WO2020253503A1 - 人才画像的生成方法、装置、设备及存储介质 - Google Patents

人才画像的生成方法、装置、设备及存储介质 Download PDF

Info

Publication number
WO2020253503A1
WO2020253503A1 PCT/CN2020/093409 CN2020093409W WO2020253503A1 WO 2020253503 A1 WO2020253503 A1 WO 2020253503A1 CN 2020093409 W CN2020093409 W CN 2020093409W WO 2020253503 A1 WO2020253503 A1 WO 2020253503A1
Authority
WO
WIPO (PCT)
Prior art keywords
target
talent
text
training model
portrait
Prior art date
Application number
PCT/CN2020/093409
Other languages
English (en)
French (fr)
Inventor
马琳
Original Assignee
平安科技（深圳）有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 平安科技（深圳）有限公司 filed Critical 平安科技（深圳）有限公司
Publication of WO2020253503A1 publication Critical patent/WO2020253503A1/zh

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
    • G06F16/33—Querying
    • G06F16/335—Filtering based on additional data, e.g. user or group profiles
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
    • G06F16/35—Clustering; Classification

Definitions

  • This application relates to the field of artificial intelligence, and in particular to a method, device, equipment and storage medium for generating talent portraits.
  • HR human resources
  • This application provides a method, device, equipment and storage medium for generating a talent portrait, which are used to refine the core skills in a resume and generate a talent portrait, which is convenient for recruiters to view and improves the efficiency of resume screening.
  • the first aspect of the embodiments of this application provides a method for generating a talent portrait, including: obtaining a target resume; extracting resume data from the target resume; and cleaning the resume data based on a text classification algorithm based on natural language processing NLP, Obtain the target text, which indicates the core skills of the target talent; call a preset training model to analyze the target text, and generate prediction labels, each predicted label corresponds to a core skill of the target talent; The prediction tag generates a talent portrait of the target talent.
  • the second aspect of the embodiments of this application provides a device for generating a talent portrait, including: an acquisition unit for acquiring a target resume; an extraction unit for extracting resume data from the target resume; and a cleaning unit for The natural language processing NLP text classification algorithm cleans the resume data to obtain the target text, which indicates the core skills of the target talent; the first generating unit is used to call a preset training model to perform the target text Analyze and generate predicted labels, each predicted label corresponds to a core skill of the target talent; the second generating unit is used to generate a talent portrait of the target talent according to the predicted label.
  • the third aspect of the embodiments of the present application provides a device for generating a portrait of a talent, including a memory, a processor, and computer-readable instructions stored on the memory and running on the processor, and the processor executes
  • the computer-readable instruction implements the method for generating the talent portrait described in any of the foregoing embodiments.
  • the fourth aspect of the embodiments of the present application provides a computer-readable storage medium, including instructions, which when run on a computer, cause the computer to execute the steps of the talent portrait generation method described in any of the foregoing embodiments.
  • the target resume is obtained; resume data is extracted from the target resume; the resume data is cleaned based on the text classification algorithm of natural language processing NLP to obtain the target text, which indicates the core skills of the target talent; Call the preset training model to analyze the target text and generate prediction labels.
  • Each prediction label corresponds to a core skill of the target talent; according to the prediction label, a talent portrait of the target talent is generated.
  • the embodiment of the present invention refines the core skills in the resume to generate a talent portrait, which is convenient for recruiters to check and improves the efficiency of resume screening.
  • FIG. 1 is a schematic diagram of an embodiment of a method for generating a talent portrait in an embodiment of the application
  • FIG. 2 is a schematic diagram of another embodiment of the method for generating a talent portrait in an embodiment of the application
  • Fig. 3 is a schematic diagram of an embodiment of a device for generating a talent portrait in an embodiment of the application
  • FIG. 4 is a schematic diagram of another embodiment of the device for generating a talent portrait in an embodiment of the application
  • Fig. 5 is a schematic diagram of an embodiment of a device for generating a talent portrait in an embodiment of the application.
  • FIG. 1 a flowchart of a method for generating a talent portrait provided by an embodiment of the present application, which specifically includes:
  • the terminal obtains the target resume. Among them, the terminal selects a resume that meets the requirements from a large number of resumes according to preset screening conditions.
  • the target resume here can be one or multiple. For ease of description, this application uses one target resume as an example.
  • the terminal obtains a large number of resumes according to actual conditions, and filters out target resumes that meet the basic requirements, where the target resumes can be different types of resumes.
  • the format of the target resume can be adjusted for different groups.
  • the target resume can include: 1. Basic personal information, which can include name, age, gender, ethnicity, contact number, hometown, etc.; 2. Education background, Educational background can include middle school education experience and university education experience. Among them, middle school education experience includes school name and length of study, university education experience includes school name, length of study, and major of study; 3. Major courses, major courses can include English , Advanced Mathematics, Linear Algebra, College Physics, etc. It can also include other courses such as Building Mechanics, Material Mechanics, Contract Management, etc. The specifics are not limited here; 4. Self-evaluation, self-evaluation includes hobbies, etc.
  • the target resume can include: 1. Basic personal information, which can include name, age, gender, ethnicity, contact number, hometown, etc.; 2. Educational background, education The background can include middle school education experience and university education experience. Among them, the middle school education experience includes the name of the school and the length of study, and the university education experience includes the name of the school, the length of study, and the major of study; 3. Work experience, which can include the name of the work unit, Working hours, work content and nature of the work unit, etc.; 4. Self-evaluation, including hobbies, work attitude, etc. 5. Honorary awards. Honorary awards include study awards, work awards, and government awards. The specifics are not limited here. 6. Master the language. The master language can include Chinese, English, French, Cantonese, etc., which is not limited here.
  • resumes are suitable for different groups, and the resume content contained in the target resume is also different.
  • the resume can also be customized, and part or all of the content contained in the resume of fresh graduates and social personnel can be included in the resume. The details are not limited here.
  • the execution subject of the embodiment of the present application may be a device for generating a talent portrait, or may be a terminal or a server, which is not specifically limited here.
  • This application takes the terminal as the execution subject as an example for description.
  • the terminal extracts resume data from the target resume. Specifically, the terminal extracts resume data from the target resume.
  • the resume data includes basic information and professional information of the target talent.
  • the basic information includes name, age, gender, contact information, etc.
  • the professional information includes work unit, job Years, graduate school, graduate major, certificates obtained, professional skills, job position, etc.
  • the resume data is the various options included in the above steps.
  • the extracted resume data includes name: Li Yi; age: 18; gender: female; contact information: 151XXXXXX54, and can also include other Information, such as graduate school: Massachusetts Institute of Technology, etc. The specifics are not limited here.
  • a text classification algorithm based on natural language processing NLP cleans resume data to obtain target text, which indicates the core skills of the target talent.
  • the terminal cleans the resume data based on the text classification algorithm of natural language processing (NLP) to obtain the target text, which indicates the core skills of the target talent.
  • NLP natural language processing
  • the terminal invokes a convolutional neural network CNN algorithm (TextCNN) or a logistic regression (logistic regression, LR) algorithm to perform text classification on resume data.
  • CNN convolutional neural network CNN algorithm
  • LR logistic regression
  • the features used in text classification are the words that best represent the classification.
  • resume data can be divided into two categories, namely valid data and invalid data. Among them, valid data can include work unit, name, age, gender, etc., and invalid data can include application position, working years, etc., which is not done here. limited.
  • the text classification algorithm of NLP is a preset convolutional neural network CNN algorithm or LR algorithm. According to actual needs, the resume data is classified and then filtered, invalid data is deleted, and valid data is retained. There are target talents in the valid data. Core skill information, for example, good at English, etc., that is, effective data "good at English" as the target text.
  • text classification can also be part-of-speech tagging, sentence segmentation, recognition of dialogue behavior types, and recognition of textual implication.
  • sentence segmentation e.g., sentence segmentation
  • recognition of dialogue behavior types e.g., dialogue behavior types
  • recognition of textual implication
  • the terminal calls a preset training model to analyze the target text and generates prediction labels, each of which corresponds to a core skill of the target talent. Specifically, the terminal sequentially traverses each piece of application information in the target text, and each piece of application information corresponds to a core skill of the target talent; the terminal performs word segmentation on each piece of application information based on a preset training model to obtain the word segmentation result; the terminal according to the word segmentation The result is searched in the business word bag to obtain the business keywords of each application information; the terminal arranges the business keywords in descending order of frequency to obtain the sorting result; the terminal determines the top preset number of keywords in the sorting result as the predicted label .
  • label value of the abstract classification and generalization of a certain characteristic of a certain type of specific group or object is classifiable.
  • label value can be divided into “male”, “female” and “unknown” according to the enumeration analysis (mutually exclusive collectively exhaustive, MECE) principle;
  • label “age” its label value can be divided into “ 0-18”, “18-35”, “35-60”, “60-100”, etc., which are not specifically limited here.
  • the terminal generates a talent portrait of the target talent based on the predicted label.
  • the terminal creates a target class, the target class includes multiple attributes; the terminal determines each attribute in the target class according to the prediction label, and each attribute corresponds to a prediction label; the terminal determines the output result of the target class, and the output result includes multiple predictions Label: The terminal generates a talent portrait of the target talent according to the label value of each predicted label in the output result.
  • Talent portraits can be simply understood as labels for massive amounts of data. According to the differences in users' goals, behaviors and opinions, they are divided into different types, and then typical characteristics are extracted from each type, and given names, photos, and some demographics. The description of elements, scenes, etc. forms a personas.
