Talent information processing method, device, computer equipment and readable storage medium

Abstract

The present invention proposes a talent information processing method, device, computer equipment and readable storage medium. The talent information processing method includes: determining the M basic label information possessed by any talent data in the talent database; determining the upper-level label information used to summarize the N basic label information, where N is less than or equal to M; associating any talent data with the upper-level label information. The technical solution of the present invention adds upper-level labels corresponding to the talent data to the talent database, avoiding the problem in the human resources system of the related technology that when searching for upper-level labels, multiple basic labels need to be extracted together due to the lack of upper-level labels in the human resources system. This not only improves the accuracy of the searched information, but also can provide more satisfactory talent data when searching for talent data.

Description

Technical Field

[0001] The present application relates to the field of software development, and in particular, to a talent information processing method, a talent information processing apparatus, a computer device, and a readable storage medium. Background Art

[0002] In the related art, the labels of talents in the talent pool are fixed, relatively simple, and not general. With the diversification of talent labels and the complexity and personalization of search requirements, the existing labels cannot meet the actual talent information search needs. Summary of the invention

[0003] The present application aims to solve at least one of the technical problems existing in the prior art or related art.

[0004] To this end, the first aspect of the present invention proposes a talent information processing method.

[0005] A second aspect of the present invention provides a talent information processing device.

[0006] A third aspect of the present invention provides a computer device.

[0007] A fourth aspect of the present invention provides a readable storage medium.

[0008] In view of this, according to the first aspect of the present invention, a talent information processing method is proposed, including: determining M basic label information possessed by any talent data in a talent database; determining upper-level label information used to summarize N basic label information, wherein N is less than or equal to M; and associating any talent data with the upper-level label information.

[0009] In this technical solution, any talent data is extracted from the talent database, the M basic labels corresponding to the talent data are determined, and the talent data in the talent database are relabeled to form basic labels. N basic label information is extracted from the M basic label information determined above, and the N basic label information is summarized, wherein N is a positive integer less than or equal to M, and N is a positive integer greater than or equal to 1, and any talent data mentioned above is associated with the upper-level label information, and the upper-level label corresponding to the talent data is added to the talent database, avoiding the problem in the human resources system of the related technology that when searching for the upper-level label, multiple basic labels need to be extracted together due to the lack of upper-level labels in the human resources system, which not only improves the accuracy of the searched information, but also can provide more satisfactory talent data when searching for talent data.

[0010] It should be noted that the numbers M and N of the above basic tag information are not fixed values, but can be changed with the talent data in the talent database and the received user search information.

[0011] The talent information processing method according to the present invention may also have the following additional technical features:

[0012] In the above technical solution, determining the upper-level label information for summarizing the N basic label information specifically includes: inputting the N basic label information into a pre-trained target classification model to obtain the upper-level label information.

[0013] In this technical solution, a pre-trained target classification model is pre-stored in the talent database, and the N basic label information is summarized by the target classification model to obtain the corresponding upper label information. The technical solution of the present invention, through the above-mentioned method of summarizing the basic label information, not only ensures the matching degree of the acquired upper label information with the basic label information, so as to ensure the accuracy of the upper label information, but also ensures the formation of the diversity of the upper label information, thereby improving the diversity of labels in the talent database.

[0014] In any of the above technical solutions, the basic label information includes at least one of the following labels: gender label, nationality label, health label, education label, professional title label, industry label, age label, and achievement label.

[0015] In this technical solution, the basic label information is limited to one or more of the following labels: gender label, nationality label, health label, education label, professional title label, industry label, age label, and achievement label. This provides a basis for label content in order to determine the basic label corresponding to any talent data in the talent database, so as to ensure the accuracy and reliability of the talent data cleansing process in the talent database.

[0016] It should be noted that the basic label information of the present invention is not limited to the above-mentioned label information, but can be adaptively modified and supplemented according to the talent data information in the talent pool or the search information input by the user. For example, the information on the overseas work and study experience such as the number of years studied abroad, the number of years lived abroad, and the number of years worked abroad in the talent data information can be used to form a basic label "overseas background label".

