TL;DR: The authors pretrain the model to classify images into Wordnet synsets[a] that appear in the caption, using a standard Cross Entropy loss. They keep the number of classes relatively small by removing any synsets that don't show up in captions at least 500 times in the dataset. It seems to work well. My immediate question is: Why not classify among the entire hierarchy of all Wordnet synsets? --- [a]... - Source: Hacker News / about 1 month ago

To operationalize this intuition, the Microsoft and UC Berkeley researchers use WordNet and Wiktionary to augment the text in image-text pairs. The concept itself is augmented for isolated concepts, such as the class labels in ImageNet, whereas for captions (such as from GCC), the least common noun phrase is augmented. Equipped with this additional structured knowledge, contrastively pretrained models exhibit... - Source: dev.to / 3 months ago

If you like this, definitely check out WordNet (https://wordnet.princeton.edu/). - Source: Hacker News / 5 months ago

I didn't understand well what you meant, but maybe this site can help you: https://wordnet.princeton.edu/. Source: over 1 year ago

What I'd do is work with a huge database like WordNet and then try to "extrapolate" BIP39 to 4096 words by creating queries against WordNet to obtain words meeting the constraints you'd like to keep. Source: over 1 year ago

I looked for some alternatives and found http://wordgrabbag.com/ which definitely meets my needs. Thanks again! Looks like I was a bit off in the year estimate: https://gcemetery.co/google-sets/ says it existed from 2002-2011. Source: over 2 years ago