  • the obtained target talent portrait can include age, gender, working years, core skills, job search expectations, and so on.
  • the talent portrait obtained by refining A resume can be: Internet, married, corporate executives, high income.
  • the talent portrait can also be other tags, which are not limited here.
  • the definition of the target talent based on the characteristic value helps Human Resource (HR) to grasp the characteristics of the target talent at a glance, such as "working in computer development for many years", “excellent technology” “Etc., you can quickly understand that the target talent is good at working in the computer field, so HR can judge whether it is the talent the company needs.
  • HR Human Resource
  • HR can quickly associate with this kind of people, the sense of fashion is very important, that is, the design sense and appearance of the product, and the word “adults” means that such people are not blindly pursuing trends , They have their own aesthetics and can influence the people around them.
  • HR can determine that the target talent is suitable for related work such as fashion media.
  • a talent will have multiple prediction tags, and different talents will have overlapping prediction tags.
  • the weight of the prediction tags reflects the core skills of different talents. For example, “fashionistas” and “technological pioneers” both have female labels. At this time, it is necessary to compare the weight of women's labels among different talents to determine which group of people to interpret the label. Generally, for a good portrait of a talent, the label overlap between different groups of people is small, and only those labels with lower weights will overlap slightly.
  • the target resume is obtained; the resume data is extracted from the target resume; the resume data is cleaned based on the text classification algorithm of natural language processing NLP to obtain the target text, which indicates the core skills of the target talent; Call the preset training model to analyze the target text and generate prediction labels.
  • Each prediction label corresponds to a core skill of the target talent; according to the prediction label, a talent portrait of the target talent is generated.
  • This application embodiment refines the core skills in the resume, generates a talent portrait, which is convenient for recruiters to view, and improves the efficiency of resume screening.
  • FIG. 2 another flowchart of the method for generating a talent portrait provided by the embodiment of the present application, which specifically includes:
  • the terminal obtains the target resume. Among them, the terminal selects a resume that meets the requirements from a large number of resumes according to preset screening conditions.
  • the target resume here can be one or multiple. For ease of description, this application uses one target resume as an example.
  • the terminal obtains a large number of resumes according to actual conditions, and filters out target resumes that meet the basic requirements, where the target resumes can be different types of resumes.
  • the format of the target resume can be adjusted for different groups.
  • the target resume can include: 1. Basic personal information, which can include name, age, gender, ethnicity, contact number, hometown, etc.; 2. Education background, Educational background can include middle school education experience and university education experience. Among them, middle school education experience includes school name and length of study, university education experience includes school name, length of study, and major of study; 3. Major courses, major courses can include English , Advanced Mathematics, Linear Algebra, College Physics, etc. It can also include other courses such as Building Mechanics, Material Mechanics, Contract Management, etc. The specifics are not limited here; 4. Self-evaluation, self-evaluation includes hobbies, etc.
  • the target resume can include: 1. Basic personal information, which can include name, age, gender, ethnicity, contact number, hometown, etc.; 2. Educational background, education The background can include middle school education experience and university education experience. Among them, the middle school education experience includes the name of the school and the length of study, and the university education experience includes the name of the school, the length of study, and the major of study; 3. Work experience, which can include the name of the work unit, Working hours, work content and nature of the work unit, etc.; 4. Self-evaluation, including hobbies, work attitude, etc. 5. Honorary awards. Honorary awards include study awards, work awards, and government awards. The specifics are not limited here. 6. Master the language. The master language can include Chinese, English, French, Cantonese, etc., which is not limited here.
  • resumes are suitable for different groups, and the resume content contained in the target resume is also different.
  • the resume can also be customized, and part or all of the content contained in the resume of fresh graduates and social personnel can be included in the resume. The details are not limited here.
  • the execution subject of the embodiment of the present application may be a device for generating a talent portrait, or may be a terminal or a server, which is not specifically limited here.
  • This application takes the terminal as the execution subject as an example for description.
  • the terminal extracts resume data from the target resume. Specifically, the terminal extracts resume data from the target resume.
  • the resume data includes basic information and professional information of the target talent.
  • the basic information includes name, age, gender, contact information, etc.
  • the professional information includes work unit, job Years, graduate school, graduate major, certificates obtained, professional skills, job position, etc.
  • the resume data is the various options included in the above steps.
  • the extracted resume data includes: name: Li Yan; age: 18; gender: female; contact information: 151XXXXXX54, and can also include others Information, such as graduate school: Massachusetts Institute of Technology, etc. The specifics are not limited here.
  • the text classification algorithm based on natural language processing NLP cleans the resume data to obtain the target text, which indicates the core skills of the target talent.
  • the terminal cleans the resume data based on the text classification algorithm of natural language processing (NLP) to obtain the target text, which indicates the core skills of the target talent.
  • NLP natural language processing
  • the terminal invokes a convolutional neural network CNN algorithm (TextCNN) or a logistic regression (logistic regression, LR) algorithm to perform text classification on resume data.
  • CNN convolutional neural network CNN algorithm
  • LR logistic regression
  • the features used in text classification are the words that best represent the classification.
  • resume data can be divided into two categories, namely valid data and invalid data. Among them, valid data can include work unit, name, age, gender, etc., and invalid data can include application position, working years, etc., which is not done here. limited.
  • the text classification algorithm of NLP is a preset convolutional neural network CNN algorithm or LR algorithm. According to actual needs, the resume data is classified and then filtered, invalid data is deleted, and valid data is retained. There are target talents in the valid data. Core skill information, for example, good at English, etc., that is, effective data "good at English" as the target text.
  • text classification can also be part-of-speech tagging, sentence segmentation, recognition of dialogue behavior types, and recognition of textual implication.
  • sentence segmentation e.g., sentence segmentation
  • recognition of dialogue behavior types e.g., dialogue behavior types
  • recognition of textual implication
  • the terminal calls a preset training model to analyze the target text and generates prediction labels, each of which corresponds to a core skill of the target talent. Specifically, the terminal sequentially traverses each piece of application information in the target text, and each piece of application information corresponds to a core skill of the target talent; the terminal performs word segmentation on each piece of application information based on a preset training model to obtain the word segmentation result; the terminal according to the word segmentation The result is searched in the business word bag to obtain the business keywords of each application information; the terminal arranges the business keywords in descending order of frequency to obtain the sorting result; the terminal determines the top preset number of keywords in the sorting result as the predicted label .
  • label value of the abstract classification and generalization of a certain characteristic of a certain type of specific group or object is classifiable.
  • label value can be divided into “male”, “female” and “unknown” according to the enumeration analysis (mutually exclusive collectively exhaustive, MECE) principle;
  • label “age” its label value can be divided into “ 0-18”, “18-35”, “35-60”, “60-100”, etc., which are not specifically limited here.
  • the terminal generates a talent portrait of the target talent based on the predicted label.
  • the terminal creates a target class, the target class includes multiple attributes; the terminal determines each attribute in the target class according to the prediction label, and each attribute corresponds to a prediction label; the terminal determines the output result of the target class, and the output result includes multiple predictions Label: The terminal generates a talent portrait of the target talent according to the label value of each predicted label in the output result.
  • Talent portraits can be simply understood as labels for massive amounts of data. According to the differences in users' goals, behaviors and opinions, they are divided into different types, and then typical characteristics are extracted from each type, and given names, photos, and some demographics. The description of elements, scenes, etc. forms a personas.
  • the obtained target talent portrait can include age, gender, working years, core skills, job search expectations, and so on.
  • the talent portrait obtained by refining A resume can be: Internet, married, corporate executives, high income.
  • the talent portrait can also be other tags, which are not limited here.
  • the definition of the target talent based on the characteristic value helps Human Resource (HR) to grasp the characteristics of the target talent at a glance, such as "working in computer development for many years", “excellent technology” “Etc., you can quickly understand that the target talent is good at working in the computer field, so HR can judge whether it is the talent the company needs.
  • HR Human Resource
  • HR can quickly associate with this kind of people, the sense of fashion is very important, that is, the design sense and appearance of the product, and the word “adults” means that such people are not blindly pursuing trends , They have their own aesthetics and can influence the people around them.
  • HR can determine that the target talent is suitable for related work such as fashion media.
  • a talent will have multiple prediction tags, and different talents will have overlapping prediction tags.
  • the weight of the prediction tags reflects the core skills of different talents. For example, “fashionistas” and “technological pioneers” both have female labels. At this time, it is necessary to compare the weight of women's labels among different talents to determine which group of people to interpret the label. Generally, for a good portrait of a talent, the label overlap between different groups of people is small, and only those labels with lower weights will overlap slightly.
  • the terminal scores the talent portrait of the target talent according to the preset standard to obtain the target score. Specifically, the terminal determines the weight of each piece of information in the talent portrait and the score corresponding to each piece of information according to preset standards, and calculates the target score of the talent portrait based on the weight and the score. For example, suppose that a talent profile includes 2 pieces of information, working hours and working levels, and the weights of working hours and working levels are both 0.5.
