On Tripadvisor, customers can query anything ranging from restaurants, tour guides, flights, and hotel booking. The website suggests related reviews, pictures, ratings and suggestions given by other visitors. The client also supports online booking, allowing customers to easily book their travels with one-click. Our project is to implement an on-line clustering algorithm for review data written by customers. On some timed-interval basis, we would like to classify new clusters of those reviews into categories and also connect them with sentimental analysis, e.g. to be able to identify items that customers liked the most or complained about the most.
Currently, the client still faces some challenges. One problem suggested to our team is how to better utilize reviews given by customers that have been collected. By analyzing those data, the company hopes to better understand business providers’ services so as to improve the website’s recommending accuracy. Furthermore, understanding customers’ needs will also help TripAdvisor’s service and discover the hidden potential markets in certain areas. We (the Team, comprising the individuals listed above) hope to meet this challenge.
As of right now, our team has managed to reorganize data given by the company and scraped online. With the large amount of data available, the team is currently working on hotels and restaurants in Boston area. Utilizing many packages available online, the team has been trying to apply sentimental analysis and topic models on those data. In the meantime, our team is also exploring other potential tools available mentioned in the recent literature as well.