An optimization method and classification system for job function standardization

Abstract

The present invention discloses an optimization method and classification system for the standardization of job functions. The method can classify and train the data by collecting ranking position information attached to data files and collecting historical information of user clicks on file data, and finally obtain a relevant training model. The accuracy of intelligent matching recommendations and faster and shorter processing time can be greatly improved through partial matching training and direct training matching. Through the optimization classification system for the standardization of job functions, decentralized and diversified job title data information can be modularized and standardized, and the corresponding data information can be integrated and processed. The information attached to the uploaded data files can be analyzed to obtain more file data historical information. This information can be combined to provide favorable assistance for data training and make intelligent matching recommendations more accurate.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent recruitment systems, and in particular to an optimization method and classification system for job function standardization. Background Art

[0002] With the rapid development of artificial intelligence technology, precise matching and personalized recommendation systems have been greatly developed and applied in various fields. How to apply them to intelligent recruitment platforms and accurately complete the precise matching and recommendation of job positions and user resumes from big data such as user data and recruitment data is an important research direction in the field of artificial intelligence.

[0003] Intelligent recruitment is an important application of matching recommendation algorithms and has been widely used in actual recruitment platforms. In intelligent recruitment algorithms, job title standardization is a very important step. The names of job titles are very decentralized and diversified. For example, software developers and software engineers represent the same job position. Too many decentralized and redundant job titles and a huge number of job titles will lead to the degradation of the performance of intelligent matching recommendation algorithms, such as lower accuracy or longer processing time.

[0004] In the intelligent recruitment system, job title standardization is an extremely important step, because the names of job titles are too scattered and diverse, and the number is huge, so job title standardization is a very necessary process. Job title standardization is a form of classification that aims to classify millions of different job positions and map them to a smaller fixed set of occupations; the present invention aims to apply a method and system for improving the existing job title standardization by collecting information on user search clicks. Summary of the invention

[0005] In view of the problems existing in the prior art, the present invention discloses an optimization method for job function standardization, comprising the following steps:

[0006] Step 1: Provide a standardized template composed of models with associated relationships. Users fill in standardized job titles according to the template and create a data file;

[0007] Step 2: Find and remove the "level" term in the data file name, find and correct the starting words and qualification phrases of the job title, delete the information of "part-time" and "full-time", and standardize the job titles in the data file;

[0008] Step 3: Obtain the information filled in by the user and the information of the user's search and click in the data file, and attach the position information of the relevant ranking, remove the keywords stripped in step 2 and the "qualification" information in the job title, replace the keywords or replace the job title according to the database information, generate a relative click threshold according to the relevant position ranking information, and combine the user's historical click data to form training data;

[0009] Step 4: The job titles of the clicks on the binarized training data are calculated using the Bag of words algorithm, and the possibility of each n-gram is calculated, the number of occurrences is calculated, and the final feature value is obtained. The data is partially matched and directly trained and matched through the Naive Bayes classifier to finally obtain the training model. The Naive Bayes algorithm formula is: P(A|B1,B2)=P(B1,B2|A)*P(A) / P(B1,B2)=P(B1|A)*P(B2|A)*P(A) / P(B1,B2), P(standardized job title|feature)=P(feature|standardized job title)*P(standardized job title) / P(feature).

[0010] As a preferred technical solution of the present invention, the step 1 uses Chinese regular expressions to achieve regular Chinese matching, and the regular expression list maps words and phrases to a standardized job title data file.

[0011] As a preferred technical solution of the present invention, the step two, starting word: a certain "level" is obtained only when the title begins with "assistant", "senior", "senior", etc., qualification phrase: different levels include "director", "manager", "project leader", "project person in charge", and "fresh graduate".