[0017] In any of the above technical solutions, any talent data includes one or more of reward information, patent achievement information and scientific research achievement information. Based on M basic label information including achievement labels, the M basic label information possessed by any talent data in the talent database is determined, including: using the textRank keyword extraction algorithm of the word graph model to extract keywords from the reward information, patent achievement information and scientific research achievement information to obtain achievement labels.

[0018] In this technical solution, any talent data is extracted from the talent database. When the talent data includes one or more of reward information, patent achievement information and scientific research achievement information, and the M basic label information corresponding to the talent data includes an achievement label, the textRank keyword extraction algorithm of the word graph model is used to extract keywords related to the reward information, patent achievement information and scientific research achievement information in the above talent data, and the extracted keywords are used as achievement labels. The above achievement label generation method does not require learning multiple documents in advance, and is very suitable for extracting keywords such as reward information, patent information, scientific research achievement information, etc., ensuring that when determining the basic label for any talent data, the information related to the achievement involved in any talent data can be accurately converted into an achievement label, thereby ensuring that the talent database can form achievement labels more accurately, and then ensuring the accuracy of the basic label information in the talent database.

[0019] In any of the above technical solutions, any talent data includes date of birth information. Based on M basic label information including an age label, the M basic label information possessed by any talent data in the talent database is determined, including: obtaining the division relationship between the date of birth information and the age group; according to the division relationship, determining the age group corresponding to the date of birth information to obtain the age label.

[0020] In this technical solution, relevant data on the division of age groups are pre-stored in the talent database. When any talent data extracted from the talent database includes date of birth information, and the M basic label information corresponding to the any talent data includes an age label, the division is performed according to the correspondence between the date of birth information and the age group, and then the age group corresponding to the date of birth information is determined based on the above division result, and the age group is used as the age label of the date of birth information in the above any talent data, thereby determining a method for generating an age label for any talent data from the talent database, so as to ensure that when determining the basic label for any talent data, the age-related information involved in the above any talent data can be accurately converted into an age label, thereby ensuring that the talent database can form age labels more accurately and quickly, so as to ensure the accuracy of the basic label information in the talent database.

[0021] In any of the above technical solutions, the talent information processing method also includes: using the Bert model to encode M basic label information and N upper-level label information to obtain a vector distribution corresponding to any talent data; associating the vector distribution with any talent data; and receiving talent search information, performing unsupervised text matching between the talent search information and the vector distribution corresponding to any talent data in the talent database to output the target talent data.

[0022] In this technical solution, the M basic label information and N upper label information determined above are encoded using the Bert model to obtain a vector distribution corresponding to any talent data, and the above vector distribution is associated with any talent data, so that the talent database can convert any talent data into a corresponding vector distribution, so that the talent database can realize vector-based search, avoiding the problems in related technologies such as low search recall rate due to the ambiguity and diversity of language expression when searching based on keywords, and low search accuracy caused by different semantics of keywords in different sentences. After receiving the talent search information, the talent search information is matched with the vector distribution corresponding to any talent data in the talent database by unsupervised text matching to adapt to the randomness of the search information input by the user, and output more accurate target talent data for the user, thereby improving the user's satisfaction with the searched information and the user's search recall rate.

[0023] In any of the above technical solutions, the talent information processing method also includes: using a standardized flow method to perform a Gaussian transformation on a vector distribution corresponding to any talent data.

[0024] In this technical solution, after using the Bert model to encode M basic label information and N upper-level label information to obtain the vector distribution corresponding to any talent data, the vector distribution corresponding to any of the above talent data is converted into a smooth, isotropic standard Gaussian vector distribution by using a standardized flow method, and the Gaussian vector distribution is associated with any of the above talent data to improve the representation ability of the vector semantic similarity in the vector distribution, thereby improving the speed and accuracy of the search, and further enhancing the search effect.

[0025] According to a second aspect of the present invention, a talent information processing device is proposed, comprising: a first determination unit, used to determine M basic label information possessed by any talent data in a talent database; a second determination unit, used to determine upper-level label information used to summarize N basic label information, wherein N is less than or equal to M; and an association unit, used to associate any talent data with the upper-level label information.

