Triple Pattern Fragments: A Low-Cost Knowledge Graph Interface for the Web

Abstract

Billions of Linked Data triples exist in thousands of RDF knowledge graphs on the Web, but few of those graphs can be queried live from Web applications. Only a limited number of knowledge graphs are available in a queryable interface, and existing interfaces can be expensive to host at high availability. To mitigate this shortage of live queryable Linked Data, we designed a low-cost Triple Pattern Fragments interface for servers, and a client-side algorithm that evaluates SPARQL queries against this interface. This article describes the Linked Data Fragments framework to analyze Web interfaces to Linked Data and uses this framework as a basis to define Triple Pattern Fragments. We describe client-side querying for single knowledge graphs and federations thereof. Our evaluation verifies that this technique reduces server load and increases caching effectiveness, which leads to lower costs to maintain high server availability. These benefits come at the expense of increased bandwidth and slower, but more stable query execution times. These results substantiate the claim that lightweight interfaces can lower the cost for knowledge publishers compared to more expressive endpoints, while enabling applications to query the publishers' data with the necessary reliability.

Distributed Subweb Specifications for Traversing the Web

“…Federated querying over expressive interfaces such as sparql endpoints only addresses part of the problem: empirical evidence suggests that, counterintuitively, less expressive interfaces can lead to faster processing times for several queries (Verborgh et al 2016), while being less expensive to host. A document-based interface is about the simplest interface imaginable, and is thereby partly responsible for the Web's scalability.…”

“…As with Big Data, query processing on the Semantic Web has traditionally focused on single databases. The sparql query language allows querying such a single rdf store through the sparql protocol, which places significantly more constraints on the server than a document-based interface (Verborgh et al 2016). While federated query processing enables incorporating data from multiple sparql endpoints, federated queries have very limited link traversal capabilities and sparql endpoints easily experience performance degradation (Buil-Aranda et al 2013).…”

Using a Personal Health Library–Enabled mHealth Recommender System for Self-Management of Diabetes Among Underserved Populations: Use Case for Knowledge Graphs and Linked Data

“…Executing federated Linked Data queries on the Web requires accessing multiple data sources, which involves the discovery of data sources and determining relevant ones. Researchers have proposed discovery approaches for Linked Data interfaces based on hypermedia links and controls and applied them to federated query execution with Triple Pattern Fragments [34]. SPARQL endpoints are expensive for the server and not always available for all data sets.…”

“…Downloadable dumps are expensive for clients and do not allow live querying on the web. The Linked Data Fragments framework enables client-side SPARQL querying of live Linked Data on the Web and federated querying through a triple-pattern interface, providing a much faster, less expensive solution [34].…”

BimSPARQL: Domain-specific functional SPARQL extensions for querying RDF building data

“…It has been applied on some geospatial cases and a plugin to include functions for spatial computation is provided based on the PostGIS API. Recently, an approach is presented in [12] to define functions by extending Triple Pattern Fragments (TPF) [53] on the client side, hence extended functions are compatible with any SPARQL server. As showed in [12], it however might have issues regarding performance and data traffics since additional functions are computed in web browsers and raw data needs to be retrieved to the client side.…”