You're right, there are a lot of academics that like the idea of a semantic web, it rings true to a lot of scientific principles. There are also a lot of ideas and far fewer day-to-day applications. Science ticks a long a lot slower than startup culture, however, so it's not suprising that a) understanding of the issues comes slower, but also b) some "experiments" that would utilize semantic data have not yet been fully tested. Read a list of points that address "why the semantic web is dead", many of those points are precisely what science seeks, i.e. principles that promote a slow, and deep understanding of a domain of knowledge.

With open-science mandates coming from governments around the world researchers are looking for ways to share their data in meaningful ways. I can think of a significant amount of research that regularly consumes RDF, particularly in the fields of medical biology and genomics where it's used to annotate data. This is where I'd guess you'll see it take a foothold, for example medical diagnosis codes are notoriously disparate and there is a strong appreciation for what semantics could address. Unify, exchange, and consume medical diagnoses ... proffit.

Links etc. off the top of my head-

* GO - The gene ontology, used in hundreds of thousands of genomic anotations https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3944782/ * UBERON - https://genomebiology.biomedcentral.com/articles/10.1186/gb-... * The second year of US2TS - http://us2ts.org/2019/posts/registration.html * OBO foundry - https://github.com/OBOFoundry/OBOFoundry.github.io