基于深度学习的端到端人岗匹配模型

End to end person-job matching model based on deep learning

针对现有人岗匹配推荐算法主要采用人工评估求职者与职位的匹配度,存在招聘速度慢、成本高且易受主观判断所误导等问题,提出一种基于深度学习的端到端人岗匹配模型BATPJF。首先,运用TextCNN提取简历和职位描述数据的局部特征。同时,运用BiLSTM提取简历和职位描述文本数据的上下文特征,再将BiLSTM隐藏层产生的特征作为Attention层的输入,利用注意力机制对BiLSTM层提取的特征采用加权的方式体现不同的经历和能力对岗位能力需求重要程度的影响。然后,将2种模型提取到的特征进行融合。最后,通过全连接层进行预测。实验结果表明,与其他5种人岗匹配模型对比,本文提出的模型可以更有效地匹配工作要求和简历文本信息。

Aiming at the problems of low recruitment efficiency,high cost and being easily misled by subjective judgment,the existing person position matching recommendation algorithm mainly uses manual evaluation of the matching degree between job seekers and positions,and proposes an end-to-end person-job matching model BATPJF based on deep learning.First of all,the paper uses TextCNN to extract the local features of the resume and job description data.At the same time,BiLSTM is used to extract the contextual features of resume and job description text data,and then the features generated by the BiLSTM hidden layer are used as the input of the Attention layer.The attention mechanism is used to weigh the features extracted by the BiLSTM layer to reflect the impact of different experiences and abilities on the importance of job competency requirements.Secondly,the features extracted from the two models are fused.Finally,the prediction is conducted through the complete connection layer.The experimental results show that the model proposed in this paper can match job requirements and resume text information more effectively than the other five person-job matching models.