• bias
• italian word embeddings
• italiano
• rnsb
• weat
• word embeddings

Questo lavoro di tesi ha l’obiettivo di studiare e analizzare la presenza di stereotipi socio-culturali in due diversi modelli distribuzionali dell’italiano, utilizzando le due metriche WEAT (Word Embedding Association Test) e RNSB (Relative Negative Sentiment Bias). Dopo aver effettuato la ricerca tramite il framework WEFE (Word Embeddings Fairness Evaluation Framework), è stato effettuato un confronto tra le due metriche.

This thesis work aims to study and analyze the presence of socio-cultural stereotypes in two different distributional models of Italian, using the two metrics Word Embedding Association Test (WEAT) and Relative Negative Sentiment Bias (RNSB). After the research, using WEFE framework (Word Embeddings Fairness Evaluation Framework), on the basis of the results obtained from each metric in the two different models, a comparison was made between the two metrics.