Semantic Relatedness Judgments of Words and Pseudowords: An Experimental Study with Italian Speakers

Understanding the mechanisms subserving the semantic processing of words and pseudowords is an issue that may provide valuable insight into human language processing. The present study aims to investigate whether Italian native speakers can identify -among triplets of letter strings. which word is semantically more related to the given pseudowords, comparing them with the classification of Distributional Semantic Model (DSM), namely fastText. The model was used to calculate the semantic relatedness between words and pseudowords considering cosine similarity; in addition, we controlled for a series of possible confounding factors such as word length, word frequency, and orthographic distance. Sixty native Italian speakers were recruited in an online experiment. They were asked to determine which word was semantically similar to the pseudoword. Participants’ responses of semantic similarity between words and pseudowords were consistent with DSM classification, with an accuracy of 58%. Further analysis showed that only the cosine similarity predicted participants’ rates successfully. This research provides evidence for the alignment of DSMs on predicting human judgments on semantic relationships between words and out-of-vocabulary words, as demonstrated by an explicit task for the first time in literature. Hence, the findings support the non-arbitrary view of natural language.