Applying NLP techniques to characterize what makes an online review trustworthy
, J. C. ROMERO
, R. MARTÍNEZ TORRES
, S. TORAL
2022, Editorial Universitat Politècnica de València
Source credibility , Trustworthiness , Helpfulness , Online reviews , Classifier , Natural Language Processing
Abstract
Users spend a significantly amount of time reading and exchanging reviews online in e‑commerce and eWOM communities that help them with their purchase decisions. Source credibility theory is gaining more importance as some online reviews are currently being damaged by those fake reviews that promote an untruthful image not only of the products but also of those online websites. Thus, trustworthiness of online reviews is a key aspect not only for the users that want to make more informed decisions regarding the products, but also for the websites whose credibility might be affected. In this regard, this study proposes a classification system using two Natural Language Processing (NLP) models that can predict trustworthy online reviews (helpful and truthful) applied to the product category “Cell phones & accessories” of Amazon. After using a keyword extractor among those trustworthy online reviews we can characterize their most important features. The results reveal that those features are related to brands, physical and technical features and the UX of the mobile phones.