[0012] As a preferred technical solution of the present invention, the step 4 Naive Bayes is a method for constructing a classifier. The classifier model will assign a class label represented by a feature value to the problem instance. The class label is taken from a finite set. The Naive Bayes classifier can estimate the necessary parameters, including the mean and variance of the variable, based on a small amount of training data;

[0013] The probability model classification calculation formula is: p(C|F1,...,F n );

[0014] Bayes' theorem has the following formula: In plain language it can be expressed as:

[0015] As a preferred technical solution of the present invention, the fourth step of partial matching includes the following steps:

[0016] 1): Find the longest standardized name of the job title, which is the "sub-name" of the original job title;

[0017] 2): The number of names of standardized work titles is less than or equal to 2;

[0018] 3): If multiple standardized job title parts match the original job title with the same number of words, then the one with the highest Naive Bayes is selected.

[0019] As a preferred technical solution of the present invention, the step 4 directly trains matching: obtaining the standardized number of clicks from the original job title to the standardized job title, and selecting the standardized job title with the highest number of clicks;

[0020] The calculation formula is: Where A(i,j) is the absolute clicks of the original job title i and the standardized job title j, and a(i,j) is the standardized clicks;

[0021] If the standardized hits and their associated parts are above some specific threshold, then this standardized job title is selected;

[0022] The calculation formula is:

[0023] As a preferred technical solution of the optimization classification system for the standardization of job functions of the present invention, it includes: a standardized template module, a data file job title correction module, a data file collection module, a training data formation module and a calculation module;

[0024] The standardized template module will be compiled according to the standardized job title data filled in by the user to obtain a data file for uploading;

[0025] The data file job title correction module searches and corrects the uploaded data file names, deletes other irrelevant information, and standardizes the job titles in the data files;

[0026] The data file collection module obtains the data information in the standard data file, collects the position ranking information attached to the data file, and collects the historical information of the user clicking the file data;

[0027] The training data formation module replaces the keywords in the collected data files and the information about "qualifications" in the job titles with the data information in the database and combines them with the collected related position ranking information and historical click data to form training data;

[0028] The calculation module uses the Bag of words algorithm to calculate the number of times the job title appears in the training data and form the feature value. The naive Bayes classifier is used to perform partial matching and direct training matching on the data to finally obtain the training model.

[0029] As a preferred technical solution of the present invention, it also includes a data storage and a data processor, the data processor and the data storage communicate with each other through a bus, and the data storage stores program instructions that can be executed by the data processor.

[0030] As a preferred technical solution of the present invention, during the data file uploading process, the data processor merges and imports the content, creation time, generation time, upload time, and creator information of the data file. After execution, the data file acquisition module can obtain the corresponding historical data information.

[0031] Beneficial effects of the present invention: The present invention proposes an optimization method for the standardization of job functions, which can be used for the standardized classification of job titles. The improved classification method is based on the results of naive Bayes classification, and then partial matching and direct training matching are performed. At the same time, the data can be classified and trained by collecting the ranking position information attached to the data file and collecting the historical information of the user clicking on the file data, and finally a relevant training model is obtained. Through partial matching training and direct training matching, the accuracy of intelligent matching recommendations can be greatly improved and the processing time can be faster and shorter. Through the standardized optimization classification system of job functions, the decentralized and diversified job title data information can be modularized, standardized, and the corresponding data information can be integrated and processed. In addition, the information attached to the uploaded data file can be analyzed to obtain more file data history information, and this information can be combined to provide favorable assistance for data training, making the intelligent matching recommendation more accurate. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] Figure 1 A flow chart of the standardized classification system for the job functions of the present invention;

[0033] Figure 2 The present invention provides a flow chart of the method for standardizing and optimizing the working functions. DETAILED DESCRIPTION

[0034] Example 1

[0035] like Figure 1 to Figure 2 As shown, the present invention discloses an optimization method and classification system for job function standardization, and the technical solution adopted is, including the following steps:

[0036] Step 1: Provide a standardized template composed of models with associated relationships. Users fill in standardized job titles according to the template and create a data file;

[0037] Step 2: Find and remove the "level" term in the data file name, find and correct the starting words and qualification phrases of the job title, delete the information of "part-time" and "full-time", and standardize the job titles in the data file;

[0038] Step 3: Obtain the information filled in by the user and the information of the user's search and click in the data file, and attach the position information of the relevant ranking, remove the keywords stripped in step 2 and the "qualification" information in the job title, replace the keywords or replace the job title according to the database information, generate a relative click threshold according to the relevant position ranking information, and combine the user's historical click data to form training data;

