Feature Validated Sentiment Recommendation of Hotel Reviews

Tourists can easily find the hotels of any particular place and share their opinions on social media and other tourism apps. These reviews give an abstraction to the readers about a particular hotel. Somehow, the readers may get confused on choosing the right hotel due to ambiguity in many reviews. Sentiment classification methods will be used to differentiate the positive and negative sentiments. However, each hotel review might have a binary statement which makes it difficult to identify whether the statement is positive or negative. In order to tackle the ambiguity, the tokenized sentences are considered. Naive Bayes classifier, a supervised machine learning algorithm, is applied to recognize the features and to classify the reviews into positive or negative. Then, the proposed framework will be implemented as a Web application using open-source methodologies and micro frameworks. The dashboard of the Web application comprises three areas, namely data, recommendation and a plot. Data shows the reviews along with the classified sentiments of the selected hotel. Recommendation part shows the levels of recommendations for the selected hotel, and the plot shows the positive and negative statements of that particular hotel. A word cloud will be presented for the effective visualization of key terminologies in the positive and negative statements. Previously, review classification was taken into consideration as a part of sentiment analysis. Here, in this approach, sentiment recommendation combines the techniques of sentiment classification and recommendation of deserved entity.