[0026] In this technical solution, the first determination unit extracts any talent data from the talent database, determines the M basic labels corresponding to the talent data, and relabels the talent data in the talent database to form basic labels. The second determination unit extracts N basic label information from the above-determined M basic label information, and summarizes the N basic label information, wherein N is a positive integer less than or equal to M, and N is a positive integer greater than or equal to 1. And any of the above talent data is associated with the upper-level label information through an association unit, and the upper-level label corresponding to the talent data is added to the talent database, avoiding the problem in the human resources system of the related technology that when searching for the upper-level label, multiple basic labels need to be extracted together due to the lack of upper-level labels in the human resources system, which not only improves the accuracy of the searched information, but also can provide more satisfactory talent data when searching for talent data.

[0027] It should be noted that the numbers M and N of the above basic tag information are not fixed values, but can be changed with the talent data in the talent database and the received user search information.

[0028] In the above technical solution, the second determination unit is used to summarize the upper-level label information of N basic label information, and is specifically used to: input the N basic label information into a pre-trained target classification model to obtain the upper-level label information.

[0029] In this technical solution, a pre-trained target classification model is pre-stored in the talent database, and the target classification model is used to summarize N basic label information through the second determination unit to obtain the corresponding upper label information. The technical solution of the present invention, through the above-mentioned method of summarizing the basic label information, not only ensures the matching degree between the acquired upper label information and the basic label information to ensure the accuracy of the upper label information, but also ensures that the acquired upper label information can be adaptively modified based on the basic label information and the language environment, thereby ensuring the formation of the diversity of the upper label information, and thus improving the diversity of labels in the talent database.

[0030] In any of the above technical solutions, the basic label information includes at least one of the following labels: gender label, nationality label, health label, education label, professional title label, industry label, age label, and achievement label.

[0031] In this technical solution, the basic label information in the talent information processing device is limited to one or more of the following labels: gender label, nationality label, health label, education label, professional title label, industry label, age label, and achievement label. This provides a basis for label content in order to determine the basic label corresponding to any talent data in the talent database, so as to ensure the accuracy and reliability of the talent data cleansing process in the talent database.

[0032] It should be noted that the basic label information of the present invention is not limited to the above-mentioned label information, but can be adaptively modified and supplemented according to the talent data information in the talent pool or the search information input by the user. For example, the information on the overseas work and study experience such as the number of years studied abroad, the number of years lived abroad, and the number of years worked abroad in the talent data information can be used to form a basic label "overseas background label".

[0033] In any of the above technical solutions, any talent data includes one or more of reward information, patent achievement information and scientific research achievement information. Based on M basic label information including achievement labels, the M basic label information possessed by any talent data in the talent database is determined, including: using the textRank keyword extraction algorithm of the word graph model to extract keywords from the reward information, patent achievement information and scientific research achievement information to obtain achievement labels.

[0034] In this technical solution, any talent data is extracted from the talent database of the talent information processing device. When the talent data includes one or more of reward information, patent achievement information and scientific research achievement information, and the M basic label information corresponding to the talent data includes an achievement label, the textRank keyword extraction algorithm of the word graph model is used to extract keywords related to the reward information, patent achievement information and scientific research achievement information in the above talent data, and the extracted keywords are used as achievement labels. Thus, it is determined that for any talent data from the talent database, the above-mentioned method of generating achievement labels does not require learning multiple documents in advance, and is very suitable for extracting keywords such as reward information, patent information, scientific research achievement information, etc., so as to ensure that when determining the basic label for any talent data, the information related to the achievement involved in any talent data can be accurately converted into an achievement label, thereby ensuring that the talent database can form achievement labels more accurately, and then ensuring the accuracy of the basic label information in the talent database.

[0035] In any of the above technical solutions, any talent data includes date of birth information. Based on M basic label information including an age label, the M basic label information possessed by any talent data in the talent database is determined, including: obtaining the division relationship between the date of birth information and the age group; according to the division relationship, determining the age group corresponding to the date of birth information to obtain the age label.

[0036] In this technical solution, relevant data on the division of age groups are pre-stored in the talent database. When any talent data extracted from the talent database includes date of birth information, and the M basic label information corresponding to the any talent data includes an age label, the division is performed according to the correspondence between the date of birth information and the age group, and then the age group corresponding to the date of birth information is determined based on the above division result, and the age group is used as the age label of the date of birth information in the above any talent data, thereby determining a method for generating an age label for any talent data from the talent database, so as to ensure that when determining the basic label for any talent data, the age-related information involved in the above any talent data can be accurately converted into an age label, thereby ensuring that the talent database can form age labels more accurately and quickly, so as to ensure the accuracy of the basic label information in the talent database.

