WARBERT: Hierarchical BERT Model Addresses Web API Recommendation Challenges

A novel BERT-based model, named WARBERT, has been created to enhance Web API recommendation systems. This model tackles three ongoing issues: semantic ambiguities in the comparison of API and mashup descriptions, inadequate progressive semantic refinement between mashup needs and specific API details, and inefficiencies in computation within large repositories that necessitate exhaustive comparisons between mashups and APIs. As Web 2.0 technologies and microservices architectures expand, the demand for effective Web API recommendations has surged. Current methods generally fall into two categories: those that classify APIs using labels and those that match APIs directly with mashups. WARBERT utilizes dual-component feature fusion and attention mechanisms for accurate semantic representation, featuring WARBERT(R) for initial candidate selection. This research was published on arXiv, identified as arXiv:2509.23175v2, under the announcement type replace-cross.