wordassociations

Predicting human similarity judgments with distributional models: The value of word associations

Predicting human similarity judgments with distributional models: The value of word associations

Finds that semantic networks built from word association data capture similarity judgments better than state-of-the-art corpus models like word2vec.