metadata
language: eng
tags:
- bert
license: apache-2.0
widget:
- text: The hotel is very nicely located
example_title: Example 1
- text: The reception staff were extremely helpful and very welcoming
example_title: Example 2
- text: There is no balcony in the rooms on the mountain side
example_title: Example 3
- text: A bit pricey
example_title: Example 4
German Hotel Review Sentiment Classification
A model trained on English Hotel Reviews from Switzerland. The base model is the bert-base-uncased. The last hidden layer of the base model was extracted and a classification layer was added. The entire model was then trained for 5 epochs on our dataset.
Model Performance
| Classes | Precision | Recall | F1 Score |
|---|---|---|---|
| Positive | 86.36% | 95.00% | 90.48% |
| Negative | 90.91% | 76.92% | 83.33% |
| Neutral | 50.00% | 42.86% | 46.15% |
| Accuracy | 83.00% | ||
| Macro Average | 75.76% | 71.59% | 73.32% |
| Weighted Average | 82.45% | 83.00% | 82.41% |
Confusion Matrix
