I'm the developer of OneLook and happy to answer any questions about it! It's been running continuously since 1996 and I've been (very slowly) adding features to it over the past couple of decades.
OneLook crawls and indexes the headwords from ~1000 online dictionaries and glossaries. When you do a wildcard search, it searches all of these lists at once. Clicking on one of the results will show you the sources. The sites range from small subject-specific glossaries with a few dozen words, to commercial dictionaries like Merriam-Webster and OED.
And Nutrimatic's own instructions link to http://www.oneacross.com/ which is crossword-specific. I've been using OneLook when I get really stuck on a crossword, but of course there are more options!
I could really use something like that, with the additional ability to add custom sources.
My use cases:
I read e-books in Calibre[1]. When I read novels in The Witcher series or in the Forgotten Realms meta-series, I need to lookup words that may or may not be just a creature in the universe of the novel. Having a page that automatically looks up a word in both real dictionaries and in The Witcher or Forgotten Realms wiki would be really cool.
My daughter also reads e-books in Calibre, but her English is nowhere near my level. It's a second language for her. So she needs to be able to look up words in English-Chinese dictionaries.
If there's any other word-geeks out there, check out http://wordo.co
and follow me there: https://wordo.co/@aminozuur
i'll follow back so we can both learn interesting words
I use the OneLook reverse dictionary all the time to try to find the right word (not for crosswords), but I do wish it were smarter. If only someone would make a deep-learning-based reverse dictionary.
Me too! I've carved out some time this summer to look into some approaches to the problem based on the BERT architecture. A few years ago I tried some (pre-BERT) deep-learning based approaches, but nothing panned out. The "dumb" information-retrieval based approach that my service currently uses is certainly bad, but stubbornly not as bad as all the alternatives I've tried in the 15 years since creating it!