Transfer learning is a machine learning technique that uses existing knowledge from pre-trained models to solve problems in different domains. Many researchers have applied transfer learning to the field of sentiment analysis to create state-of-the-art models. Usually, applications in sentiment analysis are data-driven, and the resources currently available mostly cover only a couple of text genres in specific contexts. In this project, we propose an approach that uses transfer learning to classify the sentiment of Portuguese texts of various genres. To this end, we create two resources: a dataset from affective knowledge obtained from a collection of lexicons/corpora and a sentiment classifier based on a pre-trained model with Portuguese language data. We compare the accuracy of the proposed classifier through eight benchmark datasets. Experimental results show a consistent improvement of our approach over conventional models that were tested in our experiments
BibTeX:
In process of review
The repository has the following structure
| File | Description | |
|---|---|---|
| SentPt | Presents a classifier to detect sentiment polarity (notebook: SentPt.ipynb) |  |
| Dataset | Link to the folder containing the created datasets |