"Polarity detection" is a technique used in the field of sentiment analysis in texts, a subset of natural language processing (NLP), which determines the emotional polarity of a text, i.e., whether it expresses a positive, negative, or neutral opinion. Despite the impact registered in the field of Natural Language Processing by Deep Learning techniques and, in particular, by Transformers, these approaches come not without downsides, generally related to low-sources languages and domains. In this article, a lexicon-based approach is applied to extract polarity from medical reviews written in Italian from an Italian portal. Specifically, Vader-IT, an adaptation of the popular VADER sentiment analysis tool tailored for the Italian language, is employed to predict the overall sentiment expressed in 5,491 Italian-language reviews about several Italian hospital departments specialized in heart-related diseases. The reviews are accessible on the Qsalute website urlhttps://www.qsalute.it/. The results show a micro averaged F1 − score = 90.8% and a micro averaged Jaccard − score = 83%.
VADER-IT: a Sentiment Analysis Tool for the Italian Language
Martinis M. C.;Zucco C.;Cannataro M.
2024-01-01
Abstract
"Polarity detection" is a technique used in the field of sentiment analysis in texts, a subset of natural language processing (NLP), which determines the emotional polarity of a text, i.e., whether it expresses a positive, negative, or neutral opinion. Despite the impact registered in the field of Natural Language Processing by Deep Learning techniques and, in particular, by Transformers, these approaches come not without downsides, generally related to low-sources languages and domains. In this article, a lexicon-based approach is applied to extract polarity from medical reviews written in Italian from an Italian portal. Specifically, Vader-IT, an adaptation of the popular VADER sentiment analysis tool tailored for the Italian language, is employed to predict the overall sentiment expressed in 5,491 Italian-language reviews about several Italian hospital departments specialized in heart-related diseases. The reviews are accessible on the Qsalute website urlhttps://www.qsalute.it/. The results show a micro averaged F1 − score = 90.8% and a micro averaged Jaccard − score = 83%.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.