An LLM-Based Methodology for RESTful Service Publication and Discovery

Large Language Models (LLMs) promise to automatically solve various research and industry tasks, including service discovery. However, the main research question is in which ways service discovery can be supported such that its accuracy is increased. The article’s answer to this question is twofold. First, it proposes a novel LLM-based methodology that semantically enriches the OpenAPI description of RESTful services. Second, on top of that methodology, it places novel service discovery algorithms: (a) hybrid ones that exploit both LLM-based embeddings and ontology-based annotations and (b) a configurable and innovative LLM-based service discovery algorithm. All these algorithms are then evaluated in terms of their accuracy against a specific dataset to investigate: (a) whether OpenAPI description enrichment caters for higher accuracy and (b) which of them delivers more accurate results. The main conclusion drawn is that not only is service discovery accuracy elevated, but also the LLM-based algorithm is far better, thus indicating that LLMs can directly tackle the service discovery problem rather than just provide assistance to it.