Food Review Sentiment Analysis In Bangla Language: A Comparative Analysis Of Boosting And Traditional Machine Learning Algorithm

Abstract

Even in this day and age of advanced internet technology, businesses that deliver meals via the internet in Bangladesh are doing quite well. Even after the COVID-19 pandemic, the majority of people living in urban areas of Bangladesh are unfazed when it comes to placing meal orders with a variety of food delivery services. The expansion of the food supply has made it possible for consumers to experience previously unattainable levels of convenience. People frequently consult food reviews and ratings before making decisions about what to eat. Ratings are not the only factor that goes into determining the caliber of a dish. The menu, the delivery monitoring program and the attitude of the delivery person are all elements that go into the rating. As a consequence of this, the reader is required to read every single comment pertaining to food. However, this process does require some patience, mainly due to the fact that it is difficult to read all of the comments on the different dishes. As a consequence of this, we attempted to develop a system for providing insightful feedback regarding meals. Around two thousand Bangla phrases have been compiled by our team for use in a variety of online meal delivery applications. Throughout the course of our investigation, we made utilized a number of different methods for analyzing sentiment and classifying data. Some of these traditional machine learning methods include DT, RM, and LR. Compare with Boosting classifier algorithms. We made utilized an algorithm that not only gave us an F1 score but also gave us the highest possible level of accuracy in our prediction.