Text-rating review discrepancy (TRRD): an integrative review and implications for research

The large number of online product and service review websites has created a substantial information resource for both individuals and businesses. Researching the abundance of text reviews can be a daunting task for both customers and business owners; however, rating scores are a concise form of evaluation. Traditionally, it is assumed that user sentiments, which are expressed in the text reviews, should correlate highly with their score ratings. To better understand this relationship, this study aims to determine whether text reviews are always consistent with the combined numeric evaluations. This paper reviews the relevant literature and discusses the methodologies used to analyse reviews, with suggestions of possible future research directions. From surveying the literature, it is concluded that the quality of the rating scores used for sentiment analysis models is questionable as it might not reflect the sentiment of the associated reviews texts. Therefore, it is suggested considering both types of sources, reviews’ texts and scores in developing Online Consumer Reviews (OCRs) solution models. In addition, quantifying the relationship degree between the text reviews and the scores might be used as an instrument to understand the quality of rating scores, hence its usefulness as labels for building OCRs solution models.