Local Score Solving

Average solving time : 5 seconds (with non-filtered model)

Average solving time : 2 seconds (with filtered model)

This method relies on having a model that produces the same scores as Cemantix.

1. I request the scores for 3 starting words directly from Cemantix.

2. I compute the similarity between every word in my model and the starting words locally. Since we’re using the gensim and numpy libraries, this is much faster than querying Cemantix.

   • If I find a word with the same similarity as the starting words’ score, I test it on Cemantix. If it has a score of 1.0, the solution is found and I stop.

   • Otherwise, I continue the search.

If no matching word is found, it likely means that the target word is not present in the model.