Imagine a vast library of software, each project meticulously categorized and easily discoverable. That's the dream of every developer, researcher, and software enthusiast. But building such a comprehensive, interconnected taxonomy of software application domains is no easy task. Traditional methods, relying on manual curation or rigid structures, often fall short, failing to capture the dynamic, evolving nature of the software world. In this post, we explore how a novel approach, combining the strengths of multiple data sources and AI, can revolutionize software taxonomy construction. Previous attempts at classifying software applications have encountered significant roadblocks due to inconsistencies in existing classifications, including mixing categories, overlapping topics, and outdated labeling practices. This research tackles these challenges head-on by integrating diverse data sources: a specialized Computer Science Ontology (CSO), the collaborative knowledge base Wikidata, and the generative power of Large Language Models (LLMs). Each source brings unique advantages and limitations to the table. The CSO offers structured knowledge but lacks coverage of newer technical terms. Wikidata, while comprehensive, introduces noise from abstract or irrelevant concepts. LLMs excel at linking terms but struggle to create a cohesive hierarchical structure. By strategically combining these sources in an ensemble approach, researchers were able to construct a more robust and complete taxonomy. This approach not only reduces inconsistencies but also creates a more interconnected and navigable structure, making it easier to classify and retrieve software projects. The results are promising. Human evaluations confirmed the accuracy of the majority of term relationships within the taxonomy, highlighting the effectiveness of this ensemble method. Furthermore, the research explores the use of LLMs as automated evaluators, offering a scalable and cost-effective way to assess taxonomy quality. While not yet as precise as human judgment, LLMs provide valuable insights and complement human evaluation. This work represents a significant leap toward a more organized and accessible software landscape. Future applications of this research could transform code search and recommendation systems, enabling developers to discover and reuse relevant software components more efficiently. By automating the taxonomy construction process, and leveraging the strengths of AI, we move closer to a future where navigating the vast software universe is not a chore but a seamless, enlightening experience.