  • the preset standard is: when working hours are 0-2 years, the score is 2 points, and working hours are For 3-5 years, the score is 4 points, when the working hours are 5-8 years, the score is 6 points, and when the working hours are more than 9 years, the score is 8 points; when the working level is general staff, the score is 8 points.
  • the value is 2 points.
  • the score is 4 points.
  • the score is a high-level cadre, the score is 6 points.
  • the job level is the main leader, the score is 8 points.
  • the terminal judges whether the target score of the talent portrait of the target talent is greater than the preset score.
  • the preset score value is preset according to the situation, for example, the preset score value is 3 points, then the target resume whose template score value in the above step is 4 points is greater than the preset score value. If the target score of the talent portrait of the target talent is greater than the preset score, step 208 is executed. If the target score of the talent portrait of the target talent is less than or equal to the preset score, step 209 is executed.
  • the target score of the target talent's talent portrait is greater than the preset score, it is determined that the target talent's talent portrait meets the requirements, and the target talent's talent portrait is saved.
  • the terminal determines that the target talent's talent portrait meets the requirements, and saves the target talent's talent portrait.
  • the terminal saves the talent portrait that meets the demand, so that when needed, the corresponding resume can be retrieved according to the talent portrait.
  • the target score of the target talent's talent portrait is less than or equal to the preset score, it is determined that the target talent's talent portrait does not meet the requirements, and the target talent's talent portrait is marked.
  • the terminal determines that the target talent's talent portrait does not meet the requirements, and marks the target talent's talent portrait.
  • the terminal may also generate prompt information.
  • the prompt information may also include the identification of the talent portrait. The identification is used to indicate whether the talent portrait meets the requirements, so that the staff can receive the prompt , To differentiate and mark the portraits of talents that meet the requirements.
  • the target resume is obtained; the resume data is extracted from the target resume; the resume data is cleaned based on the text classification algorithm of natural language processing NLP to obtain the target text, which indicates the core skills of the target talent; Call the preset training model to analyze the target text and generate prediction labels.
  • Each prediction label corresponds to a core skill of the target talent; according to the prediction label, a talent portrait of the target talent is generated.
  • This application embodiment refines the core skills in the resume, generates a talent portrait, which is convenient for recruiters to view, and improves the efficiency of resume screening.
  • An embodiment of the device for generating the talent portrait in the embodiment of the application includes:
  • the obtaining unit 301 is used to obtain a target resume
  • the extraction unit 302 is configured to extract resume data from the target resume
  • the cleaning unit 303 is configured to clean the resume data based on a natural language processing NLP text classification algorithm to obtain a target text, the target text indicating the core skills of the target talent;
  • the first generating unit 304 is configured to call a preset training model to analyze the target text and generate prediction labels, each prediction label corresponding to a core skill of the target talent;
  • the second generating unit 305 is configured to generate a talent portrait of the target talent according to the predicted tag.
  • FIG. 4 another embodiment of the device for generating a talent portrait in the embodiment of the present application includes:
  • the obtaining unit 301 is used to obtain a target resume
  • the extraction unit 302 is configured to extract resume data from the target resume
  • the cleaning unit 303 is configured to clean the resume data based on a natural language processing NLP text classification algorithm to obtain a target text, the target text indicating the core skills of the target talent;
  • the first generating unit 304 is configured to call a preset training model to analyze the target text and generate prediction labels, each prediction label corresponding to a core skill of the target talent;
  • the second generating unit 305 is configured to generate a talent portrait of the target talent according to the predicted tag.
  • the first generating unit 304 is specifically configured to: traverse each piece of application information in the target text in turn, and each piece of application information corresponds to a core skill of the target talent; The application information is segmented to obtain the word segmentation result; the business word bag is searched according to the word segmentation result to obtain the business keywords of each application information; the business keywords are arranged in descending order of frequency to obtain the sorting result; The keywords with the preset number at the top of the result are determined as predicted tags.
  • the second generating unit 305 is specifically configured to: create a target class, the target class including multiple attributes; determine each attribute in the target class according to the prediction label, and each attribute corresponds to one prediction Label; determine the output result of the target class, the output result including a plurality of the predicted labels; generate the talent portrait of the target talent according to the label value of each predicted label in the output result.
  • the device for generating a talent portrait further includes: a third generating unit 306, configured to generate a preset training model, and the preset training model is used to generate a predicted label based on text data.
  • a third generating unit 306 configured to generate a preset training model, and the preset training model is used to generate a predicted label based on text data.
  • the third generation unit 306 includes: an acquisition module 3061, configured to acquire a preset number of resume data; a cleaning module 3062, configured to clean the preset number of resume data to obtain valid data; first determination Module 3063, used to determine the corresponding label for each valid data to obtain a corpus; input module 3064, used to input the corpus into the text training model; adjustment generation module 3065, used to parameterize the text training model Adjust to generate a preset training model, and the preset training model is used to generate a predicted label based on the text data.
  • a second determining module 3066 configured to determine the accuracy rate P and recall rate R of the preset training model
  • a generating module 3067 configured to determine the accuracy rate P and the recall rate R according to the accuracy rate P and the The recall rate R
  • the device for generating a talent portrait further includes: a scoring unit 307 for scoring the talent portrait of the target talent according to a preset standard to obtain a target score; a judging unit 308 for judging the talent of the target talent Whether the target score of the portrait is greater than the preset score; the storage unit 309, if the target score of the talent portrait of the target talent is greater than the preset score, it is used to determine that the talent portrait of the target talent meets the requirements, And save the talent portrait of the target talent; the marking unit 310, if the target score of the talent portrait of the target talent is less than or equal to the preset score, is used to determine the talent portrait of the target talent Does not meet the requirements, and mark the talent portrait of the target talent.
  • FIG. 5 is a schematic structural diagram of a device for generating a talent portrait provided by an embodiment of the present application.
  • the device 500 for generating a talent portrait may have relatively large differences due to different configurations or performance, and may include one or more processors (central processing units, CPU) 501 (for example, one or more processors) and memory 509, one or more storage media 508 for storing application programs 507 or data 506 (for example, one or one storage device with a large amount of storage), and storage media 508 may be It is non-volatile or volatile.
  • the memory 509 and the storage medium 508 may be short-term storage or persistent storage.
  • the program stored in the storage medium 508 may include one or more modules (not shown in the figure), and each module may include a series of command operations in the device for generating the talent portrait.
  • the processor 501 may be configured to communicate with the storage medium 508, and execute a series of instruction operations in the storage medium 508 on the device 500 for generating the talent portrait.
  • the talent portrait generation device 500 may also include one or more power supplies 502, one or more wired or wireless network interfaces 503, one or more input and output interfaces 504, and/or one or more operating systems 505, such as Windows Serve, Mac OS X, Unix, Linux, FreeBSD, etc.
  • operating systems 505 such as Windows Serve, Mac OS X, Unix, Linux, FreeBSD, etc.
  • the processor 501 can execute the acquisition unit 301, the extraction unit 302, the cleaning unit 303, the first generation unit 304, the second generation unit 305, the third generation unit 306, the scoring unit 307, the judgment unit 308, and the marking unit 310 in the above embodiments. Function.
  • the processor 501 is the control center of the device for generating the talent portrait, and can perform processing according to the set method for generating the talent portrait.
  • the processor 501 uses various interfaces and lines to connect the various parts of the entire talent portrait generation device, and executes the talent portrait by running or executing software programs and/or modules stored in the memory 509, and calling data stored in the memory 509.
  • the various functions and processing data of the generation equipment can refine the core skills in the resume, generate talent portraits, which are convenient for recruiters to view, and improve the efficiency of resume screening.
  • the storage medium 508 and the memory 509 are both carriers for storing data.
  • the storage medium 508 may refer to an internal memory with a small storage capacity but high speed, and the storage 509 may have a large storage capacity but a slow storage speed. External memory.
  • the memory 509 may be used to store software programs and modules.
  • the processor 501 executes various functional applications and data processing of the talent portrait generation device 500 by running the software programs and modules stored in the memory 509.
  • the memory 509 may mainly include a storage program area and a storage data area.
  • the storage program area may store an operating system, an application program required by at least one function (such as extracting resume data from a target resume), etc.; The data (such as predicted labels, etc.) created by the use of the portrait generation device.
  • the memory 509 may include a high-speed random access memory, and may also include a non-volatile memory, such as at least one magnetic disk storage device, a flash memory device, or other volatile solid-state storage devices.
  • the method and program for generating the talent portrait provided in the embodiment of the present application and the received data stream are stored in the memory, and the processor 501 is called from the memory 509 when needed.
  • the computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices.
  • the computer instructions may be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another computer-readable storage medium.
  • the computer instructions may be transmitted from a website, computer, server, or data center. Transmission to another website site, computer, server or data center via wired (such as coaxial cable, optical fiber, twisted pair) or wireless (such as infrared, wireless, microwave, etc.).
  • the computer-readable storage medium may be any available medium that can be stored by a computer or a data storage device such as a server or data center integrated with one or more available media.
  • the usable medium may be a magnetic medium (for example, a floppy disk, a hard disk, and a magnetic tape), an optical medium (for example, an optical disc), or a semiconductor medium (for example, a solid state disk (SSD)).

