Assume that someone creates a glossary with one word definitions extracted from a bilingual dictionary with multiple word definitions, choosing only one word to define every word in the other language. could that glossary be published without copyright infringement?
I don't think the question has been put to the test in court. Words of a language are facts which are not copyright-protectable. But simply copying all of the headwords and definitions of a Finnish-English dictionary would be flagrant copyright violation. Randomly (via computer algorithm) extracting a subset of such entries is still copying, just not complete copying. However, you could extract a word or two plus a couple of translations, under the Fair Use exception. It's hard to know whether what you propose might fall under Fair Use, without knowing what you plan to do with the work, and what the original dictionary is. I think you would likely end up copying about 10% of the original, which is pretty extensive (disfavoring Fair Use); if you commercially exploited the result, that makes things worse, but if the dictionary is out of print that makes it better. The more creative your contribution and the greater the difference from the original, the greater the "transformative" nature of the work (supporting Fair Use). An obvious next step would be (if possible) to not just copy words and translations from one source, but to organize a new wordlist, where multiple dictionaries are used as data resources in creating a new dictionary-like object.
I am unsure whether this would be related but facts and false facts are not copyright-able. Fred L. Worth vs Trivial Pursuit 1984, has ruled that facts are not able to be copyrighted due to them being facts.