Ontologies of research topics are crucial for structuring scientific knowledge, enabling scientists to navigate vast amounts of research, and forming the backbone of intelligent systems such as search engines and recommendation systems. However, manual creation of these ontologies is expensive, slow, and often results in outdated and overly general representations. As a solution, researchers have been investigating ways to automate or semi-automate the process of generating these ontologies. One of the key challenges in this domain is accurately assessing the semantic relationships between pairs of research topics. This paper presents an analysis of the capabilities of large language models (LLMs) in identifying such relationships, with a specific focus on the field of engineering. To this end, we introduce a novel benchmark based on the IEEE Thesaurus for evaluating the task of identifying three types of semantic relations between pairs of topics: broader , narrower , and same-as . Our study evaluates the performance of seventeen LLMs, which differ in scale, accessibility (open vs. proprietary), and model type (full vs. quantised), while also assessing four zero-shot reasoning strategies. Several models with varying architectures and sizes have achieved excellent results on this task, including Mixtral-8 × 7B, Dolphin-Mistral-7B, and Claude 3 Sonnet, with F1-scores of 0.847, 0.920, and 0.967, respectively. Furthermore, our findings demonstrate that smaller, quantised models, when optimised through prompt engineering, can achieve strong performance while requiring very limited computational resources.

Ontologies of research areas are important tools for characterising, exploring, and analysing the research landscape. Some fields of research are comprehensively described by large-scale taxonomies, e.g., MeSH in Biology and PhySH in Physics. Conversely, current Computer Science taxonomies are coarse-grained and tend to evolve slowly. For instance, the ACM classification scheme contains only about 2K research topics and the last version dates back to 2012. In this paper, we introduce the Computer Science Ontology (CSO), a large-scale, automatically generated ontology of research areas, which includes about 26K topics and 226K semantic relationships. It was created by applying the Klink-2 algorithm on a very large dataset of 16M scientific articles. CSO presents two main advantages over the alternatives: (i) it includes a very large number of topics that do not appear in other classifications, and (ii) it can be updated automatically by running Klink-2 on recent corpora of publications. CSO powers several tools adopted by the editorial team at Springer Nature and has been used to enable a variety of solutions, such as classifying research publications, detecting research communities, and predicting research trends. To facilitate the uptake of CSO we have developed the CSO Portal, a web application that enables users to download, explore, and provide granular feedback on CSO at different levels. Users can use the portal to rate topics and relationships, suggest missing relationships, and visualise sections of the ontology. The portal will support the publication of and access to regular new releases of CSO, with the aim of providing a comprehensive resource to the various communities engaged with scholarly data.