Final Project for CS 410: Text Information Systems at UIUC

By Kavya Moharana, Pranav Chandra, Sriya Mikkilineni, Aryan Vaswani

Languages, Libraries, and Tools

python NLTK pytorch pandas scikit_learn vega-lite

View Full Report The Analysis (Jupyter Notebook)

Abstract

Green Street’s wide variety of restaurants and cafes is an integral part of student life at the University of Illinois at Urbana-Champaign. Known as the “Heart of Campustown”, these establishments provide an important cultural and social hub for students, faculty, and residents. We intend to conduct sentiment analysis of the Yelp reviews for multiple major restaurants on this historical street. We will analyze online reviews over time periods to determine how students and food enthusiasts perceive these restaurants. Through this evaluation, we can see how the reception of these popular food spots has changed over the past decade.

The Data

We utilized two different datasets during this project: the first from Kaggle for the testing of five different models and one for the evaluation of our actual project. To understand why we needed two models, we need to look at Yelp’s GraphQL API documentation. For this project, we have identified 12 restaurants on Green Street for which we were interested in understanding peoples’ sentiments towards them over time, as shown in Figure 1. However, because Yelp API is still in its beta stage, we are limited to only the first three reviews on each restaurant’s Yelp page. While severely limits our ability in trying to understand Champaign’s preferences, our code still shows sentiments over a couple of months. When the API limits increase, we would only have to re-run the code as is to get all the data.

Figure 1

Figure 2