【摘要】 简历文本作为体现人才能力信息的载体，面对智能制造机械行业人才能力需求向多元化变化，单一能力属性的人才筛选存在人才遗漏或难以满足企业需求问题，因此对简历进行多标签分类更有价值。提出了一种融合卷积神经网络（Convolutional Neural Networks, CNN）的注意力机制序列模型（Sequence to Sequence, Seq2Seq）方法将多标签分类问题转化为序列输出问题，实现多标签简历行业分类。首先，对简历文本进行分词与停用词去除，并经过词嵌入操作获得简历文本的向量表示。然后，使用单层多卷积核的卷积网络进行初步局部文本特征提取，将提取的词向量特征输入编码器，经过双向门控循环（Bidirectional Gate Recurrent Unit, Bi-GRU）获得全文向量编码，并在解码过程使用注意力机制关注不同输入时刻上的重点内容，输出所属行业标签。在最后的实验验证中，该方法模型的F₁值达到0.841,证明融入CNN后的模型效果更好。更多还原

【Abstract】 Resume is the carrier of talent ability information.In the face of the diversified demand for talents in the intelligent manufacturing machinery industry, there are problems with talent omission or difficulty in meeting the needs of enterprises in talent selection with a single competency attribute, therefore, multi label classification of resumes is more valuable.A CNN integrated attention mechanism Seq2Seq method is proposed to transform the multi label classification problem into a sequence output problem, and realize the multi label resume industry classification.First, the resume text is segmented and the stop words are removed, and the vector representation of the resume text is obtained through the word embedding operation.Then, the convolution network with single layer and multiple convolution cores is used for preliminary local text feature extraction.The extracted word vector feature is input into the encoder, and the full text vector code is obtained through Bi-GRU.In the decoding process, attention mechanism is used to focus on the key content at different input times, and the industry label is output.In the final experimental verification, the F₁ value of the proposed method model reaches 0.841,It turns out that the model after incorporating CNN works better.更多还原