavy query), which usually lead to suboptimal response time (latency) as well as throughput collapse. In this article, we present Wukong+G, the first graph-based distributed RDF query processing system that efficiently exploits the hybrid parallelism of CPU and GPU. Wukong+G is made fast and concurrent with four key designs. First, Wukong+G tames massive random memory accesses in graph exploration by efficiently mapping data between CPU and GPU for latency hiding, including a set of techniques like query-aware prefetching, pattern-aware pipelining and fine-grained swapping. Second, Wukong+G scales up by introducing a GPU-friendly RDF store to support RDF graphs exceeding GPU memory size, by using techniques like predicate-based grouping, pairwise caching and look-ahead replacing to narrow the gap between host and device memory scale. Third, Wukong+G scales out through a communication layer that decouples the transferring process for query metadata and intermediate results, and further leverages both native and GPUDirect RDMA to enable efficient communication on a CPU/GPU cluster. Finally, Wukong+G simultaneously runs multiple queries on a single GPU to improve overall throughput and fully exploits hardware heterogeneity (CPU/GPU) by scheduling a single query on CPU and GPU adaptively. We have implemented Wukong+G by extending a state-of-the-art distributed RDF store (i.e., Wukong) with distributed GPU support. Evaluation on a heterogeneous CPU/GPU cluster with RDMA-capable network shows that Wukong+G outperforms Wukong by up to 9.0× (from 2.3×) and scales well on 10 GPU cards for heavy queries. Wukong+G can also improve both latency and throughput by more than one order of magnitude when facing hybrid workloads.

Toward Mapping an NGSI-LD Context Model on RDF Graph Approaches: A Comparison Study.

A considerable number of Internet of Things deployments are isolated from specific solutions, from devices to data platforms. Standardized data models were proposed to overcome the interoperability gap between these deployments. Next generation service interfaces-linked data (NGSI-LD) is one of the proposed platforms that exploits linked data and proposes an information model and an application programming interface (API) for easy use and standard management of context information. The NGSI-LD information model is based on JSON for Linked Data (JSON-LD) as a serialization format for context information. This efficiently exploits the potential of semantics and linked open data. However, the NGSI-LD graph API and query language are still theoretically defined and limited to some preliminary works. Consequently, current NGSI-LD implementations are mainly based on traditional databases, where the JSON-LD serialization is supported but not exploited owing to the difficulties in defining and implementing new NGSI-LD based Graph APIs. One of the basic solutions is the use of an RDF store for NGSI-LD payloads because these types of databases are well defined and maintained and will not need any added effort for JSON-LD based payloads. However, the main complication at this level is the use of reification to annotate relationships. This study focused on both aspects of exploiting the semantics of NGSI-LD by proposing standardized mapping mechanisms to RDF graphs without reifying JSON-LD payloads and with respect to the NGSI-LD context model and API. Our main proposals highlight that exploiting the RDF store for processing NGSI-LD data semantically is feasible and uncomplicated. We illustrated the proposed mapping approaches with real use-case examples and a possible exploitation of semantic approaches.

An Ontology and Data Converter from RDF to the i2b2 Data Model.

In a national effort aiming at cross-hospitals data interoperability, the Swiss Personalized Health Network elected RDF as preferred data and meta-data representation format. Yet, most clinical research software solutions are not designed to interact with RDF databases. We present a modular Python toolkit allowing easy conversion from RDF graphs to i2b2, adaptable to other common data models (CDM) with reasonable efforts. The tool was designed with feedback from clinicians in both oncology and laboratory research.

Context mining and graph queries on giant biomedical knowledge graphs

Contextual information is widely considered for NLP and knowledge discovery in life sciences since it highly influences the exact meaning of natural language. The scientific challenge is not only to extract such context data, but also to store this data for further query and discovery approaches. Classical approaches use RDF triple stores, which have serious limitations. Here, we propose a multiple step knowledge graph approach using labeled property graphs based on polyglot persistence systems to utilize context data for context mining, graph queries, knowledge discovery and extraction. We introduce the graph-theoretic foundation for a general context concept within semantic networks and show a proof of concept based on biomedical literature and text mining. Our test system contains a knowledge graph derived from the entirety of PubMed and SCAIView data and is enriched with text mining data and domain-specific language data using Biological Expression Language. Here, context is a more general concept than annotations. This dense graph has more than 71M nodes and 850M relationships. We discuss the impact of this novel approach with 27 real-world use cases represented by graph queries. Storing and querying a giant knowledge graph as a labeled property graph is still a technological challenge. Here, we demonstrate how our data model is able to support the understanding and interpretation of biomedical data. We present several real-world use cases that utilize our massive, generated knowledge graph derived from PubMed data and enriched with additional contextual data. Finally, we show a working example in context of biologically relevant information using SCAIView.

