Research On The Application Of Bidirectional Recommendation Method In Enterprise Recruitment

With the expansion of network coverage,the progress of software and hardware technology and the improvement of government policies and regulations on network recruitment,the traditional offline recruitment shows low coverage,poor efficiency and high cost defects.At the same time,the recognition of online recruitment in net citizens and enterprises has been greatly improved.In addition,it has the advantages of low cost,unlimited,high efficiency and large amount of information,network recruitment has become the mainstream channel of recruitment.Many online recruitment websites,such as 51 job,Zhaopin and Chinahr,provide an information exchange platform for job seekers and enterprises to reach mutual choice,which greatly promoted the exchange of information between the recruiting and the applicants,which offset some of the recruitment pressure.However,there are still some aspects of recruitment that need to be optimized and improved.For example,the electronic resume on the website cannot be automatically analyzed and recommended,it requires manual screening by enterprises or headhunters,and the delivery process cannot be tracked,the time to get feedback is too long,and there are few websites that implement the two-way recommendation function of individual and enterprise.This paper mainly solves the problems of automatic extraction of semi-structured text information and two-way recommendation of personnel and posts.By analyzing the structure and content of the two semi-structured texts,job resumes and job announcements,it is found that although they differ in grammatical structure and content organization order according to the different grammar and vocabulary usage habits of each person,some commonalities and rules can still be found to support information extraction.Because different files have different formats,firstly,the format of the files needs to be differentiated,and the files of different formats are batch-converted by the format conversion tool to obtain a unified set of text documents,and then read through the r language for data pre-processing.Then,the text is hierarchically divided.By drawing the hierarchical structure diagram,thegeneral modules of the two texts and the subdivision attribute relationship are clarified,the module labels are extracted according to the rules,and then the texts of the corresponding classes are clustered by using the k-means method.The block is extracted and saved as the original text of the module.Among them,the extraction method of simple information is mainly to introduce all kinds of dictionaries in the process of word segmentation,such as professional subject catalogue,national college directory and name dictionary,and so on.The specific words are identified by self-defined part of speech,and the information is extracted according to the rules.For complex information,it mainly includes the extraction and analysis of two free texts,such as work content and job responsibilities.According to the steps of word segmentation,denoising,and similar words,the text is represented by keyword sequence,and then the text similarity comparison is completed.In order to solve the problem of two-way recommendation of personnel posts and optimize the website recommendation function,the article uses the content-based two-way recommendation method to recommend to both the staff and the enterprise,and matches the preferences in the resume with the post-related attributes to calculate the enterprise positions to satisfy the job seekers.The degree of preference,matching the post's personnel preferences with the relevant attributes in the resume to calculate the degree to which the job seeker satisfies the post preference,and selecting the feature texts that are more important to the subject,including the two free text segments of job responsibilities and job content.First of all,the weighted keyword is calculated by the TFIDF algorithm after the word segmentation,and then the VSM model is used to map the free text to the feature vector composed of keywords to calculate the similarity between the vectors.Finally,the recommendation list is obtained according to the comprehensive result of the text matching degree.Complete two-way recommendations for personnel and positions.The information extraction method used in this paper has a recognition rate of95% for module tags,and the accuracy of extracting simple texts such as name,gender,and place name is high.The result of extracting free texts such as job responsibilities and work content is subject to the module tag set.The effect will appear redundant text,but as the richness and accuracy of the label set will increaseaccordingly,the de-stopping and feature dimension reduction process for keywords reduces the computational complexity in time and space.Most of the recruitment websites only match the job title and certificate skills filled in by the user,and the content-based two-way recommendation method used in this article also introduces text segments such as job responsibilities and work content,which can help the recruitment website optimize the recommendation results,thereby improving user experience.