Presentation Title

Which restaurant to choose? - Exploring topics and sentiments of Yelp restaurant reviews

Faculty Mentor

Sonya Zhang

Start Date

23-11-2019 8:00 AM

End Date

23-11-2019 8:45 AM

Location

117

Session

poster 1

Type of Presentation

Poster

Subject Area

business_economics_public_administration

Abstract

Our research study intends to develop a recommender system based on the topic and sentiment for online reviews using text mining. Previous research shows that consumers and businesses use online reviews for a variety of reasons. For many products/services, there are a large number of reviews that make it difficult for consumers to decide which reviews to pay attention to, or businesses to make decisions which areas to improve on. Our prior study developed a content filtering recommender system based upon 8 restaurant review factors of a Yelp dataset. Our current project plans to extend the previous study by adding topics extraction and sentiment analysis. We also used a more updated Yelp round 13 dataset, which after cleaning contains reviews of 11,685 restaurants that are located mostly in Ontario (2,902), Arizona (2,274), and Nevada (1,575). We then used a natural language processing method called Latent Dirichlet Allocation (LDA) to extract the top 10 topics from the reviews. Topics of Food, Price, Service, Location, Ambiance, Menu Options, and Waiting Time stood out from our preliminary findings.Then we used a Python library called TextBlob to determine if the sentiment of each review, review sentence, and review topic. Interestingly, we found more positive and neutral reviews than negative reviews from our preliminary findings.

This document is currently not available here.

Share

COinS
 
Nov 23rd, 8:00 AM Nov 23rd, 8:45 AM

Which restaurant to choose? - Exploring topics and sentiments of Yelp restaurant reviews

117

Our research study intends to develop a recommender system based on the topic and sentiment for online reviews using text mining. Previous research shows that consumers and businesses use online reviews for a variety of reasons. For many products/services, there are a large number of reviews that make it difficult for consumers to decide which reviews to pay attention to, or businesses to make decisions which areas to improve on. Our prior study developed a content filtering recommender system based upon 8 restaurant review factors of a Yelp dataset. Our current project plans to extend the previous study by adding topics extraction and sentiment analysis. We also used a more updated Yelp round 13 dataset, which after cleaning contains reviews of 11,685 restaurants that are located mostly in Ontario (2,902), Arizona (2,274), and Nevada (1,575). We then used a natural language processing method called Latent Dirichlet Allocation (LDA) to extract the top 10 topics from the reviews. Topics of Food, Price, Service, Location, Ambiance, Menu Options, and Waiting Time stood out from our preliminary findings.Then we used a Python library called TextBlob to determine if the sentiment of each review, review sentence, and review topic. Interestingly, we found more positive and neutral reviews than negative reviews from our preliminary findings.