Crop Pest and Disease Question Classification Method Based on Self-Attention Mechanism
Abstract：In crop pest and disease intelligent question-answering systems, question classification is a key technical link for efficient knowledge retrieval and accurate answer generation. To address the issues of long training time and weak ability to learn text context information in the Bidirectional Long Short-Term Memory (BiLSTM) network, this paper proposes a crop pest and disease question classification method based on the self-attention mechanism. First, a BERT pre-trained model built on an agricultural domain corpus is used to vectorize the text data. Then, a Bidirectional Gated Recurrent Unit (BiGRU) is utilized to extract deep features of the text. Next, a self-attention mechanism layer is applied to assign corresponding weights to the extracted deep text information. Finally, the text features with different weights are fed into a Softmax classifier for crop pest and disease question classification. Experimental results show that the proposed method outperforms traditional methods such as template matching, SVM, and CNN-LSTM, and can provide technical support for crop pest and disease intelligent question-answering systems.
Keywords：Crop Pest and Disease QuestionBERTBidirectional Gated Recurrent Unit (BiGRU)Self-Attention Mechanism
Publication Date：2025-10-25
Online Publishing Date：2025-11-14(First online date of this platform, not the publication date of the document)
Pages：5( 14-18 )