[Medical text classification model integrating medical entity label semantics].

Area of Science:

• Natural Language Processing
• Artificial Intelligence
• Medical Informatics

Context:

• Online medical services require efficient and accurate classification of user queries.
• Intent recognition is crucial for understanding user needs in healthcare.
• Existing datasets often lack essential entity annotations, hindering model development.

Purpose:

• To propose a novel medical text classification model, BRELS (bidirectional encoder representation based on transformer-recurrent convolutional neural network-entity-label-semantics).
• To integrate medical entity label semantics for enhanced intent recognition.
• To address the limitations of existing datasets by incorporating prior knowledge of medical entity labels.

Summary:

• The BRELS model employs an adaptive fusion mechanism for local feature enhancement and a lightweight recurrent convolutional neural network (LRCNN) for global feature extraction.
• This approach effectively preserves text semantics while managing parameter growth.
• Experiments on three public datasets demonstrate the model's superior performance, achieving F1 scores of 87.34%, 81.71%, and 77.74%.

Impact:

• The BRELS model significantly improves the identification and understanding of medical terminology.
• This leads to more accurate user intent recognition, enhancing the quality and efficiency of online medical services.
• Provides a robust solution for automated medical question classification in digital health platforms.