Abstract
This paper explores the application of Natural Language Processing (NLP) methods in sentiment analysis of restaurant reviews available online, for a sample of restaurants in the Algarve region. The primary objective was to develop an automated method that could efficiently extract and categorize relevant sentiments relating to five key attributes of customer satisfaction, namely food quality, service, ambient, price and restaurant’s location. Using the F1 Score the proposed method was compared against human classification benchmarks. The results showed that Universal Sentence Encoding (USE) was a suitable method for implementation due to its acceptable F1 score performance, ease of accessibility and reduced cost. The use of semantic embeddings can provide valuable insights from online reviews that could benefit the restaurant management and in general the data-driven decision-making processes businesses in the gastronomic sector.
Recommended Citation
Carrasco, Paulo and Dias, Sandra, "Exploring Natural Language Processing and Sentence Embeddings for Sentiment Analysis of Online Restaurant Reviews" (2023). CAPSI 2023 Proceedings. 9.
https://aisel.aisnet.org/capsi2023/9