| 摘 要: 针对单一模型在文本分类时难以捕捉丰富语义信息的问题,提出了一种基于双通道和注意力机制的多特征融合的文本情感分类模型(MFF-DCAM)。首先,利用BERT-WWM-EXT获取文本词向量;其次,分别使用多尺度卷积神经网络和双向长短期记忆网络提取局部特征和上下文特征;再次,拼接两个特征并使用注意力机制对该特征加权;然后,使用全局平均池化突出句子整体信息;最后,使用Softmax函数进行分类。实验结果表明,MFF-DCAM模型的准确率、精确率和F1值优于8个基准模型,验证了该模型在提升文本情感分类方面的有效性。 |
| 关键词: 双通道 注意力机制 BiLSTM模型 Text-CNN模型 |
|
中图分类号:
文献标识码: A
|
| 基金项目: 新疆维吾尔自治区高校科研基金资助项目(XJEDU2024094) |
|
| A Multi-Feature Fusion of Dual-Channel and Attention Mechanism for Text Sentiment Classification Model |
|
DENG Zhiwen1, ZHANG Longjian1, LI Jinhui1, WEI Qiwu2
|
(1.Basic Teaching Research Department, Xinjiang College of Science & Technology, Bayingolian Mongolian Autonomous Prefecture 841000, China;
2.College of Culture and Media, Xinjiang College of Science & Technology, Bayingolian Mongolian Autonomous Prefecture 841000, China)
965213513@qq.com; 394463995@qq.com; 1187357069@qq.com; 1300130705@qq.com
|
| Abstract: To address the challenge of capturing rich semantic information with single models in text classification, this paper proposes a Mult-i Feature Fusion of Dua-l Channel and Attention Mechanism for Text Sentiment Classification Model (MFF-DCAM). First, BERT-WWM-EXT is employed to obtain text word vectors.Second, mult-i scale convolutional neural networks and bidirectional long shor-t term memory networks are used to extract local features and contextual features, respectively. Then, the two features are concatenated and weighted using an attention mechanism. Next, global average pooling is applied to highlight the overall sentence information. Finally,the Softmax function is used for classification. Experimental results demonstrate that the MFF-DCAM model outperforms eight baseline models in accuracy, precision, and F1-score, validating the effectiveness of this model in enhancing text sentiment classification. |
| Keywords: dual channel attention mechanism BiLSTM model Tex-t CNN model |
