File size: 736 Bytes
877049d
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
# Appendix Code Snippets

Các đoạn mã nên đưa vào phụ lục báo cáo:

1. `src/preprocess.py`: chuẩn hóa schema, map nhãn, tạo `node_id`.
2. `src/split_data.py`: chia `train/val/test` theo `stratified split`.
3. `src/extract_embeddings.py`: trích xuất embedding PhoBERT cho toàn bộ tiêu đề.
4. `src/build_graph.py`: xây đồ thị `k-NN` từ cosine similarity.
5. `src/model_fusion.py`: mô hình fusion giữa nhánh text và nhánh graph.
6. `src/train_fusion.py`: vòng lặp train, early stopping và lưu checkpoint.

Có thể đặt nhãn minh họa theo dạng:

- `Listing A.1 Data preprocessing`
- `Listing A.2 Graph construction`
- `Listing A.3 Fusion model`
- `Listing A.4 Training procedure`