[0037] In any of the above technical solutions, the talent information processing device also includes: an encoding unit, used to use the Bert model to encode M basic label information and N upper-level label information to obtain a vector distribution corresponding to any talent data; an association unit, also used to associate the vector distribution with any talent data; a matching unit, used to receive talent search information, and perform unsupervised text matching between the talent search information and the vector distribution corresponding to any talent data in the talent database to output the target talent data.

[0038] In this technical solution, the M basic label information and N upper label information determined above are encoded by the Bert model through the encoding unit, so as to obtain a vector distribution corresponding to any talent data. Through the association unit, the above vector distribution is associated with any talent data, so that the talent database can convert any talent data into a corresponding vector distribution, so that the talent database can realize vector-based search, avoiding the problems in related technologies, such as low search recall rate due to the ambiguity and diversity of language expression when searching based on keywords, and keywords have different semantics in different sentences, resulting in low search accuracy. After receiving the talent search information, the matching unit performs unsupervised text matching on the talent search information and the vector distribution corresponding to any talent data in the talent database, so as to adapt to the randomness of the search information input by the user, and output more accurate target talent data for the user, thereby improving the user's satisfaction with the searched information and the user's search recall rate.

[0039] In any of the above technical solutions, the talent information processing device also includes: a vector conversion unit, which is used to perform Gaussian transformation on the vector distribution corresponding to any talent data by adopting a standardized flow method.

[0040] In this technical solution, through a vector conversion unit, after using the Bert model to encode M basic label information and N upper-level label information to obtain the vector distribution corresponding to any talent data, a standardized flow method is used to convert the vector distribution corresponding to any of the above talent data into a smooth, isotropic standard Gaussian vector distribution, and the Gaussian vector distribution is associated with any of the above talent data to improve the representation ability of the vector semantic similarity in the vector distribution, thereby improving the speed and accuracy of the search, and further enhancing the search effect.

[0041] According to a third aspect of the present invention, a computer device is proposed, comprising: a memory storing a computer program; and a controller executing the computer program to implement the steps of the talent information processing method in any of the above technical solutions.

[0042] Specifically, the memory may include a large capacity memory for data or instructions. By way of example and not limitation, the memory may include a hard disk drive (HDD), a floppy disk drive, a flash memory, an optical disk, a magneto-optical disk, a magnetic tape, or a universal serial bus (USB) drive, or a combination of two or more of these. Where appropriate, the memory may include a removable or non-removable (or fixed) medium. Where appropriate, the memory may be inside or outside the integrated gateway disaster recovery device. In a specific embodiment, the memory is a non-volatile solid-state memory. In a specific embodiment, the memory includes a read-only memory (ROM). Where appropriate, the ROM may be a mask-programmed ROM, a programmable ROM (PROM), an erasable PROM (EPROM), an electrically erasable PROM (EEPROM), an electrically rewritable ROM (EAROM), or a flash memory, or a combination of two or more of these. The above-mentioned processor may include a central processing unit (CPU), or an application specific integrated circuit (ASIC), or may be configured to implement one or more integrated circuits of an embodiment of the present invention.

[0043] According to a fourth aspect of the present invention, a readable storage medium is proposed, on which a program or instruction is stored, and when the program or instruction is executed by a processor, the steps of the talent information processing method in any of the above technical solutions are implemented.

[0044] The readable storage medium provided by the present invention implements the steps of the talent information processing method of any of the above-mentioned technical solutions when the processor executes the computer program. Therefore, the readable storage medium includes all the beneficial effects of the talent information processing method of any of the above-mentioned technical solutions.

