Threshold-based Na¨ıve Bayes Classificr: Customer satisfaction evaluation

Considering online reviews on Booking.com we propose an ad-hoc classification model for evaluate the Customer Satisfaction, starting from the reviews’ content for predicting them as positive/negative. The log-likelihood ratios attributed to each word included in a review are then used to estimate a numeric score. Such a sentiment score, obtained by a review, is then decomposed w.r.t. the different business areas of an hotel for assess the Customer Satisfaction. The proposed approach is evaluated analysing the reviews provided by tourists who stayed in Sardinian hotels.

Considerando le recensioni online presenti su Booking.com, si propone un classificatore ad hoc per valutare la Customer Satisfaction che, partendo dal contenuto delle recensioni stesse, le classifica in positive/negative. La log-verosimiglianza attribuita ad ogni parola contenuta in una recensione e successivamente utilizzate ` per stimare uno score numerico. Tale sentiment-score, ottenuto da una recensione, e successivamente scomposto considerando le differenti aree d’interesse presenti in ` un hotel al fine di valutare la Customer Satisfaction. L’approccio cos`ı proposto e` validato analizzando le recensioni fornite da turisti che hanno alloggiato in strutture ricettive sarde.