metadata
language: en
pipeline_tag: text-classification
tags:
- sentiment-analysis
model_bert
Sentiment classifier (positive / negative) fine-tuned from distilbert-base-uncased.
Group members: Gumoshabe Sandra and Keeya Tracy Dataset: stanfordnlp/imdb
Results on the test set
| Metric | Score |
|---|---|
| Accuracy | 0.8868 |
| F1 | 0.89 |
Example predictions
- "Terrible experience, I want my money back." -> Negative
- "This is boring" -> Negative
- "Absolutely wonderful movie with great acting!" -> Positive
Limitations
Trained on English movie reviews only; struggles with sarcasm and highly complex or mixed language text.