Dual-channel feature fusion network for sheep diseases question classification

Area of Science:

• Veterinary Medicine
• Artificial Intelligence
• Natural Language Processing

Background:

• Sheep disease diagnosis relies on accurate question classification.
• Existing methods struggle with feature sparsity, semantic ambiguity, and insufficient feature extraction.

Purpose of the Study:

• To propose a novel Dual-Channel Feature Fusion Network for Sheep Diseases Question Classification (DFF-SDQC).
• To enhance semantic representation and feature extraction for improved classification accuracy.

Main Methods:

• Utilized CINO pre-trained model for dynamic word embeddings.
• Employed BiLSTM for global textual features and an attention mechanism for local contextual features.
• Introduced a question-word attention mechanism for better intent capture and fused dual-channel features.

Main Results:

• The DFF-SDQC model achieved an F1-score of 93.18% on the D-SDQC dataset.
• Demonstrated a 2.22 percentage point improvement over the strongest baseline.
• Validated the effectiveness of dual-channel fusion and the attention mechanism design.

Conclusions:

• The DFF-SDQC model significantly improves sheep disease question classification.
• The proposed model effectively addresses challenges in feature representation and semantic ambiguity.
• Dual-channel fusion and attention mechanisms are crucial for robust and accurate classification.