[0039] Step 4: The job titles of the clicks on the binarized training data are calculated using the Bag of words algorithm, and the possibility of each n-gram is calculated, the number of occurrences is calculated, and the final feature value is obtained. The data is partially matched and directly trained and matched through the Naive Bayes classifier to finally obtain the training model. The Naive Bayes algorithm formula is: P(A|B1,B2)=P(B1,B2|A)*P(A) / P(B1,B2)=P(B1|A)*P(B2|A)*P(A) / P(B1,B2), P(standardized job title|feature)=P(feature|standardized job title)*P(standardized job title) / P(feature).

[0040] The step 1 uses Chinese regular expressions to achieve regular Chinese matching, and the regular expression list maps words and phrases to a standardized job title data file.

[0041] Step 2, starting words: a certain "level" is obtained only when the title begins with "assistant", "senior", "senior", etc. Qualification phrases: different levels include "director", "manager", "project leader", "project director", and "fresh graduate".

[0042] The step 4 Naive Bayes is a method for constructing a classifier. The classifier model assigns a class label represented by a feature value to the problem instance. The class label is taken from a finite set. The Naive Bayes classifier can estimate the necessary parameters, including the mean and variance of the variable, based on a small amount of training data.

[0043] The probability model classification calculation formula is: p(C|F1,...,F n );

[0044] Bayes' theorem has the following formula:

[0045] In plain language it can be expressed as:

[0046] The step 4 partial matching includes the following steps:

[0047] 1): Find the longest standardized name of the job title, which is the "sub-name" of the original job title;

[0048] 2): The number of names of standardized work titles is less than or equal to 2;

[0049] 3): If multiple standardized job title parts match the original job title with the same number of words, then the one with the highest Naive Bayes is selected.

[0050] The step 4 directly trains the matching: obtaining the standardized number of clicks from the original job title to the standardized job title, and selecting the standardized job title with the highest number of clicks;

[0051] The calculation formula is:

[0052] Where A(i,j) is the absolute clicks of the original job title i and the standardized job title j, and a(i,j) is the standardized clicks;

[0053] If the standardized hits and their associated parts are above some specific threshold, then this standardized job title is selected;

[0054] The calculation formula is:

[0055] A technical solution for optimizing the classification system for standardized job functions, comprising: a standardized template module, a data file job title correction module, a data file collection module, a training data formation module and a calculation module;

[0056] The standardized template module will be compiled according to the standardized job title data filled in by the user to obtain a data file for uploading;

[0057] The data file job title correction module searches and corrects the uploaded data file names, deletes other irrelevant information, and standardizes the job titles in the data files;

[0058] The data file collection module obtains the data information in the standard data file, collects the position ranking information attached to the data file, and collects the historical information of the user clicking the file data;

[0059] The training data formation module replaces the keywords in the collected data files and the information about "qualifications" in the job titles with the data information in the database and combines them with the collected related position ranking information and historical click data to form training data;

[0060] The calculation module uses the Bag of words algorithm to calculate the number of times the job title appears in the training data and form the feature value. The naive Bayes classifier is used to perform partial matching and direct training matching on the data to finally obtain the training model.

[0061] It also includes a data storage and a data processor. The data processor and the data storage communicate with each other through a bus. The data storage stores program instructions that can be executed by the data processor.

[0062] During the data file upload process, the data processor merges and imports the content, creation time, generation time, upload time, and creator information of the data file. After the execution is completed, the data file acquisition module can obtain the corresponding historical data information.

[0063] Components not described in detail herein are prior art.

[0064] Although the specific embodiments of the present invention are described in detail above, the present invention is not limited to the above embodiments. Various changes can be made within the knowledge of ordinary technicians in this field without departing from the purpose of the present invention, and modifications or deformations without creative labor are still within the protection scope of the present invention.