Abstract

本申请涉及人工智能领域，尤其涉及自然语言处理领域，公开了一种人才画像的生成方法、装置、设备及存储介质，用于对简历中的核心技能进行提炼，生成人才画像，便于招聘人员查看，提高了简历筛选效率。本申请方法包括：获取目标简历；从所述目标简历中提取简历数据；基于自然语言处理NLP的文本分类算法对所述简历数据进行清洗，得到目标文本，所述目标文本指示目标人才的核心技能；调用预置的训练模型对所述目标文本进行分析，生成预测标签，每个预测标签都对应所述目标人才的一个核心技能；根据所述预测标签生成所述目标人才的人才画像。

Description

人才画像的生成方法、装置、设备及存储介质
本申请要求于2019年6月18日提交中国专利局、申请号为201910524634.X，发明名称为“人才画像的生成方法、装置、设备及存储介质”的中国专利申请的优先权，其全部内容通过引用结合在本申请中。
技术领域
本申请涉及人工智能领域，尤其涉及一种人才画像的生成方法、装置、设备及存储介质。
背景技术
目前，在一些大企业招聘外包人员时，通常是由人力资源(human resource,HR)直接筛选简历，然后将筛选后的简历发送到专业部门，企业的专业部门需对HR发送的应聘人员简历进行确认，然后开始进行招聘流程。然而，HR不可能对所有岗位职能都了解，因此HR可能会发送不符合要求人员的简历，推荐不符合用人单位要求的人员参加面试。
发明人发现，HR查看候选人的简历时需要自己提炼简历的关键信息，对于自己不了解的行业，提取关键信息则费时费力，效率低。
技术问题
本申请提供了一种人才画像的生成方法、装置、设备及存储介质，用于对简历中的核心技能进行提炼，生成人才画像，便于招聘人员查看，提高了简历筛选效率。
技术解决方案
本申请实施例的第一方面提供一种人才画像的生成方法，包括：获取目标简历；从所述目标简历中提取简历数据；基于自然语言处理NLP的文本分类算法对所述简历数据进行清洗，得到目标文本，所述目标文本指示目标人才的核心技能；调用预置的训练模型对所述目标文本进行分析，生成预测标签，每个预测标签都对应所述目标人才的一个核心技能；根据所述预测标签生成所述目标人才的人才画像。
本申请实施例的第二方面提供了一种人才画像的生成装置，包括：获取单元，用于获取目标简历；提取单元，用于从所述目标简历中提取简历数据；清洗单元，用于基于自然语言处理NLP的文本分类算法对所述简历数据进行清洗，得到目标文本，所述目标文本指示目标人才的核心技能；第一生成单元，用于调用预置的训练模型对所述目标文本进行分析，生成预测标签，每个预测标签都对应所述目标人才的一个核心技能；第二生成单元，用于根据所述预测标签生成所述目标人才的人才画像。
本申请实施例的第三方面提供了一种人才画像的生成设备，包括存储器、处理器及存储在所述存储器上并可在所述处理器上运行的计算机可读指令，所述处理器执行所述计算机可读指令时实现上述任一实施方式所述的人才画像的生成方法。
本申请实施例的第四方面提供了一种计算机可读存储介质，包括指令，当所述指令在计算机上运行时，使得计算机执行上述任一实施方式所述的人才画像的生成方法的步骤。
有益效果
本发明实施例提供的技术方案中，获取目标简历；从目标简历中提取简历数据；基于自然语言处理NLP的文本分类算法对简历数据进行清洗，得到目标文本，目标文本指示目标人才的核心技能；调用预置的训练模型对目标文本进行分析，生成预测标签，每个预测标签都对应目标人才的一个核心技能；根据预测标签生成目标人才的人才画像。本发明实施例，对简历中的核心技能进行提炼，生成人才画像，便于招聘人员查看，提高了简历筛选效率。
附图说明
图1为本申请实施例中人才画像的生成方法的一个实施例示意图；
图2为本申请实施例中人才画像的生成方法的另一个实施例示意图；
图3为本申请实施例中人才画像的生成装置的一个实施例示意图；
图4为本申请实施例中人才画像的生成装置的另一个实施例示意图；
图5为本申请实施例中人才画像的生成设备的一个实施例示意图。
本发明的最佳实施方式
请参阅图1，本申请实施例提供的人才画像的生成方法的流程图，具体包括：
101、获取目标简历。
终端获取目标简历。其中，终端根据预置的筛选条件，从海量的简历中选择出符合要求的简历。这里的目标简历可以是一个，也可以是多个，为了便于描述，本申请以一个目标简历为例进行说明。
具体的，终端根据实际的情况获取海量的简历，并筛选出符合基本要求的目标简历，其中，目标简历可以是不同类型的简历。目标简历可以针对不同的群体，对简历的格式进行调整。例如，若目标简历为应届毕业生简历，那么该目标简历中可以包括：1、个人基本信息，个人基本信息中可以包括姓名、年龄、性别、民族、联系电话、籍贯等；2、教育背景，教育背景中可以包括中学教育经历和大学教育经历，其中，中学教育经历包括学校名称和学习时长，大学教育经历包括学校名称、学习时长、学习专业；3、主修课程，主修课程可以包括英语、高等数学、线性代数、大学物理等，还可以包括建筑力学、材料力学、合同管理等其他课程，具体此处不做限定；4、自我评价，自我评价包括兴趣爱好等。
例如，若目标简历为社会人员简历，那么该目标简历中可以包括：1、个人基本信息，个人基本信息中可以包括姓名、年龄、性别、民族、联系电话、籍贯等；2、教育背景，教育背景中可以包括中学教育经历和大学教育经历，其中，中学教育经历包括学校名称和学习时长，大学教育经历包括学校名称、学习时长、学习专业；3、工作经验，工作经验可以包括工作单位名称、工作时长、工作内容和工作单位性质等；4、自我评价，自我评价包括兴趣爱好，工作态度等。5、荣誉奖项，荣誉奖项包括学习奖项、工作奖项以及政府奖项等，具体此处不做限定。6、掌握语言，掌握语言可以包括汉语、英语、法语、粤语等，具体此处不做限定。
需要说明的是，不同的类型的简历，适应的群体也不同，目标简历中包含的简历内容也不相同。简历也可以自定义，可以在简历中同时包括应届毕业生简历和社会人员简历中包含的部分或全部内容，具体此处不做限定。
可以理解的是，本申请实施例的执行主体可以为人才画像的生成装置，还可以是终端或者服务器，具体此处不做限定。本申请以终端为执行主体为例进行说明。
102、从目标简历中提取简历数据。
终端从目标简历中提取简历数据。具体的，终端从目标简历中提取简历数据，其中，该简历数据包括了目标人才的基本信息和专业信息，其中，基本信息包括姓名、年龄、性别、联系方式等，专业信息包括工作单位、工作年限、毕业院校、毕业专业、获得的证书、专业技能、应聘职位等。
其中，简历数据为上述步骤中包括的各个选项，例如，对于其中一份简历，提取到的简历数据包括，姓名：李一；年龄：18；性别：女；联系方式：151XXXXXX54，还可以包括其他信息，如毕业院校：麻省理工等，具体此处不做限定。
103、基于自然语言处理NLP的文本分类算法对简历数据进行清洗，得到目标文本，目标文本指示目标人才的核心技能。
终端基于自然语言处理(natural language processing，NLP)的文本分类算法对简历数据进行清洗，得到目标文本，目标文本指示目标人才的核心技能。具体的，终端调用卷积神经网络CNN算法(TextCNN)或逻辑回归(logistic regression，LR)算法对简历数据进行文本分类。可以理解的是，不管是什么分类，最重要的是要知道哪些特征是最能反 映这个分类的特点，也就是特征选取，文本分类使用的特征就是最能代表这个分类的词。例如，可以将简历数据分为两类，即有效数据和无效数据，其中，有效数据可以包括工作单位、姓名、年龄、性别等，无效数据可以包括应聘职位、工作年限等，具体此处不做限定。
其中，NLP的文本分类算法为预置的卷积神经网络CNN算法或LR算法，按照实际需要，将简历数据分类后，进行筛选，删除无效数据，保留有效数据，该有效数据中有目标人才的核心技能信息，例如，擅长英语等，即将有效数据“擅长英语”作为目标文本。
需要说明的是，文本分类除了可以是文档分类，还可以是词性标注、句子分割、识别对话行为类型和识别文字蕴含等，具体可以参见现有技术，此处不再赘述。
104、调用预置的训练模型对目标文本进行分析，生成预测标签，每个预测标签都对应目标人才的一个核心技能。
终端调用预置的训练模型对目标文本进行分析，生成预测标签，每个预测标签都对应目标人才的一个核心技能。具体的，终端依次遍历目标文本中的每条应聘信息，每条应聘信息都对应目标人才的一个核心技能；终端基于预置的训练模型对每条应聘信息进行分词，得到分词结果；终端根据分词结果在业务词袋中进行检索，得到每条应聘信息的业务关键词；终端将业务关键词按照频数降序排列得到排序结果；终端将排序结果中靠前的预置数目的关键词确定为预测标签。
其中，对某一类特定群体或对象的某项特征进行的抽象分类和概括，其值(标签值)具备可分类性。例如，“性别”，其标签值根据枚举分析(mutually exclusive collectively exhaustive，MECE)原则可分为“男”、“女”、“未知”；对于标签“年龄”，其标签值可分为“0-18”、“18-35”、“35-60”、“60-100”等，具体此处不做限定。
105、根据预测标签生成目标人才的人才画像。
终端根据预测标签生成目标人才的人才画像。具体的，终端创建一个目标类，目标类包括多个属性；终端根据预测标签确定目标类中的各个属性，每个属性对应一个预测标签；终端确定目标类的输出结果，输出结果包括多个预测标签；终端根据输出结果中各个预测标签的标签值生成目标人才的人才画像。
