Applied Research Of Deep Learning On Resume Analysis

With the rapid growth of text information data over the Internet,the massive freeformat text resumes bring us much convenience.Meanwhile,challenges and difficulties arise due to information overload.In order to improve the performance of analyzing unstructured text resumes,this thesis applies deep learning techniques to resume analysis.The main work is as follows.1.Research of Chinese resume analysis based on character sequence.To solve the defects of word embedding generated by shallow neural networks,a Bidirectional Long Short-Term Memory(BLSTM)neural network is employed to model the character sequences and obtain the corresponding internal features of words.Furthermore,the BLSTM combined with Conditional Random Fields(CRF)(BLSTM-CRF)is utilized to model the generated word embedding and tune the model.Finally,the unstructured resumes are analyzed by the trained model.The experimental results show that our method is superior to the resume analysis model of traditional word embedding scheme,and the F1-score is improved by 2.31%.2.Resarch of Chinese resume analysis based on feature fusion.It is difficult to improve the performance of resume analysis model by using a single feature,and a scheme of integrating multiple effective features to improve the performance of resume analysis model is proposed.The method combines the semantic features generated by traditional shallow neural networks and the character sequence features generated by BLSTM.(1)The concat method is used to fuse the two features,and then,the BLSTMCRF is utilized to model the fused features and tune the model.Finally,the unstructured resumes are analyzed by the trained model.The experimental results show that the F1-score of this method is increased by 3.27% and 0.96% respectively compared with the resume analysis model of the traditional word embedding scheme and the character sequence scheme.(2)The attention mechanism is introduced to fuse the above two features,which is also introduced into BLSTM-CRF model.Finally,the BLSTM-CRF model based on the attention mechanism is utilized to analyze the unstructured resume.The experimental results show that compared with the resume analysis model of traditional word embedding scheme,character sequence scheme and concat feature fusion scheme,the F1-score of the proposed method is improved by 6.39%,4.39% and 3.43% respectively.3.Design of Chinese resume analysis system based on deep learning.A Chinese reume analysis system is designed on the basis of the previous experiments and theories.The results of the resume analysis are displayed on the web page and applied to the knowledge map of the characters and companies.