https://drive.google.com/file/d/1U69x-1YmM3H_V7EW9i-_uX7HHSSHZYs7/view?usp=share_link
This joint talk, by two researchers who have coined well known ‘slogans’ in the knowledge graph community, Jim Hendler, credited with “a little semantics goes a long way” and Deborah McGuinness, known for saying developers need to use “just enough ontology.” Both aphorisms have been in use in the AI and knowledge community since before the advent of deep learning and LLMs and they both were concerned with issues of how to scale semantics as the Web and amount of online data grew.

In the early 2000s, the advent of the Semantic Web, open government data sites, semantic search and Google’s 2012 re-popularization of the term Knowledge Graphs, led to a resurgence of interest in symbolic AI. But, as scalable gen-AI systems created from massive data consumption grew in usage, semantics was downplayed. However, as the challenges relating to these technologies –hallucinations, data protection, data integration and guardrail design) – have become more clear, and with increasing use of AI in complex scientific domains, researchers and business developers have increasingly been recognizing the need for a reintegration of semantic knowledge.

We argue the scale of the systems in use today such for massive knowledge graphs, like wikidata, and large enterprise-wide ontologies requires changes to earlier definitions of the borders of KGs vs. ontologies. In short, a large-scale knowledge graph with “a little semantics” can be a great starting place for “just enough ontology.