人才画像可以简单理解成是海量数据的标签，根据用户的目标、行为和观点的差异，将他们区分为不同的类型，然后每种类型中抽取出典型特征，赋予名字、照片、一些人口统计学要素、场景等描述，形成了一个人物原型(personas)。例如，得到的目标人才画像中可以包括年龄，性别，工作年限，核心技能，求职期望等等。具体的，例如，通过对A简历进行提炼后得到的人才画像可以为：互联网、已婚、企业高管、高收入。人才画像还可以是其他标签，具体此处不做限定。
可以理解的是，根据特征值(即预测标签)对目标人才进行定义，有助于人力资源(Human Resource，HR)一目了然掌握该目标人才的特性，如“从事多年计算机开发工作”、“技术过硬”等，可以快速的了解到目标人才为擅长从事计算机领域的工作，由此HR可以判断是否为公司需要的人才。又例如，如“时尚达人”，HR可以快速的联想到针对这类人，时尚感至关重要，即产品的设计感、外观等，并且达人两字表明该类人并不盲目追求潮流，他们有自己的审美观，并且能够影响身边的人。HR可以确定目标人才适合从事时尚传媒等相关工作。
同时，一个人才会有多个预测标签，不同的人才之间也会有预测标签的重合，此时预测标签的权重反映了不同人才的核心技能。例如“时尚达人”和“科技先锋”两类人群中都有女性标签，此时需要比较女性在不同人才中的标签权重，以决定将该标签解读给哪类人群。通常，一个好的人才画像，不同人群之间的标签重合度较小，只有在那些权重较小的标签上会有些许重合。
本申请实施例提供的技术方案中，获取目标简历；从目标简历中提取简历数据；基于自然语言处理NLP的文本分类算法对简历数据进行清洗，得到目标文本，目标文本指示目 标人才的核心技能；调用预置的训练模型对目标文本进行分析，生成预测标签，每个预测标签都对应目标人才的一个核心技能；根据预测标签生成目标人才的人才画像。本申请实施例，对简历中的核心技能进行提炼，生成人才画像，便于招聘人员查看，提高了简历筛选效率。
请参阅图2，本申请实施例提供的人才画像的生成方法的另一个流程图，具体包括：
201、获取目标简历。
终端获取目标简历。其中，终端根据预置的筛选条件，从海量的简历中选择出符合要求的简历。这里的目标简历可以是一个，也可以是多个，为了便于描述，本申请以一个目标简历为例进行说明。
具体的，终端根据实际的情况获取海量的简历，并筛选出符合基本要求的目标简历，其中，目标简历可以是不同类型的简历。目标简历可以针对不同的群体，对简历的格式进行调整。例如，若目标简历为应届毕业生简历，那么该目标简历中可以包括：1、个人基本信息，个人基本信息中可以包括姓名、年龄、性别、民族、联系电话、籍贯等；2、教育背景，教育背景中可以包括中学教育经历和大学教育经历，其中，中学教育经历包括学校名称和学习时长，大学教育经历包括学校名称、学习时长、学习专业；3、主修课程，主修课程可以包括英语、高等数学、线性代数、大学物理等，还可以包括建筑力学、材料力学、合同管理等其他课程，具体此处不做限定；4、自我评价，自我评价包括兴趣爱好等。
例如，若目标简历为社会人员简历，那么该目标简历中可以包括：1、个人基本信息，个人基本信息中可以包括姓名、年龄、性别、民族、联系电话、籍贯等；2、教育背景，教育背景中可以包括中学教育经历和大学教育经历，其中，中学教育经历包括学校名称和学习时长，大学教育经历包括学校名称、学习时长、学习专业；3、工作经验，工作经验可以包括工作单位名称、工作时长、工作内容和工作单位性质等；4、自我评价，自我评价包括兴趣爱好，工作态度等。5、荣誉奖项，荣誉奖项包括学习奖项、工作奖项以及政府奖项等，具体此处不做限定。6、掌握语言，掌握语言可以包括汉语、英语、法语、粤语等，具体此处不做限定。
需要说明的是，不同的类型的简历，适应的群体也不同，目标简历中包含的简历内容也不相同。简历也可以自定义，可以在简历中同时包括应届毕业生简历和社会人员简历中包含的部分或全部内容，具体此处不做限定。
可以理解的是，本申请实施例的执行主体可以为人才画像的生成装置，还可以是终端或者服务器，具体此处不做限定。本申请以终端为执行主体为例进行说明。
202、从目标简历中提取简历数据。
终端从目标简历中提取简历数据。具体的，终端从目标简历中提取简历数据，其中，该简历数据包括了目标人才的基本信息和专业信息，其中，基本信息包括姓名、年龄、性别、联系方式等，专业信息包括工作单位、工作年限、毕业院校、毕业专业、获得的证书、专业技能、应聘职位等。
其中，简历数据为上述步骤中包括的各个选项，例如，对于其中一份简历，提取到的简历数据包括，姓名：李燕；年龄：18；性别：女；联系方式：151XXXXXX54，还可以包括其他信息，如毕业院校：麻省理工等，具体此处不做限定。
203、基于自然语言处理NLP的文本分类算法对简历数据进行清洗，得到目标文本，目标文本指示目标人才的核心技能。
终端基于自然语言处理(natural language processing，NLP)的文本分类算法对简历数据进行清洗，得到目标文本，目标文本指示目标人才的核心技能。具体的，终端调用卷积神经网络CNN算法(TextCNN)或逻辑回归(logistic regression，LR)算法对简历数据进行文本分类。可以理解的是，不管是什么分类，最重要的是要知道哪些特征是最能反映这个分类的特点，也就是特征选取，文本分类使用的特征就是最能代表这个分类的词。例如，可以将简历数据分为两类，即有效数据和无效数据，其中，有效数据可以包括工作 单位、姓名、年龄、性别等，无效数据可以包括应聘职位、工作年限等，具体此处不做限定。
其中，NLP的文本分类算法为预置的卷积神经网络CNN算法或LR算法，按照实际需要，将简历数据分类后，进行筛选，删除无效数据，保留有效数据，该有效数据中有目标人才的核心技能信息，例如，擅长英语等，即将有效数据“擅长英语”作为目标文本。
需要说明的是，文本分类除了可以是文档分类，还可以是词性标注、句子分割、识别对话行为类型和识别文字蕴含等，具体可以参见现有技术，此处不再赘述。
204、调用预置的训练模型对目标文本进行分析，生成预测标签，每个预测标签都对应目标人才的一个核心技能。
终端调用预置的训练模型对目标文本进行分析，生成预测标签，每个预测标签都对应目标人才的一个核心技能。具体的，终端依次遍历目标文本中的每条应聘信息，每条应聘信息都对应目标人才的一个核心技能；终端基于预置的训练模型对每条应聘信息进行分词，得到分词结果；终端根据分词结果在业务词袋中进行检索，得到每条应聘信息的业务关键词；终端将业务关键词按照频数降序排列得到排序结果；终端将排序结果中靠前的预置数目的关键词确定为预测标签。
其中，对某一类特定群体或对象的某项特征进行的抽象分类和概括，其值(标签值)具备可分类性。例如，“性别”，其标签值根据枚举分析(mutually exclusive collectively exhaustive，MECE)原则可分为“男”、“女”、“未知”；对于标签“年龄”，其标签值可分为“0-18”、“18-35”、“35-60”、“60-100”等，具体此处不做限定。
205、根据预测标签生成目标人才的人才画像。
终端根据预测标签生成目标人才的人才画像。具体的，终端创建一个目标类，目标类包括多个属性；终端根据预测标签确定目标类中的各个属性，每个属性对应一个预测标签；终端确定目标类的输出结果，输出结果包括多个预测标签；终端根据输出结果中各个预测标签的标签值生成目标人才的人才画像。
人才画像可以简单理解成是海量数据的标签，根据用户的目标、行为和观点的差异，将他们区分为不同的类型，然后每种类型中抽取出典型特征，赋予名字、照片、一些人口统计学要素、场景等描述，形成了一个人物原型(personas)。例如，得到的目标人才画像中可以包括年龄，性别，工作年限，核心技能，求职期望等等。具体的，例如，通过对A简历进行提炼后得到的人才画像可以为：互联网、已婚、企业高管、高收入。人才画像还可以是其他标签，具体此处不做限定。
可以理解的是，根据特征值(即预测标签)对目标人才进行定义，有助于人力资源(Human Resource，HR)一目了然掌握该目标人才的特性，如“从事多年计算机开发工作”、“技术过硬”等，可以快速的了解到目标人才为擅长从事计算机领域的工作，由此HR可以判断是否为公司需要的人才。又例如，如“时尚达人”，HR可以快速的联想到针对这类人，时尚感至关重要，即产品的设计感、外观等，并且达人两字表明该类人并不盲目追求潮流，他们有自己的审美观，并且能够影响身边的人。HR可以确定目标人才适合从事时尚传媒等相关工作。
同时，一个人才会有多个预测标签，不同的人才之间也会有预测标签的重合，此时预测标签的权重反映了不同人才的核心技能。例如“时尚达人”和“科技先锋”两类人群中都有女性标签，此时需要比较女性在不同人才中的标签权重，以决定将该标签解读给哪类人群。通常，一个好的人才画像，不同人群之间的标签重合度较小，只有在那些权重较小的标签上会有些许重合。
206、对目标人才的人才画像按照预设标准进行评分得到目标分值。
终端对目标人才的人才画像按照预设标准进行评分得到目标分值。具体的，终端按照预设标准，确定人才画像中每条信息的权重和每条信息对应的分值，根据权重和分值计算得到人才画像的目标分值。例如，假设人才画像包括2条信息，工作时长和工作级别，工 作时长和工作级别的权重都为0.5，预设标准为：当工作时长为0-2年时，分值为2分，工作时长为3-5年时，分值为4分，工作时长为5-8年时，分值为6分，工作时长为9年以上时，分值为8分；工作级别为普通职员时，分值为2分，工作级别为中层干部时，分值为4分，工作级别为高层干部时，分值为6分，工作级别为主要领导时，分值为8分。当目标人才的人才画像中工作时长为4年，工作级别为中层干部时，计算得到4*0.5+4*0.5＝4，得到的目标分值为4分。
