Searcharxiv⌕ Search

arXiv subjects

Robert Stevens

Publications and source records attributed to Robert Stevens.

8 recordsLinked to original sources

Cross-cultural Adaptation of Recipes for Low-Resource Languages: A Benchmark and Multilingual Culinary Lexicon for Hausa, Yorùbá, and Igbo

Recent multilingual large language models (LLMs) have improved machine translation, yet culturally adaptive translation for low-resource languages (LRLs) remains underexplored. This is particularly challenging for recipes, where ingredients, cooking tools, measurements, and procedures may require adaptation to the target cultural context. In this work, we introduce ReciLoReL, a benchmark of 600 expert-translated and culturally adapted recipes across Yorùbá, Hausa, and Igbo (three of the most prominent languages in Nigeria), and ReciLex, a multilingual culinary lexicon containing approximately 15,000 translated entities across ten recipe entity categories, including ingredients, tools, chef actions, state of tools, and durations. We evaluate state-of-the-art multilingual LLMs for English-to-LRL recipe translation under zero-shot and few-shot settings and investigate whether augmenting LLMs with ReciLex improves culturally adaptive recipe translation without requiring large target-language recipe corpora for training or fine-tuning. Beyond conventional reference-based translation evaluation using BLEU, chrF, and TER, we introduce five criteria for evaluating cultural adaptation, adapted from established cross-cultural adaptation methods. Across the selected LLMs and languages, ReciLex improves average BLEU from 13.57 to 14.51 and chrF from 37.89 to 38.74, while reducing TER from 77.14 to 75.12. Human evaluation further shows improvements ranging from 0.11 to 0.36 points across the five cultural adaptation evaluation criteria on a five-point scale. Overall, our work provides new resources, evaluation criteria, and empirical evaluation for studying culturally adaptive recipe translation in low-resource settings.

cs.CL↗

SCODEL: Shared Canonical Ontology Definitions for Food Entity Linking and Ontology-Grounded Recipe Knowledge Graphs

Identifying recipe entities and linking them to ontology concepts is fundamental for semantic recipe understanding. However, food mentions in recipes are often ambiguous, context-dependent, and lexically diverse, making ontology entity linking (EL) challenging. We propose SCODEL, a streamlined, model-agnostic approach based on a central idea: contextual food mentions and ontology concepts that refer to the same entity should learn the same canonical ontology-oriented definition. SCODEL uses these shared definitions as a semantic normalisation layer, providing complementary semantic information to existing EL models without requiring a new linking architecture. We further introduce ReciLinkD, the first benchmark for contextual food ontology entity linking in recipes. Experiments show that SCODEL consistently improves three modular EL models (ReFinED, GSM-EL, and BLINK): SCODEL x BLINK achieves the highest Hit@1 on seen concepts (88.33%), while SCODEL x GSM-EL achieves the highest Hit@1 on unseen concepts (83.15%), outperforming the strongest large language model (LLM) based EL baseline in each setting. Experiments on a biomedical ontology benchmark further demonstrate that SCODEL generalises beyond the food domain. Finally, we use SCODEL to ground food mentions at scale and construct an ontology-grounded recipe knowledge graph, demonstrating its utility for ontology-aware recipe retrieval and corpus-level food analysis.

cs.AI↗

Measuring Expert Performance at Manually Classifying Domain Entities under Upper Ontology Classes

Classifying entities in domain ontologies under upper ontology classes is a recommended task in ontology engineering to facilitate semantic interoperability and modelling consistency. Integrating upper ontologies this way is difficult and, despite emerging automated methods, remains a largely manual task. Little is known about how well experts perform at upper ontology integration. To develop methodological and tool support, we first need to understand how well experts do this task. We designed a study to measure the performance of human experts at manually classifying classes in a general knowledge domain ontology with entities in the Basic Formal Ontology (BFO), an upper ontology used widely in the biomedical domain. We conclude that manually classifying domain entities under upper ontology classes is indeed very difficult to do correctly. Given the importance of the task and the high degree of inconsistent classifications we encountered, we further conclude that it is necessary to improve the methodological framework surrounding the manual integration of domain and upper ontologies.

