Are humans always needed for relevance

I’ve been thinking about the claim that you always need humans to define relevance or meaning – trying to imagine a case in which this is not true. Search engines today give you a list of relevant links and sometimes they cluster queries or results or try to link clusters of queries to clusters of results. Are you one of the “spring as in metal coil” people looking for a “spring as in metal coil” result? This kind of relevance is coming from user data, maybe also a team of people whose job it is to judge results, and maybe using a structure like WordNet to disambiguate.

I found this post from Freebasics interesting about Wikipedia-esque goals for structured data.

Control of the structure is given to the people who understand their domains the most — sports fans, scientists, trainspotters, computer gamers, journalists, historians, gadget freaks and eccentrics who collect plaster gnomes.

Even if we have machines suggesting new, better structure than the community of enthusiasts can come up with, people are still at the beginning of the process setting up the parameters and at the end saying, “This looks good, let’s use it.”