Loading...
Publication
Effectiveness of sentiment and emotion detection to unveil customer preferences in portuguese restaurants: insights for restaurant owners
| Name: | Description: | Size: | Format: | |
|---|---|---|---|---|
| 27.92 MB | Adobe PDF |
Authors
Abstract(s)
O trabalho apresentado nesta tese investiga o impacto das avaliações online de uma empresa
na indústria de serviços alimentares em Portugal, enfatizando a importância das plataformas de
avaliação online na formação do comportamento do consumidor e na melhoria dos serviços de
restauração. A pesquisa utiliza dados do Zomato, uma plataforma de descoberta de restaurantes e
entrega de comida, para analisar as avaliações dos clientes e extrair informações valiosas.
O estudo começa com uma visão geral das metodologias usadas na recolha, preparação e pro cessamento de dados, destacando a eficácia do uso de uma base de dados NoSQL, especificamente
MongoDB. É realizada Análise Exploratória de Dados para descobrir padrões e tendências, focando se no tamanho das avaliações, deteção de idioma, tradução e deteção de género.
A deteção de emoções é explorada com uma técnica personalizada que utiliza o léxico EMO TAIX.PT, revelando o impacto significativo do tom emocional nas classificações das avaliações.
A seleção de características é realizada usando os métodos de correlação de Spearman e mRMR,
identificando as características relacionadas com o sentimento como os preditores mais fortes das
classificações das avaliações.
Uma contribuição essencial desta pesquisa é o desenvolvimento do ReviewSight, uma plataforma
web projetada para fornecer aos proprietários de restaurantes informações acionáveis. O Re viewSight integra técnicas de processamento de dados e processamento de linguagem natural numa
interface fácil de usar. A plataforma gera relatórios detalhados incluindo visualizações como gráfi cos de linhas, gráficos de pizza e nuvens de palavras, bem como resumos de texto e identificação
de tópicos. Estes relatórios ajudam os proprietários de restaurantes a entender o sentimento dos
clientes e assim tomar decisões baseadas em dados adaptados às suas necessidades.
Finalmente, é discutida a implementação do ReviewSight numa Máquina Virtual Linux na
Google Cloud Platform, mostrando a robustez e escalabilidade dos serviços de cloud para soluções
empresariais.
The work presented in this thesis investigates the impact of online reviews of a company on the food service industry in Portugal, emphasizing the importance of online review platforms in shaping consumer behavior and improving restaurant services. The research utilizes data from Zomato, a prominent restaurant discovery and food delivery platform, to analyze customer reviews and extract valuable insights. The study begins with an overview of the methodologies used for data collection, preparation, and processing, highlighting the effectiveness of using a NoSQL database, specifically MongoDB. Exploratory Data Analysis is conducted to uncover patterns and trends, focusing on review length, language detection, translation, and gender detection. Emotion detection is explored using a customized technique that employs the EMOTAIX.PT lexicon, revealing the significant impact of emotional tone on review ratings. Feature selection is performed using Spearman correlation and mRMR methods, identifying sentiment-related features as the strongest predictors of review ratings. An essential contribution of this research is the development of ReviewSight, a web platform designed to provide restaurant owners with actionable insights. ReviewSight integrates data pro cessing techniques and natural language processing tools into a user-friendly interface. The plat form generates detailed reports that include visualizations such as line charts, pie charts, and word clouds, as well as text summaries and topic detection. These reports help restaurant owners understand customer sentiment and make data-driven decisions tailored to their needs. Finally, the deployment of ReviewSight on a Linux Virtual Machine on the Google Cloud Platform is discussed, showcasing the robustness and scalability of cloud services for enterprise solutions.
The work presented in this thesis investigates the impact of online reviews of a company on the food service industry in Portugal, emphasizing the importance of online review platforms in shaping consumer behavior and improving restaurant services. The research utilizes data from Zomato, a prominent restaurant discovery and food delivery platform, to analyze customer reviews and extract valuable insights. The study begins with an overview of the methodologies used for data collection, preparation, and processing, highlighting the effectiveness of using a NoSQL database, specifically MongoDB. Exploratory Data Analysis is conducted to uncover patterns and trends, focusing on review length, language detection, translation, and gender detection. Emotion detection is explored using a customized technique that employs the EMOTAIX.PT lexicon, revealing the significant impact of emotional tone on review ratings. Feature selection is performed using Spearman correlation and mRMR methods, identifying sentiment-related features as the strongest predictors of review ratings. An essential contribution of this research is the development of ReviewSight, a web platform designed to provide restaurant owners with actionable insights. ReviewSight integrates data pro cessing techniques and natural language processing tools into a user-friendly interface. The plat form generates detailed reports that include visualizations such as line charts, pie charts, and word clouds, as well as text summaries and topic detection. These reports help restaurant owners understand customer sentiment and make data-driven decisions tailored to their needs. Finally, the deployment of ReviewSight on a Linux Virtual Machine on the Google Cloud Platform is discussed, showcasing the robustness and scalability of cloud services for enterprise solutions.
Description
Keywords
Avaliações online Industria de serviços alimentares Comportamento do consumidor Processamento de linguagem natural Análise de sentimentos Aprendizagem de máquina Online reviews Food service industry Consumer behavior Natural language processing Sentiment analysis Machine learning Informatics Engineering . Faculdade de Ciências Exatas e da Engenharia