The semantic similarity is not always sensible, either. The underlying model needs some work. (spoilers for today below) . . . . . . . These are some o

The semantic similarity is not always sensible, either. The underlying model needs some work. (spoilers for today below) . . . . . . . These are some of my guesses for today, in order of similarity: kilometer 3501 authority 3149 tomato 2960 exercise 2357 fruit 2008 healthy 1715 run 972 input 927 power 909 foot 598 meter 300 control 159 Ugh! I feel like I'm blindly stumbling around a maze of puns, or a Saturday NY Times crossword full of eye-rolling clues. It just feels like unsystematic guessing at what the semantic similarity model might contain.