cs.DL↗

Facets, Tiers and Gems: Ontology Patterns for Hypernormalisation

There are many methodologies and techniques for easing the task of ontology building. Here we describe the intersection of two of these: ontology normalisation and fully programmatic ontology development. The first of these describes a standardized organisation for an ontology, with singly inherited self-standing entities, and a number of small taxonomies of refining entities. The former are described and defined in terms of the latter and used to manage the polyhierarchy of the self-standing entities. Fully programmatic development is a technique where an ontology is developed using a domain-specific language within a programming language, meaning that as well defining ontological entities, it is possible to add arbitrary patterns or new syntax within the same environment. We describe how new patterns can be used to enable a new style of ontology development that we call hypernormalisation.

cs.AI↗

OntoVerbal: a Generic Tool and Practical Application to SNOMED CT

Ontology development is a non-trivial task requiring expertise in the chosen ontological language. We propose a method for making the content of ontologies more transparent by presenting, through the use of natural language generation, naturalistic descriptions of ontology classes as textual paragraphs. The method has been implemented in a proof-of- concept system, OntoVerbal, that automatically generates paragraph-sized textual descriptions of ontological classes expressed in OWL. OntoVerbal has been applied to ontologies that can be loaded into Protégé and been evaluated with SNOMED CT, showing that it provides coherent, well-structured and accurate textual descriptions of ontology classes.

cs.AI↗

Three Steps to Heaven: Semantic Publishing in a Real World Workflow

Semantic publishing offers the promise of computable papers, enriched visualisation and a realisation of the linked data ideal. In reality, however, the publication process contrives to prevent richer semantics while culminating in a `lumpen' PDF. In this paper, we discuss a web-first approach to publication, and describe a three-tiered approach which integrates with the existing authoring tooling. Critically, although it adds limited semantics, it does provide value to all the participants in the process: the author, the reader and the machine.

cs.DL↗

Lost in translation: data integration tools meet the Semantic Web (experiences from the Ondex project)

More information is now being published in machine processable form on the web and, as de-facto distributed knowledge bases are materializing, partly encouraged by the vision of the Semantic Web, the focus is shifting from the publication of this information to its consumption. Platforms for data integration, visualization and analysis that are based on a graph representation of information appear first candidates to be consumers of web-based information that is readily expressible as graphs. The question is whether the adoption of these platforms to information available on the Semantic Web requires some adaptation of their data structures and semantics. Ondex is a network-based data integration, analysis and visualization platform which has been developed in a Life Sciences context. A number of features, including semantic annotation via ontologies and an attention to provenance and evidence, make this an ideal candidate to consume Semantic Web information, as well as a prototype for the application of network analysis tools in this context. By analyzing the Ondex data structure and its usage, we have found a set of discrepancies and errors arising from the semantic mismatch between a procedural approach to network analysis and the implications of a web-based representation of information. We report in the paper on the simple methodology that we have adopted to conduct such analysis, and on issues that we have found which may be relevant for a range of similar platforms

cs.SE↗

Populous: A tool for populating ontology templates

We present Populous, a tool for gathering content with which to populate an ontology. Domain experts need to add content, that is often repetitive in its form, but without having to tackle the underlying ontological representation. Populous presents users with a table based form in which columns are constrained to take values from particular ontologies; the user can select a concept from an ontology via its meaningful label to give a value for a given entity attribute. Populated tables are mapped to patterns that can then be used to automatically generate the ontology's content. Populous's contribution is in the knowledge gathering stage of ontology development. It separates knowledge gathering from the conceptualisation and also separates the user from the standard ontology authoring environments. As a result, Populous can allow knowledge to be gathered in a straight-forward manner that can then be used to do mass production of ontology content.

cs.AI↗