PreKar: A learned performance predictor for knowledge graph stores

PreKar: A learned performance predictor for knowledge graph stores

Creating RESTful APIs over SPARQL endpoints using RAMOSE

Semantic Web technologies are widely used for storing RDF data and making them available on the Web through SPARQL endpoints, queryable using the SPARQL query language. While the use of SPARQL endpoints is strongly supported by Semantic Web experts, it hinders broader use of RDF data by common Web users, engineers and developers unfamiliar with Semantic Web technologies, who normally rely on Web RESTful APIs for querying Web-available data and creating applications over them. To solve this problem, we have developed RAMOSE, a generic tool developed in Python to create REST APIs over SPARQL endpoints. Through the creation of source-specific textual configuration files, RAMOSE enables the querying of SPARQL endpoints via simple Web RESTful API calls that return either JSON or CSV-formatted data, thus hiding all the intrinsic complexities of SPARQL and RDF from common Web users. We provide evidence that the use of RAMOSE to provide REST API access to RDF data within OpenCitations triplestores is beneficial in terms of the number of queries made by external users of such RDF data using the RAMOSE API, compared with the direct access via the SPARQL endpoint. Our findings show the importance for suppliers of RDF data of having an alternative API access service, which enables its use by those with no (or little) experience in Semantic Web technologies and the SPARQL query language. RAMOSE can be used both to query any SPARQL endpoint and to query any other Web API, and thus it represents an easy generic technical solution for service providers who wish to create an API service to access Linked Data stored as RDF in a triplestore.

Querying graph databases using context-free grammars

Path queries are used to specify paths inside a data graph to match a given pattern. Query languages such as SPARQL usually include support for regular path patterns defined by means of regular expressions. Context-free path queries define a path whose language can be defined by a context-free grammar. This kind of query is interesting in practice in domains such as genetics, data science, and source code analysis. In this paper, we present an algorithm for context-free path query processing. Our algorithm works by looking for localized paths, allowing us to process subgraphs, in contrast to other approaches that have to process the whole graph. It also takes any context-free grammar as input, avoiding the use of normal forms that can be problematic. The grammar normalization process may introduce a large number of non-terminal symbols and production rules, what, in general, reflects on more runtime and memory consumption by evaluation algorithms. We prove the correctness of our approach and show its runtime and memory complexity. We show the viability of our approach by means of prototypes implemented in Go and Python. We run experiments proposed in recent works, which include both synthetic and real RDF databases, and introduce a more realistic scenario inspired in Biology. Our algorithm shows performance gains when compared to other algorithms implemented using single-thread programs.

A design space for RDF data representations

RDF triplestores’ ability to store and query knowledge bases augmented with semantic annotations has attracted the attention of both research and industry. A multitude of systems offer varying data representation and indexing schemes. However, as recently shown for designing data structures, many design choices are biased by outdated considerations and may not result in the most efficient data representation for a given query workload. To overcome this limitation, we identify a novel three-dimensional design space. Within this design space, we map the trade-offs between different RDF data representations employed as part of an RDF triplestore and identify unexplored solutions. We complement the review with an empirical evaluation of ten standard SPARQL benchmarks to examine the prevalence of these access patterns in synthetic and real query workloads. We find some access patterns, to be both prevalent in the workloads and under-supported by existing triplestores. This shows the capabilities of our model to be used by RDF store designers to reason about different design choices and allow a (possibly artificially intelligent) designer to evaluate the fit between a given system design and a query workload.

Interoperability and Integration: An Updated Approach to Linked Data Publication at the Dutch Land Registry

Kadaster, the Dutch National Land Registry and Mapping Agency, has been actively publishing their base registries as linked (open) spatial data for several years. To date, a number of these base registers as well as a number of external datasets have been successfully published as linked data and are publicly available. Increasing demand for linked data products and the availability of new linked data technologies have highlighted the need for a new, innovative approach to linked data publication within the organisation in the interest of reducing the time and costs associated with said publication. The new approach to linked data publication is novel in both its approach to dataset modelling, transformation, and publication architecture. In modelling whole datasets, a clear distinction is made between the Information Model and the Knowledge Model to capture both the organisation-specific requirements and to support external, community standards in the publication process. The publication architecture consists of several steps where instance data are loaded from their source as GML and transformed using an Enhancer and published in the triple store. Both the modelling and publication architecture form part of Kadaster’s larger vision for the development of the Kadaster Knowledge Graph through the integration of the various linked datasets.

