Research And Application Of Resume Text Analysis

With the continuous development of the social economy,the needs of enterprises for talents are gradually increasing,which in turn has inspired more and more job seekers to submit their resumes through various channels to let companies understand themselves more specifically.During the peak recruitment period(autumn and spring recruitment in campus recruitment,mainly for graduates),companies will receive a large number of resumes from different channels.In this case,how can the staff responsible for recruitment in the company be able to quickly finding the resume you want becomes an urgent problem to be solved in the recruitment scenario.In order to realize the rapid identification of recruitment resumes,this article uses a system provided by a company to test resumes and obtain resume text data.Because the resumes are semi-structured resumes,and the format restrictions of different resume templates and the personal writing habits of applicants are different,the extracted text data often have large format differences,and at the same time,considering the data requirements of the models in the following text,this article In the data preprocessing stage,text regularization processing,text word segmentation processing,special character processing,and word segmentation processing were performed on the text data.The word segmentation processing is mainly implemented by jieba word segmentation,so as to ensure that the processed data can be successfully trained in the model.After preprocessing the resume text data,the text data is converted into word vectors through label dictionary,sentence filling,and word matrix construction,and finally the model is trained.This article is mainly based on Tensor Flow,using convolutional neural networks and long-short-term memory networks in recurrent neural networks to construct a hybrid classification model CNN-LSTM model for text analysis.In the model training stage,the model parameters are continuously adjusted through optimization methods such as back propagation and random discarding,so that the model classification effect is optimal.Through training in this article,the final accuracy rate reaches 94.02%.Based on the classification results of the resume text,this paper proposes specific applications in three recruitment scenarios,which are extraction of simplified resume,automatic labeling(skill label,school type label,academic label)and post matching calculation.The information in the streamlined resume is all the information of the applicants concerned by the company,and the irrelevant information will not be stored in the streamlined resume;automatic skill labeling can help HR quickly locate the applicants who meet their requirements;job matching degree calculation Matching is required,and the matching degree is given.The highest matching degree is the candidate who best meets the requirements of the enterprise.