Novel Features Based on Sentence Specificity for Helpfulness Prediction of Online Reviews
Beatriz Lima, Tatiane Marques Nogueira · 2019
The growth of e-commerce websites have provided a fruitful scenario for the increasing availability of user-generated content. Among the different types of data acquired from customers interaction, product reviews have received a special attention since they are a paramount part of the decision making process both for individuals and companies. However, the large volume of reviews that have a low quality hinders the process of gathering helpful information from these reviews. In the present study, we hypothesize that the specificity (in terms of the level of details) of the sentences from a review influences the perception of the helpfulness of a review. Using Amazon reviews as a case study, the findings of this study show that the proposed features extracted from the specificity degree of the sentences improve the performance when comparing against a common baseline model using unigrams and tf-idf weight. These specificity-based features are also ranked higher than unigrams by three different methods that evaluate features goodness. The results show a strong evidence about the relevance of such features in the task of predicting online review helpfulness.