Practical Near-Data-Processing Architecture for Large-Scale Distributed Graph Neural Network

Graph Neural Networks have drawn tremendous attention in the past few years due to their convincing performance and high interpretability in various graph-based tasks like link prediction and node classification. With the ever-growing graph size in the real world, especially for industrial graphs at a billion-level, the storage of graphs can easily consume Terabytes so that the process of GNNs has to be processed in a distributed manner. As a result, the execution could be inefficient due to the expensive cross-node communication and irregular memory access. Various GNN accelerators have been proposed for efficient GNN processing. They, however, mainly focused on small and medium-size graphs, which is not applicable to large-scale distributed graphs. In this paper, we present a practical Near-Data-Processing architecture based on a memory-pool system for large-scale distributed GNNs. We propose a customized memory fabric interface to construct the memory pool for low-latency and high throughput cross-node communication, which can provide flexible memory allocation and strong scalability. A practical Near-Data-Processing design is proposed for efficient work offloading and bandwidth utilization improvement. Moreover, we also introduce a partition and scheduling scheme to further improve performance and achieve workload balance. Comprehensive evaluations demonstrate that the proposed architecture can achieve up to <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$27\times $ </tex-math></inline-formula> and <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$8\times $ </tex-math></inline-formula> higher training speed compared to two state-of-the-art distributed GNN frameworks: Deep Graph Library and <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$P^{3}$ </tex-math></inline-formula> , respectively.

Incremental Knowledge Extraction From IoT-Based System for Anomaly Detection in Vegetation Crops

Precision agriculture systems collect spectral images from satellites, from which vegetation indices (VIs) can be assessed to monitor vegetation and soil condition. It requires a near-daily data acquisition to perform robust crop monitoring and data analysis. Satellites provide a periodic data acquisition that need a further data integration using multiple satellite sources along with camera-equipped drones to achieve an accurate data collection on a selected area. Moreover, VIs are not enough for a proper vegetation evaluation of the monitored areas due to differences among cultivars, the phenological season in which the vegetation is evaluated, the latitude of the areas, etc. This article introduces a system model to detect anomalies regarding the vegetation and soil conditions according to the area phenology and the historical vegetation trends. The system collects spectral images of the regions of interest (ROIs) from satellites and drones, harmonized to calculate VIs and feeds a dataset of near-daily high-resolution integrated images. The harmonic analysis allows phenological data extraction about the ROIs, hence the territorial observation model (TOM) has been extended to represent phenological stages and build knowledge on the ROIs and their phenology that is stored on a triple store. The system selects the VI values, calculated during the learned growing seasons of the ROIs, and classifies them to detect vegetation anomalies affecting those ROIs. The collected knowledge can be used by end-users (e.g., agronomists, experts, etc.) to analyze the anomalies correlated to historical results and vegetation trends.

CRAFTS: Configurable REST APIs for Triple Stores

Massive amounts of Linked Open Data are readily available to anyone who wants to use them. Unfortunately, Semantic Web technologies such as SPARQL and RDF remain unfamiliar to the majority of web developers, more used to REST APIs. This paper addresses the challenge of accessing Linked Open Data through REST APIs. Configurable REST APIs For Triple Stores (CRAFTS) is the tool devised for this purpose. CRAFTS allows knowledge engineers to configure REST APIs over multiple triple stores. CRAFTS automatically handles the translation of API calls into SPARQL queries, delivering results in JSON format. Web developers can then use a CRAFTS API to read and write Linked Open Data. The API of CRAFTS is uniform, domain-independent, and described with the OpenAPI specification. A reference implementation of CRAFTS is published in a GitHub repository, and a live test site is readily available since February 2021. CRAFTS is currently employed in eight different applications, with more than 2.4K users, and more than 260K API calls.

Cloud-Based Framework for Spatio-Temporal Trajectory Data Segmentation and Query

Trajectory segmentation is a technique of dividing sequential trajectory into segments. These segments are building blocks to various applications. Hence a system framework is essential to support trajectory segment indexing, storage, and query. When the size of segments is beyond the computing capacity of a single processing node, a distributed solution is proposed. In this article, we develop a distributed trajectory segmentation framework that includes a greedy-split segmentation method. This framework consists of distributed in-memory processing and a cluster of graph storage respectively. For fast trajectory queries, we design a distributed spatial R-tree index of trajectory segments. Using the indexes, we build scalable query operations from both in-memory processing and access to graph storage. Based on this framework, we define two metrics to measure trajectory similarity and chance of collision. These two metrics are further applied to identify moving groups of trajectories. We quantitatively evaluate the effects of data partition, parallelism, and data size on the system. We identify the bottleneck factors at the data partition stage and validate two mitigation techniques to data skew. The evaluation demonstrates our distributed segmentation method and the system framework scale as the growth of the workload and the size of the parallel cluster.

A survey of RDF stores &amp; SPARQL engines for querying knowledge graphs

RDF has seen increased adoption in recent years, prompting the standardization of the SPARQL query language for RDF, and the development of local and distributed engines for processing SPARQL queries. This survey paper provides a comprehensive review of techniques and systems for querying RDF knowledge graphs. While other reviews on this topic tend to focus on the distributed setting, the main focus of the work is on providing a comprehensive survey of state-of-the-art storage, indexing and query processing techniques for efficiently evaluating SPARQL queries in a local setting (on one machine). To keep the survey self-contained, we also provide a short discussion on graph partitioning techniques used in the distributed setting. We conclude by discussing contemporary research challenges for further improving SPARQL query engines. An extended version also provides a survey of over one hundred SPARQL query engines and the techniques they use, along with twelve benchmarks and their features.