Dataset

• Training data: 11314 texts
• Test data: 7532 texts
• Unbalanced datasets (see Report.pdf for details)
• Allocated into 20 categories
• Dataset source: https://scikit-learn.org/0.19/datasets/twenty_newsgroups.html

Workflow

Test Results

Model  | Accuracy | Precision(*) | Recall(*) | F1-Score(*) |
------------------------------------------------------------
SVM    |   0.66   |     0.68     |    0.66   |    0.66     |
BiLSTM |   0.67   |     0.68     |    0.67   |    0.67     |

(*) weighted metrics