@@ -12,11 +12,11 @@ The dataset used consist of 50k labeled IMDb movie reviews. Due to hardware cons
<center>
| | Train | Test |
| :-------: | :---: | :---: |
| Positive | 5,189 | 4,766 |
| Negative | 1,707 | 1,613 |
| **Total** | 9,955 | 3,319 |
| | Train | Valid | Test |
| :-------: | :---: | :---: | :---: |
| Positive | 4,849 | 1,033 | 1,013 |
| Negative | 4,442 | 958 | 979 |
| **Total** | 9,291 | 1,991 | 1,992 |
</center>
...
...
@@ -36,7 +36,7 @@ our model, precision, recall and f1-score will serve as a complementary to the a
As baseline for this project, a regular BERT model has been implemented and fine tuned on the task of classifying the sentiment of IMDb reviews.
Training our baseline model for 1 epoch using a batch size of 32 yielded the following results:
Training our baseline model for 1 epoch using a batch size of 32 yielded the following average results:
<center>
...
...
@@ -50,7 +50,7 @@ Training our baseline model for 1 epoch using a batch size of 32 yielded the fol
### Method 1
Method 1 implements a multi layer perceptron to combine the fine-tuned BERT model from our baseline with VAD-scores from VADER. Training the MLP implementation yielded results as follows:
Method 1 implements a multi layer perceptron to combine the fine-tuned BERT model from our baseline with VAD-scores from VADER. Training the MLP implementation yielded average results as follows:
<center>
...
...
@@ -62,7 +62,7 @@ Method 1 implements a multi layer perceptron to combine the fine-tuned BERT mode
### Method 2
Method 2 assigns weights to the individual results from the fine-tuned BERT and VADER and combines the models with different weight-combinations. The best combination of weights yielded the following results:
Method 2 assigns weights to the individual results from the fine-tuned BERT and VADER and combines the models with different weight-combinations. The best combination of weights yielded the following average results:
