BERPT

Fine-Tuning BERT for Sentiment Analysis in Portuguese

Project Description

The increasing integration of technology into our daily lives opens doors for the development of artificial intelligences capable of understanding the nuances of human emotions and opinions. This project aims to fill a crucial gap in the field of sentiment analysis in Portuguese, providing essential resources for training advanced computational models.

Main Objectives:

• Data Collection and Annotation: Implement robust techniques for collecting textual comments in Portuguese, annotated with sentiments, covering various application domains.
• Creation of a Robust Dataset: Develop an extensive and diverse dataset, ideal for training sentiment analysis models in Portuguese.
• Fine-Tuning BERT: Use the pre-trained bidirectional Transformers model - BERT for specific fine-tuning on Portuguese textual data. This process aims to maximize efficiency and accuracy in sentiment classification.
• Resource Availability: Make the trained model and annotated dataset available to the open-source community, facilitating access and encouraging the continuous development of natural language processing applications in Portuguese.

Expected Benefits:

• Advancement in Research: Contribute to significant advances in understanding and analyzing sentiments in Portuguese.
• Development Facilitation: Provide readily usable resources for researchers and developers interested in sentiment analysis.
• Community Impact: Promote sustainable development within the Portuguese NLP community, strengthening foundations for future innovations.

<i>This project not only aims to create and provide valuable resources but also aims to establish a standard for excellence in sentiment analysis in Portuguese, supporting future initiatives in artificial intelligence and NLP.</i>

Metrics and Results

• Precision/Recall/F1
• ROC Curve:

You can find the thesis associated with this research by accessing: <a href="https://github.com/RenanPeres/RenanPeres/blob/main/thesis/bachelor_final_project.pdf" target="_blank"> <img src="https://img.shields.io/badge/-GitHub-%230A0A0A?style=for-the-badge&logo=github" target="_blank"></a>

<p style="font-size: smaller;">Keywords: Sentiment Analysis, Data Mining for Portuguese Language, Word Embedding, Natural Language Processing, BERT Model, App Reviews.</p>