Extracting Relevant Elements from Consumer-Submitted Restaurant Reviews Using Natural Language Processing to Influence Business Strategy

Quick service restaurant (QSR) operators are trying to find ways to improve their customer’s experience that does not entail discounting their already cost-sensitive products. The focus of this project was to look to a different channel to devise a business strategy related to improving the customer experience. Companies in the QSR industry have shown interest in using consumer-submitted review data to understand what their customers are saying about them. However, manually extracting and analyzing consumer-submitted reviews for relevant insight is time-consuming and inefficient. This project aimed to address this problem by creating a pipeline to mine Yelp reviews for a target business and use NLP methods on the review text. A domain-specific named entity recognition (NER) model and a sentiment analysis model were trained using the review text. The NER model learned to detect mentions of food and service relating to the customer experience. The sentiment analysis model learned to accurately predict the sentiment of all sentences within the review text which enabled a nuanced view of the reviews that contain varying degrees of sentiment.