Restaurant review system with data visualization and text search using TF-IDF algorithm / Syazwina Abd Hamid

A single bad review can poison the mind of reader. That shows how online reviews are playing an important asset for decision-making. Today, restaurant review system allow feedback of customers based on personal opinion and experience to be shared with others. However, high availability of reviews made it impractical for user to read them all where users found it frustrating to go thru each one for enough information. Besides that, the text in reviews convey better information than the rating alone thus encourages deeper study on the review content. Furthermore, current approach is lacking and needs new improvement to yield better results when user search for restaurant. With that being said, feature for easy access of review through searching and to provide analysis on these data would be highly beneficial to many, from investor, business owner and Internet user. Hence, this project focuses on the integration of information retrieval with data visualization in order to produce valuable insights. Term Frequency-Inverse Document Frequency (TF-IDF) algorithm is performed for text search, where it satisfies multiple keywords at a time and the search result for review is based on relevancy score. The implementation of TF-IDF will have to go through preprocessing such as tokenization, convert to lowercase, removal of special characters and stop words and lastly, stemming. Preprocessing aims to clean the review text before the score is calculated which boost the performance of information retrieval. Meanwhile, for visualization, D3.js library is used with aim to promote interactivity and enhance user experience when using the system. Visualizations used in the system are circular bar plot, word cloud and bar chart where each visualization conveys different purpose. For dataset, the data was scrapped entirely from TripAdvisor website and it only considers review data and restaurant data. Various features are deliver from this system such as login, logout, sign up as member, post new review, view submitted review or even delete reviews, search of restaurant, view restaurants, read reviews of restaurant and lastly, view overall analysis for insights where top 10 restaurants are ranked based on its average rating and total number of review. Outcome gained from functionality and usability testing indicate that all functions in the system work perfectly and received promising feedback from users. The future work of this project is to improve data collection to build better dataset, implement sentiment analysis and provide recommendation for restaurants.