【摘要】 针对现阶段小麦问答社区的问句文本存在噪声、特征稀疏以及专业性强等问题，从词和字双粒度特征出发，提出了一种基于双粒度的小麦问句分类模型.为有效缓解农业问句语义特征稀疏的问题，采用基于字粒度和词粒度的双分支架构，并引入交互注意力机制获取词粒度和字粒度交互特征信息以实现不同粒度信息表达文本语义的一致性，最后融合双粒度特征及其交互特征构建分类模型.同时，在输入层添加农业字典和加载停用词表进行分词和分字，有效解决小麦社区问句文本专业性强和数据噪声问题.与现有六种主流农业社区问句分类模型相比，该模型在整体分类性能上表现最优，且在各类别上综合性能优于其他模型.本研究有助于提高小麦种植社区问答系统的性能，并积极推动智能农业推广进程、助力乡村振兴.更多还原

【Abstract】 In response to the issues of noise, feature sparsity, and strong agricultural expertise in the question texts of the current wheat Q&A community, a wheat question classification model based on dual-granularity features is proposed. To effectively alleviate the problem of semantic feature sparsity in agricultural questions, a dual-branch architecture is employed, extracting both character-level and word-level features in separate branches. An interactive attention mechanism is introduced to capture the interaction between word-level and character-level features, ensuring the consistency of semantic representation across different granularities. Finally, the model integrates character-level, word-level, and their interactive features to construct a robust classification framework. Additionally, an agricultural dictionary and a stop words list are incorporated at the input layer to enhance word segmentation and character splitting, addressing the challenges of high domain specificity and data noise in wheat community question texts. Compared to six existing mainstream agricultural community question classification models, this model demonstrates superior overall classification performance and improved comprehensive performance across various categories. This research contributes to enhancing the performance of question-and-answer systems in wheat farming communities and actively promotes the advancement of smart agriculture, support rural revitalization efforts.更多还原