Hacker News Reader

Thanks for mentionning it.

Beyond some very nice pieces of software created by clever SW developpers (let's mention DBPedia Spotlight, RelFinder, Semantic MediaWiki, N3.js) I mostly agree with your argument.

To be the devil's advocate, in my opinion, data lakes are just a new buzzword that will eventually become what the Semantic Web was supposed to be. With some millions dollars, companies will aggregate and interlink their datasets. There are then one step away from extracting their master data from their data lake & publish it as Linked Corporate Data. From that, the public part of will be published as Linked Open Data.

I remember my huge number of delicious bookmarks tagged with "folksonomy" and other buzz words.

That's not entirely true. It's perhaps a far cry from the original goal but Schema.org markup has definitely caught on mostly thanks to use by Google for rich snippets.

I get the impression that "dumb" crawling and machine learning have turned out to be more practicable.

It is definitely more practicable in the sense that no-one has figured out how to implement the vision of the semantic web.