I need data (preferably human judgment) for the structural/hierarchical semantic distance between many couples (at least hundreds) of words.

For example, d(computer, television) < d(radio, television) < d(dish washer, television).

If we organize all words in a dendrogram or a tree, where each node is a category ("electric device", "with screen", etc...), and words are in leaves, the number will represent several steps (nodes) we have to go from one word to another.

Does such a dataset exist? per couples ratings are enough, no need to have a full embeding/tree/specify the nodes (An example dataset will be:

Computer Television 1

Radio Television 2

DishWasher Television 3

I'm aware of the WordNet database but it has some very unintuitive definitions and the results achieved using it are not very correspond to human ratings. I would like to have a more human-rating alike data-set.

  • One keyword that might be helpful is "semantic differential" by Osgood.
    – fgregg
    Commented Nov 2, 2017 at 19:02
  • You could create such a list from wikidata.org but it'd probably be ever worse than WordNet.
    – user3856
    Commented Nov 3, 2017 at 16:43


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Browse other questions tagged or ask your own question.