DoManhQuang/phobert-cnn-text-classification

vietnamese-text-classification

vietnamese-text-classification-with-phobert-cnn

Bộ dữ liệu tiếng việt sử dụng "A Large-scale Vietnamese News Text Classification Corpus"

---

Project

• quick_start_data_NLP_Pho_Bert.ipynb : Xử lý dữ liệu
• text_classification_with_CNN.ipynb : Huấn luyện mô hình

---

Train/Test: 80/20

Classification Report

	precision	recall	f1-score	support
Class 0	0.75	0.77	0.76	200
Class 1	0.82	0.89	0.86	200
Class 2	0.87	0.79	0.83	200
Class 3	0.85	0.83	0.84	200
Class 4	0.91	0.87	0.89	200
Class 5	0.91	0.87	0.89	200
Class 6	0.85	0.88	0.86	200
Class 7	0.99	0.96	0.97	200
Class 8	0.90	0.95	0.92	200
Class 9	0.90	0.92	0.91	200
accuracy		0.87	2000
macro avg	0.87	0.87	0.87	2000
weighted avg	0.87	0.87	0.87	2000

---

Confusion Matrix

ROC KFold = 10

ROC best - Classification Report

	precision	recall	f1-score	support
Class 0	0.72	0.84	0.78	100
Class 1	0.88	0.91	0.90	100
Class 2	0.93	0.90	0.91	100
Class 3	0.85	0.86	0.86	100
Class 4	0.99	0.82	0.90	100
Class 5	0.97	0.91	0.94	100
Class 6	0.88	0.92	0.90	100
Class 7	0.95	0.97	0.96	100
Class 8	0.95	0.93	0.94	100
Class 9	0.92	0.93	0.93	100
accuracy			0.90	1000
macro avg	0.90	0.90	0.90	1000
weighted avg	0.90	0.90	0.90	1000

Confusion Matrix

ROC score

STT	Score
ROC 1	0.9910
ROC 2	0.9933
ROC 3	0.9909
ROC 4	0.9920
ROC 5	0.9908
ROC 6	0.9912
ROC 7	0.9904
ROC 8	0.9927
ROC 9	0.9896
ROC 10	0.9941

---