Small language models (SLMs) have shown promises for relation extraction (RE) when extracting RDF triples guided by SHACL shapes focused on common Datatype Properties. This paper investigates how SLMs handle both Datatype and Object Properties for a complete RDF graph extraction. We show that the key bottleneck is related to long-tail distribution of rare properties. To solve this issue, we evaluate several strategies: stratified sampling, weighted loss, dataset scaling, and template-based synthetic data augmentation. We show that the best strategy to perform equally well over unbalanced target properties is to build a training set where the number of occurrences of each property exceeds a given threshold. To enable reproducibility, we publicly released our datasets, experimental results and code. Our findings offer practical guidance for training shape-aware SLMs and highlight promising directions for future work in semantic RE.

The SPARQL query language is the standard method to access knowledge graphs (KGs). However, formulating SPARQL queries is a significant challenge for non-expert users, and remains time-consuming for the experienced ones. Best practices recommend to document KGs with competency questions and example queries to contextualise the knowledge they contain and illustrate their potential applications. In practice, however, this is either not the case or the examples are provided in limited numbers. Large Language Models (LLMs) are being used in conversational agents and are proving to be an attractive solution with a wide range of applications, from simple question-answering about common knowledge to generating code in a targeted programming language. However, training and testing these models to produce high quality SPARQL queries from natural language questions requires substantial datasets of question-query pairs. In this paper, we present Q2Forge that addresses the challenge of generating new competency questions for a KG and corresponding SPARQL queries. It iteratively validates those queries with human feedback and LLM as a judge. Q2Forge is open source, generic, extensible and modular, meaning that the different modules of the application (CQ generation, query generation and query refinement) can be used separately, as an integrated pipeline, or replaced by alternative services. The result is a complete pipeline from competency question formulation to query evaluation, supporting the creation of reference question-query sets for any target KG.

Abstract The Relational to RDF Mapping Language (R2RML) became a W3C Recommendation a decade ago. Despite its wide adoption, its potential applicability beyond relational databases was swiftly explored. As a result, several extensions and new mapping languages were proposed to tackle the limitations that surfaced as R2RML was applied in real-world use cases. Over the years, one of these languages, the RDF Mapping Language (RML), has gathered a large community of contributors, users, and compliant tools. So far, there has been no well-defined set of features for the mapping language, nor was there a consensus-marking ontology. Consequently, it has become challenging for non-experts to fully comprehend and utilize the full range of the language’s capabilities. After three years of work, the W3C Community Group on Knowledge Graph Construction proposes a new specification for RML. This paper presents the new modular RML ontology and the accompanying SHACL shapes that complement the specification. We discuss the motivations and challenges that emerged when extending R2RML, the methodology we followed to design the new ontology while ensuring its backward compatibility with R2RML, and the novel features which increase its expressiveness. The new ontology consolidates the potential of RML, empowers practitioners to define mapping rules for constructing RDF graphs that were previously unattainable, and allows developers to implement systems in adherence with [R2]RML. Resource type: Ontology/License: CC BY 4.0 International DOI: 10.5281/zenodo.7918478 /URL: http://w3id.org/rml/portal/

Abstract Faced with the ever-increasing number of scientific publications, researchers struggle to keep up, find and make sense of articles relevant to their own research. Scientific open archives play a central role in helping deal with this deluge, yet keyword-based search services often fail to grasp the richness of the semantic associations between articles. In this paper, we present the methods, tools and services implemented in the ISSA project to tackle these issues. The project aims to (1) provide a generic, reusable and extensible pipeline for the analysis and processing of articles of an open scientific archive, (2) translate the result into a semantic index stored and represented as an RDF knowledge graph; (3) develop innovative search and visualization services that leverage this index to allow researchers, decision makers or scientific information professionals to explore thematic association rules, networks of co-publications, articles with co-occurring topics, etc. To demonstrate the effectiveness of the solution, we also report on its deployment and user-driven customization for the needs of an institutional open archive of 110,000+ resources. Fully in line with the open science and FAIR dynamics, the presented work is available under an open license with all the accompanying documents necessary to facilitate its reuse. The knowledge graph produced on our use-case is compliant with common linked open data best practices.