207、判断目标人才的人才画像的目标分值是否大于预设的分值。
终端判断目标人才的人才画像的目标分值是否大于预设的分值。具体的，预设的分值为根据事情情况预先设置，例如，预设的分值为3分，那么上述步骤中的模板分值为4分的目标简历大于预设的分值。若目标人才的人才画像的目标分值大于预设的分值，则执行步骤208。若目标人才的人才画像的目标分值小于或等于预设的分值，则执行步骤209。
208、若目标人才的人才画像的目标分值大于预设的分值，则确定目标人才的人才画像满足要求，并将目标人才的人才画像进行保存。
若目标人才的人才画像的目标分值大于预设的分值，则终端确定目标人才的人才画像满足要求，并将目标人才的人才画像进行保存。终端将满足需求的人才画像进行保存，以使得在需要使用时根据该人才画像调取对应的简历。
209、若目标人才的人才画像的目标分值小于或等于预设的分值，则确定目标人才的人才画像不满足要求，并将目标人才的人才画像进行标记。
若目标人才的人才画像的目标分值小于或等于预设的分值，则终端确定目标人才的人才画像不满足要求，并将目标人才的人才画像进行标记。
可选的，终端还可以生成提示信息其中，提示信息除了包括人才画像的分值，还可以包括人才画像的标识，该标识用于指示人才画像是否满足要求，以使得工作人员在收到提示时，针对性地对满足要求的人才画像进行区分标记。
本申请实施例提供的技术方案中，获取目标简历；从目标简历中提取简历数据；基于自然语言处理NLP的文本分类算法对简历数据进行清洗，得到目标文本，目标文本指示目标人才的核心技能；调用预置的训练模型对目标文本进行分析，生成预测标签，每个预测标签都对应目标人才的一个核心技能；根据预测标签生成目标人才的人才画像。本申请实施例，对简历中的核心技能进行提炼，生成人才画像，便于招聘人员查看，提高了简历筛选效率。
上面对本申请实施例中人才画像的生成方法进行了描述，下面对本申请实施例中人才画像的生成装置进行描述，请参阅图3，本申请实施例中人才画像的生成装置的一个实施例包括：
获取单元301，用于获取目标简历；
提取单元302，用于从所述目标简历中提取简历数据；
清洗单元303，用于基于自然语言处理NLP的文本分类算法对所述简历数据进行清洗，得到目标文本，所述目标文本指示目标人才的核心技能；
第一生成单元304，用于调用预置的训练模型对所述目标文本进行分析，生成预测标签，每个预测标签都对应所述目标人才的一个核心技能；
第二生成单元305，用于根据所述预测标签生成所述目标人才的人才画像。
请参阅图4，本申请实施例中人才画像的生成装置的另一个实施例包括：
获取单元301，用于获取目标简历；
提取单元302，用于从所述目标简历中提取简历数据；
清洗单元303，用于基于自然语言处理NLP的文本分类算法对所述简历数据进行清洗，得到目标文本，所述目标文本指示目标人才的核心技能；
第一生成单元304，用于调用预置的训练模型对所述目标文本进行分析，生成预测标签，每个预测标签都对应所述目标人才的一个核心技能；
第二生成单元305，用于根据所述预测标签生成所述目标人才的人才画像。
可选的，第一生成单元304具体用于：依次遍历所述目标文本中的每条应聘信息，每条应聘信息都对应所述目标人才的一个核心技能；基于预置的训练模型对每条应聘信息进行分词，得到分词结果；根据所述分词结果在业务词袋中进行检索，得到每条应聘信息的业务关键词；将所述业务关键词按照频数降序排列得到排序结果；将所述排序结果中靠前的预置数目的关键词确定为预测标签。
可选的，第二生成单元305具体用于：创建一个目标类，所述目标类包括多个属性；根据所述预测标签确定所述目标类中的各个属性，每个属性对应一个所述预测标签；确定所述目标类的输出结果，所述输出结果包括多个所述预测标签；根据所述输出结果中各个预测标签的标签值生成所述目标人才的人才画像。
可选的，人才画像的生成装置还包括：第三生成单元306，用于生成预置的训练模型，所述预置的训练模型用于根据文本数据生成预测标签。
可选的，第三生成单元306包括：获取模块3061，用于获取预置数量的简历数据；清洗模块3062，用于对所述预置数量的简历数据进行清洗，得到有效数据；第一确定模块3063，用于为每个有效数据确定对应的标签，得到语料库；输入模块3064，用于将所述语料库输入到文本训练模型中；调整生成模块3065，用于对所述文本训练模型进行参数调整，生成预置的训练模型，所述预置的训练模型用于根据文本数据生成预测标签。
可选的，第三生成单元306还包括：第二确定模块3066，用于确定所述预置的训练模型的精确率P和召回率R；生成模块3067，用于根据所述精确率P和所述召回率R生成所述预置的训练模型的Fα值，所述Fα值满足公式：Fα＝(α 2+1)PR/(α 2P+R)，其中，所述α大于或等于1；判断模块3068，用于根据所述Fα值判断训练模型的正确率是否大于预设的阈值；第三确定模块3069，若所述训练模型的正确率大于预设的阈值，则用于确定所述预置的训练模型满足生产需求。
可选的，人才画像的生成装置还包括：评分单元307，用于对所述目标人才的人才画像按照预设标准进行评分得到目标分值；判断单元308，用于判断所述目标人才的人才画像的目标分值是否大于预设的分值；保存单元309，若所述目标人才的人才画像的目标分值大于预设的分值，则用于确定所述目标人才的人才画像满足要求，并将所述目标人才的人才画像进行保存；标记单元310，若所述目标人才的人才画像的目标分值小于或等于所述预设的分值，则用于确定所述目标人才的人才画像不满足要求，并将所述目标人才的人才画像进行标记。
图5是本申请实施例提供的一种人才画像的生成设备的结构示意图，该人才画像的生成设备500可因配置或性能不同而产生比较大的差异，可以包括一个或一个以上处理器(central processing units，CPU)501(例如，一个或一个以上处理器)和存储器509，一个或一个以上存储应用程序507或数据506的存储介质508(例如一个或一个以上海量存储设备)，存储介质508可以是非易失性，也可以是易失性。其中，存储器509和存储介质508可以是短暂存储或持久存储。存储在存储介质508的程序可以包括一个或一个以上模块(图示没标出)，每个模块可以包括对人才画像的生成设备中的一系列指令操作。更进一步地，处理器501可以设置为与存储介质508通信，在人才画像的生成设备500上执行存储介质508中的一系列指令操作。
人才画像的生成设备500还可以包括一个或一个以上电源502，一个或一个以上有线或无线网络接口503，一个或一个以上输入输出接口504，和/或，一个或一个以上操作系统505，例如Windows Serve，Mac OS X，Unix，Linux，FreeBSD等等。本领域技术人员可以理解，图5中示出的人才画像的生成设备结构并不构成对人才画像的生成设备的限定，可以包括比图示更多或更少的部件，或者组合某些部件，或者不同的部件布置。处理器501可以执行上述实施例中获取单元301、提取单元302、清洗单元303、第一生成单元304、 第二生成单元305、第三生成单元306、评分单元307、判断单元308和标记单元310的功能。
下面结合图5对人才画像的生成设备的各个构成部件进行具体的介绍：
处理器501是人才画像的生成设备的控制中心，可以按照设置的人才画像的生成方法进行处理。处理器501利用各种接口和线路连接整个人才画像的生成设备的各个部分，通过运行或执行存储在存储器509内的软件程序和/或模块，以及调用存储在存储器509内的数据，执行人才画像的生成设备的各种功能和处理数据，对简历中的核心技能进行提炼，生成人才画像，便于招聘人员查看，提高了简历筛选效率。存储介质508和存储器509都是存储数据的载体，本申请实施例中，存储介质508可以是指储存容量较小，但速度快的内存储器，而存储器509可以是储存容量大，但储存速度慢的外存储器。
存储器509可用于存储软件程序以及模块，处理器501通过运行存储在存储器509的软件程序以及模块，从而执行人才画像的生成设备500的各种功能应用以及数据处理。存储器509可主要包括存储程序区和存储数据区，其中，存储程序区可存储操作系统、至少一个功能所需的应用程序(比如从目标简历中提取简历数据)等；存储数据区可存储根据人才画像的生成设备的使用所创建的数据(比如预测标签等)等。此外，存储器509可以包括高速随机存取存储器，还可以包括非易失性存储器，例如至少一个磁盘存储器件、闪存器件、或其他易失性固态存储器件。在本申请实施例中提供的人才画像的生成方法程序和接收到的数据流存储在存储器中，当需要使用时，处理器501从存储器509中调用。
在计算机上加载和执行所述计算机可读指令时，全部或部分地产生按照本申请实施例所述的流程或功能。所述计算机可以是通用计算机、专用计算机、计算机网络、或者其他可编程装置。所述计算机指令可以存储在计算机可读存储介质中，或者从一个计算机可读存储介质向另一计算机可读存储介质传输，例如，所述计算机指令可以从一个网站站点、计算机、服务器或数据中心通过有线(例如同轴电缆、光纤、双绞线)或无线(例如红外、无线、微波等)方式向另一个网站站点、计算机、服务器或数据中心进行传输。所述计算机可读存储介质可以是计算机能够存储的任何可用介质或者是包含一个或多个可用介质集成的服务器、数据中心等数据存储设备。所述可用介质可以是磁性介质，(例如，软盘、硬盘、磁带)、光介质(例如，光盘)、或者半导体介质(例如固态硬盘(solid state disk，SSD))等。

