4 ms·
It used to be used in NLP, but never to great success. Word embeddings are a far more powerful way to achieve a lot of similar goals, but with easier computatio
by azinman2 6y ago
It used to be used in NLP, but never to great success. Word embeddings are a far more powerful way to achieve a lot of similar goals, but with easier computation, easier scalability to other languages, and accommodation for new/personal words that aren't (yet) in the dictionary.
- mark_l_watson 6y agoI agree with you, sort of, but WordNet based systems can be explainable. Word embedding (and sentence embedding, etc.) is great, but like deep learning models is a black box.
- azinman2 6y agoI wouldn't assume even that... "X was chosen because a matrix decomposition suggested it was relevant" isn't so great either... WordNet is/was rarely used as just a single hop, those hops are somewhat arbitrary to begin with, and ultimately it never really materialized as a useful tool. If it can't be used for anything, then there's nothing to explain!