Update README.md
Browse files
README.md
CHANGED
|
@@ -10,18 +10,27 @@ pipeline_tag: text-classification
|
|
| 10 |
inital commit
|
| 11 |
|
| 12 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 13 |
# Performance
|
| 14 |
|
| 15 |
-
This table presents the classification report for a 5-fold
|
| 16 |
-
|
| 17 |
-
|
| 18 |
-
|
| 19 |
-
|
| 20 |
-
|
| 21 |
-
standard deviations suggest that the split into training and test data
|
| 22 |
-
has minimal impact on model performance. Therefore, it is expected
|
| 23 |
-
that the performance of the final model will be comparable to what is
|
| 24 |
-
depicted here.
|
| 25 |
|
| 26 |
| Dimension | Precision | Recall | F1 |
|
| 27 |
|---------------------|---------------|---------------|---------------|
|
|
@@ -29,6 +38,6 @@ This table presents the classification report for a 5-fold
|
|
| 29 |
| People-Centrism | 0.670 (0.011) | 0.725 (0.040) | 0.696 (0.019) |
|
| 30 |
| Left-Wing Ideology | 0.664 (0.023) | 0.771 (0.024) | 0.713 (0.010) |
|
| 31 |
| Right-Wing Ideology | 0.654 (0.029) | 0.698 (0.050) | 0.674 (0.031) |
|
| 32 |
-
|
|
| 33 |
| micro avg | 0.732 (0.009) | 0.805 (0.006) | 0.767 (0.007) |
|
| 34 |
| macro avg | 0.700 (0.011) | 0.770 (0.010) | 0.733 (0.010) |
|
|
|
|
| 10 |
inital commit
|
| 11 |
|
| 12 |
|
| 13 |
+
# Prediction
|
| 14 |
+
|
| 15 |
+
The model outputs a Tensor of length 4.
|
| 16 |
+
The table connects the position of the predicted probability to its dimension.
|
| 17 |
+
|
| 18 |
+
| **Index** | **Dimension** |
|
| 19 |
+
|-----------|--------------------------|
|
| 20 |
+
| 0 | Anti-Elitism |
|
| 21 |
+
| 1 | People-Centrism |
|
| 22 |
+
| 2 | Left-Wing Host-Ideology |
|
| 23 |
+
| 3 | Right-Wing Host-Ideology |
|
| 24 |
+
|
| 25 |
+
|
| 26 |
# Performance
|
| 27 |
|
| 28 |
+
This table presents the classification report for a 5-fold cross-validation of our model.
|
| 29 |
+
The hyperparameters are consistent across all 5 runs. The final and published model was then trained on all data with the same hyperparameters.
|
| 30 |
+
It is evident that the model performs, on average, best for anti-elitism but performs the worst for detecting right-wing host ideology.
|
| 31 |
+
The relatively small standard deviations suggest that the split into training and test data has minimal impact on model performance.
|
| 32 |
+
Therefore, it is expected that the performance of the final model will be comparable to what is
|
| 33 |
+
depicted here.
|
|
|
|
|
|
|
|
|
|
|
|
|
| 34 |
|
| 35 |
| Dimension | Precision | Recall | F1 |
|
| 36 |
|---------------------|---------------|---------------|---------------|
|
|
|
|
| 38 |
| People-Centrism | 0.670 (0.011) | 0.725 (0.040) | 0.696 (0.019) |
|
| 39 |
| Left-Wing Ideology | 0.664 (0.023) | 0.771 (0.024) | 0.713 (0.010) |
|
| 40 |
| Right-Wing Ideology | 0.654 (0.029) | 0.698 (0.050) | 0.674 (0.031) |
|
| 41 |
+
| --- | --- | --- | --- |
|
| 42 |
| micro avg | 0.732 (0.009) | 0.805 (0.006) | 0.767 (0.007) |
|
| 43 |
| macro avg | 0.700 (0.011) | 0.770 (0.010) | 0.733 (0.010) |
|