TUSNLP at the NTCIR-18 RadNLP Task: Explainable Classification Approach by Domain Knowledge-Based Bag-of-Words

Key Points

• To develop interpretable classification models for lung cancer staging using Bag-of-Words representations based on domain knowledge.
• Developed classification models using Bag-of-Words for lung cancer stage.
• Utilized predefined key terms derived from medical domain knowledge for model training.
• Employed techniques to assess model interpretability and accuracy.
• Models showed high medical validity in classifying lung cancer stages.
• Demonstrated improved accuracy through the integration of domain knowledge.
• Provided valuable clinical insights into lung cancer staging.