Claims (20)

  1. 一种人才画像的生成方法，包括：
    获取目标简历；
    从所述目标简历中提取简历数据；
    基于自然语言处理NLP的文本分类算法对所述简历数据进行清洗，得到目标文本，所述目标文本指示目标人才的核心技能；
    调用预置的训练模型对所述目标文本进行分析，生成预测标签，每个预测标签都对应所述目标人才的一个核心技能；
    根据所述预测标签生成所述目标人才的人才画像。
  2. 根据权利要求1所述的人才画像的生成方法，所述调用预置的训练模型对所述目标文本进行分析，生成预测标签，每个预测标签都对应所述目标人才的一个核心技能包括：
    依次遍历所述目标文本中的每条应聘信息，每条应聘信息都对应所述目标人才的一个核心技能；
    基于预置的训练模型对每条应聘信息进行分词，得到分词结果；
    根据所述分词结果在业务词袋中进行检索，得到每条应聘信息的业务关键词；
    将所述业务关键词按照频数降序排列得到排序结果；
    将所述排序结果中靠前的预置数目的关键词确定为预测标签。
  3. 根据权利要求1所述的人才画像的生成方法，所述根据所述预测标签生成所述目标人才的人才画像包括：
    创建一个目标类，所述目标类包括多个属性；
    根据所述预测标签确定所述目标类中的各个属性，每个属性对应一个所述预测标签；
    确定所述目标类的输出结果，所述输出结果包括多个所述预测标签；
    根据所述输出结果中各个预测标签的标签值生成所述目标人才的人才画像。
  4. 根据权利要求1所述的人才画像的生成方法，所述基于自然语言处理NLP的文本分类算法对所述简历数据进行清洗，得到目标文本，所述目标文本指示目标人才的核心技能之后，所述调用预置的训练模型对所述目标文本进行分析，生成预测标签，每个预测标签都对应所述目标人才的一个核心技能之前，所述方法还包括：
    生成预置的训练模型，所述预置的训练模型用于根据文本数据生成预测标签。
  5. 根据权利要求4所述的人才画像的生成方法，所述生成预置的训练模型，所述预置的训练模型用于根据文本数据生成预测标签包括：
    获取预置数量的简历数据；
    对所述预置数量的简历数据进行清洗，得到有效数据；
    为每个有效数据确定对应的标签，得到语料库；
    将所述语料库输入到文本训练模型中；
    对所述文本训练模型进行参数调整，生成预置的训练模型，所述预置的训练模型用于根据文本数据生成预测标签。
  6. 根据权利要求5所述的人才画像的生成方法，所述对所述文本训练模型进行参数调整，生成预置的训练模型，所述预置的训练模型用于根据文本数据生成预测标签之后，所述方法还包括：
    确定所述预置的训练模型的精确率P和召回率R；
    根据所述精确率P和所述召回率R生成所述预置的训练模型的值，所述值满足公式：
    ，其中，所述大于或等于1；
    根据所述值判断训练模型的正确率是否大于预设的阈值；
    若所述训练模型的正确率大于预设的阈值，则确定所述预置的训练模型满足生产需求。
  7. 根据权利要求1所述的人才画像的生成方法，所述根据所述预测标签生成所述目标人才的人才画像之后，所述方法还包括：
    对所述目标人才的人才画像按照预设标准进行评分得到目标分值；
    判断所述目标人才的人才画像的目标分值是否大于预设的分值；
    若所述目标人才的人才画像的目标分值大于预设的分值，则确定所述目标人才的人才画像满足要求，并将所述目标人才的人才画像进行保存；
    若所述目标人才的人才画像的目标分值小于或等于所述预设的分值，则确定所述目标人才的人才画像不满足要求，并将所述目标人才的人才画像进行标记。
  8. 一种人才画像的生成装置，包括：
    获取单元，用于获取目标简历；
    提取单元，用于从所述目标简历中提取简历数据；
    清洗单元，用于基于自然语言处理NLP的文本分类算法对所述简历数据进行清洗，得到目标文本，所述目标文本指示目标人才的核心技能；
    第一生成单元，用于调用预置的训练模型对所述目标文本进行分析，生成预测标签，每个预测标签都对应所述目标人才的一个核心技能；
    第二生成单元，用于根据所述预测标签生成所述目标人才的人才画像。
  9. 根据权利要求8所述的人才画像的生成装置，所述第一生成单元具体用于：
    依次遍历所述目标文本中的每条应聘信息，每条应聘信息都对应所述目标人才的一个核心技能；基于预置的训练模型对每条应聘信息进行分词，得到分词结果；根据所述分词结果在业务词袋中进行检索，得到每条应聘信息的业务关键词；将所述业务关键词按照频数降序排列得到排序结果；将所述排序结果中靠前的预置数目的关键词确定为预测标签。
  10. 根据权利要求8所述的人才画像的生成装置，所述第二生成单元具体用于：
    创建一个目标类，所述目标类包括多个属性；根据所述预测标签确定所述目标类中的各个属性，每个属性对应一个所述预测标签；确定所述目标类的输出结果，所述输出结果包括多个所述预测标签；根据所述输出结果中各个预测标签的标签值生成所述目标人才的人才画像。
  11. 一种人才画像的生成设备，包括存储器、处理器及存储在所述存储器上并可在所述处理器上运行的计算机可读指令，所述处理器执行所述计算机可读指令时实现一种的人才画像的生成方法：
    其中，所述人才画像的生成方法包括以下步骤：
    获取目标简历；
    从所述目标简历中提取简历数据；
    基于自然语言处理NLP的文本分类算法对所述简历数据进行清洗，得到目标文本，所述目标文本指示目标人才的核心技能；
    调用预置的训练模型对所述目标文本进行分析，生成预测标签，每个预测标签都对应所述目标人才的一个核心技能；
    根据所述预测标签生成所述目标人才的人才画像。
  12. 根据权利要求11所述的人才画像的生成设备，所述调用预置的训练模型对所述目标文本进行分析，生成预测标签，每个预测标签都对应所述目标人才的一个核心技能包括：
    依次遍历所述目标文本中的每条应聘信息，每条应聘信息都对应所述目标人才的一个核心技能；
    基于预置的训练模型对每条应聘信息进行分词，得到分词结果；
    根据所述分词结果在业务词袋中进行检索，得到每条应聘信息的业务关键词；
    将所述业务关键词按照频数降序排列得到排序结果；
    将所述排序结果中靠前的预置数目的关键词确定为预测标签。
  13. 根据权利要求11所述的人才画像的生成设备，所述根据所述预测标签生成所述目标人才的人才画像包括：
    创建一个目标类，所述目标类包括多个属性；
    根据所述预测标签确定所述目标类中的各个属性，每个属性对应一个所述预测标签；
    确定所述目标类的输出结果，所述输出结果包括多个所述预测标签；
    根据所述输出结果中各个预测标签的标签值生成所述目标人才的人才画像。
  14. 根据权利要求11所述的人才画像的生成设备，所述基于自然语言处理NLP的文本分类算法对所述简历数据进行清洗，得到目标文本，所述目标文本指示目标人才的核心技能之后，所述调用预置的训练模型对所述目标文本进行分析，生成预测标签，每个预测标签都对应所述目标人才的一个核心技能之前，所述方法还包括：
    生成预置的训练模型，所述预置的训练模型用于根据文本数据生成预测标签。
  15. 根据权利要求14所述的人才画像的生成设备，所述生成预置的训练模型，所述预置的训练模型用于根据文本数据生成预测标签包括：
    获取预置数量的简历数据；
    对所述预置数量的简历数据进行清洗，得到有效数据；
    为每个有效数据确定对应的标签，得到语料库；
    将所述语料库输入到文本训练模型中；
    对所述文本训练模型进行参数调整，生成预置的训练模型，所述预置的训练模型用于根据文本数据生成预测标签。
  16. 一种计算机可读存储介质，包括指令，当所述指令在计算机上运行时，使得计算机执行一种人才画像的生成方法；
    其中，所述人才画像的生成方法包括以下步骤：
    获取目标简历；
    从所述目标简历中提取简历数据；
    基于自然语言处理NLP的文本分类算法对所述简历数据进行清洗，得到目标文本，所述目标文本指示目标人才的核心技能；
    调用预置的训练模型对所述目标文本进行分析，生成预测标签，每个预测标签都对应所述目标人才的一个核心技能；
    根据所述预测标签生成所述目标人才的人才画像。
  17. 根据权利要求16所述的计算机可读存储介质，所述调用预置的训练模型对所述目标文本进行分析，生成预测标签，每个预测标签都对应所述目标人才的一个核心技能包括：
    依次遍历所述目标文本中的每条应聘信息，每条应聘信息都对应所述目标人才的一个核心技能；
    基于预置的训练模型对每条应聘信息进行分词，得到分词结果；
    根据所述分词结果在业务词袋中进行检索，得到每条应聘信息的业务关键词；
    将所述业务关键词按照频数降序排列得到排序结果；
    将所述排序结果中靠前的预置数目的关键词确定为预测标签。
  18. 根据权利要求16所述的计算机可读存储介质，所述根据所述预测标签生成所述目标人才的人才画像包括：
    创建一个目标类，所述目标类包括多个属性；
    根据所述预测标签确定所述目标类中的各个属性，每个属性对应一个所述预测标签；
    确定所述目标类的输出结果，所述输出结果包括多个所述预测标签；
    根据所述输出结果中各个预测标签的标签值生成所述目标人才的人才画像。
  19. 根据权利要求16所述的计算机可读存储介质，所述基于自然语言处理NLP的文本分类算法对所述简历数据进行清洗，得到目标文本，所述目标文本指示目标人才的核心技能之后，所述调用预置的训练模型对所述目标文本进行分析，生成预测标签，每个预测标签都对应所述目标人才的一个核心技能之前，所述方法还包括：
    生成预置的训练模型，所述预置的训练模型用于根据文本数据生成预测标签。
  20. 根据权利要求19所述的计算机可读存储介质，所述生成预置的训练模型，所述预置的训练模型用于根据文本数据生成预测标签包括：
    获取预置数量的简历数据；
    对所述预置数量的简历数据进行清洗，得到有效数据；
    为每个有效数据确定对应的标签，得到语料库；
    将所述语料库输入到文本训练模型中；
    对所述文本训练模型进行参数调整，生成预置的训练模型，所述预置的训练模型用于根据文本数据生成预测标签。
PCT/CN2020/093409 2019-06-18 2020-05-29 人才画像的生成方法、装置、设备及存储介质 WO2020253503A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201910524634.XA CN110399476A (zh) 2019-06-18 2019-06-18 人才画像的生成方法、装置、设备及存储介质
CN201910524634.X 2019-06-18

