Sentiment Analysis of Restaurant Reviews in Portuguese: A Transfer Learning and Ensemble Approach with Edge Computing

• Autor(a) principal: Branco, Alexandre João Jardim
• Data de Publicação: 2024
• Tipo de documento: Dissertação
• Idioma: eng
• Título da fonte: Repositório Científico de Acesso Aberto de Portugal (Repositórios Cientìficos)
• Texto Completo: http://hdl.handle.net/10400.13/5602
• topic: Análise de sentimento; Processamento linguagem natural; Língua portuguesa; Computação de borda; Transferência de conhecimento; Transformers; Sentiment analysis; Natural language processing; Portuguese language; Edge-computing; Transfer-learning; Electrical Engineering – Telecommunications; .; Faculdade de Ciências Exatas e da Engenharia; Domínio/Área Científica::Engenharia e Tecnologia::Engenharia Eletrotécnica, Eletrónica e Informática

This research focuses on a case of applying transfer learning and transformer-based pre trained models to sentiment analysis in Portuguese in restaurant reviews. Specifically, we em ploy BERT and RoBERTa, two strong Language Models that fit into limited computational resources, like edge computing, to build a sentiment review classifier. The classifier’s perfor mance is evaluated using accuracy and AUC ROC as the primary metrics. Our results demon strate that the classifier developed using ensemble techniques outperforms the baseline model in accurately classifying restaurant reviews. This research contributes to sentiment analysis by exploring the effectiveness of transfer learning and transformer-based models in the context of Portuguese restaurant reviews. This work highlights the importance of considering the Portuguese language in sentiment analysis tasks. Furthermore, this study investigates the deployment of the model on edge com puting platforms, making sentiment analysis more accessible in resource-constrained environ ments. Combining deep learning techniques, transfer learning, and edge computing offers promising real-time sentiment analysis application opportunities. This research provides valu able insights for researchers and practitioners interested in sentiment analysis, natural language processing, and text analysis in the context of restaurant reviews.