Fine-grained sentiment analysis using neural networks to identify guest preferences based on online reviews

Online reviews are a sort of user-generated content that sway not only prospective guests' reservation decisions but also provide timely feedback to hotels on how to enhance their service offerings in response to guest feedback. Customer preferences can be inferred from online hotel evaluations by analysing the reviews for subtle indications of approval or disapproval of specific features. Sentiment component extraction, feature pair identification (FPI), and sentiment orientation analysis are all basic activities in fine-grained sentiment analysis. Yet, current fine-grained sentiment analysis techniques fail to identify FPIs reliably, particularly when dealing with review evaluations. To boost the efficiency of FPI detection, we develop a new convolutional neural network (i.e., CNN) model that can fully exploit unstructured and structured information. In addition, we present a modified fine-grained sentiment analysis methodology that integrates a term clustering algorithm for “aspect terms” to determine a correct “customer sentiment intensity value” for each evaluated aspect, rather than a simple “positive” or “negative” rating. At last, we undertake an empirical analysis of hotel web reviews to demonstrate the rationality and benefits of the suggested methodology. The empirical findings show that our suggested method can effectively identify consumer preferences from online evaluations of hotels and can enhance the effectiveness of FPI identification. Furthermore, we discover that different sorts of clients have distinct tastes in the amenities offered by hotels.