Publications (1)

Publication Number Publication Date
WO2020253503A1 true WO2020253503A1 (zh) 2020-12-24

Family

ID=68323184

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2020/093409 WO2020253503A1 (zh) 2019-06-18 2020-05-29 人才画像的生成方法、装置、设备及存储介质

Country Status (2)

Country Link
CN (1) CN110399476A (zh)
WO (1) WO2020253503A1 (zh)

Families Citing this family (12)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110399476A (zh) * 2019-06-18 2019-11-01 平安科技（深圳）有限公司 人才画像的生成方法、装置、设备及存储介质
CN111160850A (zh) * 2019-12-11 2020-05-15 浙江微元智能科技股份有限公司 一种适用于人员甄别的匹配系统及其匹配方法
CN111143517B (zh) * 2019-12-30 2023-09-05 浙江阿尔法人力资源有限公司 人选标签预测方法、装置、设备和存储介质
CN111738586B (zh) * 2020-06-17 2024-04-23 中国银行股份有限公司 人才评估方法及装置
CN111754109A (zh) * 2020-06-23 2020-10-09 重庆电子工程职业学院 一种毕业生就业管理系统
CN111813842B (zh) * 2020-09-10 2021-03-05 杭州城市大数据运营有限公司 一种数据处理方法、装置、系统、设备和存储介质
CN112434504A (zh) * 2020-11-23 2021-03-02 京东数字科技控股股份有限公司 生成文件信息的方法、装置、电子设备和计算机可读介质
CN113010727B (zh) * 2021-03-22 2024-02-02 平安科技（深圳）有限公司 直播平台画像的构建方法、装置、设备及存储介质
CN113032441A (zh) * 2021-03-24 2021-06-25 宁波莱博网络科技有限公司 一种基于大数据技术的用户画像分析系统
CN113360733A (zh) * 2021-06-17 2021-09-07 林宏佳 基于人工智能的人才数据标签分类方法、系统及云平台
CN115001856B (zh) * 2022-07-18 2022-10-21 国网浙江省电力有限公司杭州供电公司 基于数据处理的网络安全画像及攻击预测方法
CN114943037A (zh) * 2022-07-20 2022-08-26 平安银行股份有限公司 人才画像的系统化方法、计算机设备及存储介质

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20150268908A1 (en) * 2014-03-18 2015-09-24 Ricoh Company, Ltd. Information processing apparatus, information processing system, and information processing method
CN108764835A (zh) * 2018-05-24 2018-11-06 广州合摩计算机科技有限公司 逆向人才推送信息方法和装置
CN109636337A (zh) * 2018-12-12 2019-04-16 北京唐冠天朗科技开发有限公司 一种基于大数据的人才库构建方法及电子设备
CN109766438A (zh) * 2018-12-12 2019-05-17 平安科技（深圳）有限公司 简历信息提取方法、装置、计算机设备和存储介质
CN110399476A (zh) * 2019-06-18 2019-11-01 平安科技（深圳）有限公司 人才画像的生成方法、装置、设备及存储介质

Family Cites Families (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2016003742A1 (en) * 2014-07-01 2016-01-07 Piazza Technologies, Inc. Computer systems and user interfaces for learning, talent discovery, relationship management, and campaign development
CN105159962B (zh) * 2015-08-21 2018-08-17 北京全聘致远科技有限公司 职位推荐方法与装置、简历推荐方法与装置、招聘平台
CN109492164A (zh) * 2018-11-26 2019-03-19 北京网聘咨询有限公司 一种简历的推荐方法、装置、电子设备及存储介质

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20150268908A1 (en) * 2014-03-18 2015-09-24 Ricoh Company, Ltd. Information processing apparatus, information processing system, and information processing method
CN108764835A (zh) * 2018-05-24 2018-11-06 广州合摩计算机科技有限公司 逆向人才推送信息方法和装置
CN109636337A (zh) * 2018-12-12 2019-04-16 北京唐冠天朗科技开发有限公司 一种基于大数据的人才库构建方法及电子设备
CN109766438A (zh) * 2018-12-12 2019-05-17 平安科技（深圳）有限公司 简历信息提取方法、装置、计算机设备和存储介质
CN110399476A (zh) * 2019-06-18 2019-11-01 平安科技（深圳）有限公司 人才画像的生成方法、装置、设备及存储介质

Also Published As

Publication number Publication date
CN110399476A (zh) 2019-11-01

Similar Documents

Publication Publication Date Title
WO2020253503A1 (zh) 人才画像的生成方法、装置、设备及存储介质
CN108804512B (zh) 文本分类模型的生成装置、方法及计算机可读存储介质
CN112632385B (zh) 课程推荐方法、装置、计算机设备及介质
CA3129745C (en) Neural network system for text classification
US10713432B2 (en) Classifying and ranking changes between document versions
US20200210957A1 (en) Classification of job titles via machine learning
US11023503B2 (en) Suggesting text in an electronic document
US11354608B2 (en) Organization analysis platform for workforce recommendations
CN111177538A (zh) 一种基于无监督权值计算的用户兴趣标签构建方法
US11640420B2 (en) System and method for automatic summarization of content with event based analysis
CN113627797B (zh) 入职员工画像生成方法、装置、计算机设备及存储介质
CN112926308B (zh) 匹配正文的方法、装置、设备、存储介质以及程序产品
CN116882414B (zh) 基于大规模语言模型的评语自动生成方法及相关装置
Palshikar et al. Automatic Shortlisting of Candidates in Recruitment.
CN115248890A (zh) 用户兴趣画像的生成方法、装置、电子设备以及存储介质
WO2023164312A1 (en) An apparatus for classifying candidates to postings and a method for its use
US20220318253A1 (en) Search Method, Apparatus, Electronic Device, Storage Medium and Program Product
US11580499B2 (en) Method, system and computer-readable medium for information retrieval
CN112287215A (zh) 一种智能就业推荐方法和装置
JP2021149682A (ja) 学習装置、学習方法及び学習プログラム
JP2021149681A (ja) 判定装置、判定方法及び判定プログラム
CN110717008A (zh) 基于语意识别的搜索结果排序方法及相关装置
CN112954025B (zh) 基于分层知识图谱推送信息方法、装置、设备、介质
KR102600307B1 (ko) 사용자별 대표 활동 로그 추출 방법과 머신러닝 모델의 병합을 적용한 사용자 및 대화 파트너 매칭 제공 최적화 시스템 및 그 동작 방법
CN111539208B (zh) 语句处理方法和装置、以及电子设备和可读存储介质

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 20827765

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 20827765

Country of ref document: EP

Kind code of ref document: A1