This paper discusses a new approach to creating semantic resources consisting of complexassociations among words that can be used for evaluating the content of word embeddings aswell as in various language-learning scenarios. We briefly introduce Codenames – an existingparty board game – and the way of recording word associations suggested by human players.Advanced word embedding models are then compared on the collected data and it isdemonstrated that they often fail in the cases of complex word associations that go beyondsimple contextual interchangeability. We conclude with an initial evaluation of the automaticguessing of associated words based on clues provided by human players and a discussion onfurther extensions of the system towards a wide language coverage and explanations of wordassociations in the language learning context.
展开▼
