Character-based Sentiment Classifier
This sentiment classifier classifies textual data into three classes:
- Positive
- Neutral
- Negative
The classifier is based on google/canine-s, therefore, character-based. Thus, keep in mind that text pre-processing will affect results significantly.
After fine-tuning, the model these achieved these scores in the test set:
- Accuracy: 0.82174688057041
- Precision: 0.8252264610535646
- Recall: 0.8259560973518955
- F1 Macro: 0.820257587240921
- F1 Micro: 0.82174688057041