Claims

1. A method for optimizing job function standardization, characterized in that: The following steps are involved: Step 1: Provide a standardized template composed of models with associated relationships. Users fill in standardized job titles according to the template and create a data file; Step 2: Find and remove the "level" term in the data file name, find and correct the starting words and qualification phrases of the job title, delete the information of "part-time" and "full-time", and standardize the job titles in the data file; Step 3: Obtain the information filled in by the user and the information of the user's search and click in the data file, and attach the position information of the relevant ranking, remove the keywords stripped in step 2 and the "qualification" information in the job title, replace the keywords or replace the job title according to the database information, generate a relative click threshold according to the relevant position ranking information, and combine the user's historical click data to form training data; Step 4: The job titles of the clicks after binarization of the training data are calculated using the Bag of words algorithm, and the probability of each n-gram is calculated, the number of occurrences is calculated, and the final feature value is obtained. The data is partially matched and directly trained and matched through the naive Bayes classifier. The partial matching includes the following steps: 1): Find the longest standardized name of the job title, which is the "sub-name" of the original job title; 2): The number of names of standardized work titles is less than or equal to 2; 3): If multiple standardized job titles partially match the original job title with the same number of words, then the one with the highest Naive Bayes is selected; The direct training matching: obtaining the standardized number of clicks from the original job title to the standardized job title, and selecting the standardized job title with the highest number of clicks; The calculation formula is: Where A(i,j) is the absolute clicks of the original job title i and the standardized job title j, and a(i,j) is the standardized clicks; If the standardized hits and their associated parts are above some specific threshold, then this standardized job title is selected; The calculation formula is: Finally, the training model is obtained, and the formula of the Naive Bayes algorithm is: P(A|B1,B2)=P(B1,B2|A)*P(A) / P(B1,B2)=P(B1|A)*P(B2|A)*P(A) / P(B1,B2), P(standardized job title|features)=P(features|standardized job title)*P(standardized job title) / P(features).

2. The method for optimizing the standardization of job functions according to claim 1, characterized in that: The step 1 uses Chinese regular expressions to achieve regular Chinese matching, and the regular expression list maps words and phrases to a standardized job title data file.

3. The method for optimizing the standardization of job functions according to claim 1, characterized in that: The second step, starting words: a certain "level" is obtained only when the title begins with "assistant", "senior", "senior", etc., qualification phrases: different levels include "director", "manager", "project leader", "project director", "fresh graduate".

4. The method for optimizing the standardization of job functions according to claim 1, characterized in that: The step 4 Naive Bayes is a method for constructing a classifier. The classifier model assigns a class label represented by a feature value to the problem instance. The class label is taken from a finite set. The Naive Bayes classifier can estimate the necessary parameters, including the mean and variance of the variable, based on a small amount of training data. The probability model classification calculation formula is: p(C|F1,...,F n ); Bayes' theorem has the following formula: In plain language it can be expressed as:

5. A standardized and optimized classification system for job functions, used to implement the method as claimed in claim 1, characterized in that: include: Standardization template module, data file job title correction module, data file acquisition module, training data formation module and calculation module; The standardized template module will be compiled according to the standardized job title data filled in by the user to obtain a data file for uploading; The data file job title correction module searches and corrects the uploaded data file names, deletes other irrelevant information, and standardizes the job titles in the data files; The data file collection module obtains the data information in the standard data file, collects the position ranking information attached to the data file, and collects the historical information of the user clicking the file data; The training data formation module replaces the keywords in the collected data files and the information about "qualifications" in the job titles with the data information in the database and combines them with the collected related position ranking information and historical click data to form training data; The calculation module uses the Bag of words algorithm to calculate the number of times the job title appears in the training data and form the feature value. The naive Bayes classifier is used to perform partial matching and direct training matching on the data to finally obtain the training model.

6. The system according to claim 5, characterized in that It also includes a data storage and a data processor. The data processor and the data storage communicate with each other through a bus. The data storage stores program instructions that can be executed by the data processor.

7. The system according to claim 5 or 6, characterized in that: During the data file upload process, the data processor merges and imports the content, creation time, generation time, upload time, and creator information of the data file. After the execution is completed, the data file acquisition module can obtain the corresponding historical data information.

Images