[0045] Additional aspects and advantages of the present invention will become apparent from the following description or may be learned by practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] The above and / or additional aspects and advantages of the present invention will become apparent and easily understood from the description of the embodiments in conjunction with the following drawings, in which:

[0047] Figure 1 One of the flowcharts of the talent information processing method according to an embodiment of the present invention is shown;

[0048] Figure 2 A second flow chart of a method for processing talent information according to an embodiment of the present invention is shown;

[0049] Figure 3A third flowchart of a talent information processing method according to an embodiment of the present invention is shown;

[0050] Figure 4 A fourth flowchart of a talent information processing method according to an embodiment of the present invention is shown;

[0051] Figure 5 A schematic diagram of a data processing flow in a talent database according to an embodiment of the present invention is shown;

[0052] Figure 6 A schematic diagram showing the structure of a talent information processing device according to an embodiment of the present invention is shown;

[0053] Figure 7 A schematic block diagram of a computer device is shown according to one embodiment of the present invention.

[0054] in, Figure 6 and Figure 7 The corresponding relationship between the reference numerals and the component names is as follows:

[0055] 600 talent information processing device, 602 first determination unit, 604 second determination unit, 606 association unit, 608 encoding unit, 610 matching unit, 612 vector conversion unit, 700 computer device, 702 memory, 704 controller. DETAILED DESCRIPTION

[0056] In order to more clearly understand the above-mentioned purpose, features and advantages of the present invention, the present invention is further described in detail below in conjunction with the accompanying drawings and specific embodiments. It should be noted that the embodiments of the present application and the features in the embodiments can be combined with each other without conflict.

[0057] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways different from those described herein. Therefore, the protection scope of the present invention is not limited to the specific embodiments disclosed below.

[0058] The first aspect of the present invention provides a talent information processing method, and the talent information processing method is described in detail through the following embodiments 1 to 3 and specific embodiments.

[0059] Embodiment 1

[0060] Figure 1 FIG. 1 is a flow chart showing a method for processing talent information according to an embodiment of the present invention. Figure 1 As shown, the talent information processing method includes:

[0061] Step 102, determining M basic label information of any talent data in the talent database;

[0062] Step 104, determining upper-level label information for summarizing N basic label information, where N is less than or equal to M;

[0063] Step 106: associate any talent data with upper-level label information.

[0064] In this embodiment, any talent data is extracted from the talent database, the M basic labels corresponding to the talent data are determined, and the talent data in the talent database are relabeled to form basic labels to complete the cleaning of the talent data in the talent database. Further, the upper label information of N basic label information among the M basic label information is determined, and N basic label information is extracted from the above-determined M basic label information, and the N basic label information is summarized, wherein N is a positive integer less than or equal to M, and N is a positive integer greater than or equal to 1, and any talent data mentioned above is associated with the upper label information, and the upper label corresponding to the talent data is determined to be added in the talent database, thereby avoiding the problem in the human resource system of the related technology that when searching for the upper label, due to the lack of upper labels in the human resource system, multiple basic labels need to be extracted together, thereby improving the relevance of the label information in the talent database with any talent data, thereby not only improving the accuracy of the searched information to meet the user's demand for diversified and complex talent labels, but also providing more satisfactory talent data when searching for talent data.

[0065] It should be noted that the numbers M and N of the above basic tag information are not fixed values, but can be changed with the talent data in the talent database and the received user search information.

[0066] In the above embodiment, determining the upper-level label information for summarizing the N basic label information specifically includes: inputting the N basic label information into a pre-trained target classification model to obtain the upper-level label information.

[0067] In this embodiment, a pre-trained target classification model is pre-stored in the talent database, and the N basic label information is summarized by the target classification model to obtain the corresponding upper label information. The embodiment of the present invention, through the above-mentioned method of summarizing the basic label information, not only ensures the matching degree between the acquired upper label information and the basic label information to ensure the accuracy of the upper label information, but also ensures that the acquired upper label information can be adaptively modified based on the basic label information and the language environment, thereby ensuring the formation of the diversity of the upper label information, thereby improving the diversity of labels in the talent database.

[0068] In any of the above embodiments, the basic label information includes at least one of the following labels: a gender label, a nationality label, a health label, an education label, a professional title label, an industry label, an age label, and an achievement label.

[0069] In this embodiment, the basic label information is limited to one or more of the following labels: gender label, nationality label, health label, education label, professional title label, industry label, age label, and achievement label. This provides a basis for label content for determining the basic label corresponding to any talent data in the talent database, so as to ensure the accuracy and reliability of the talent data cleansing process in the talent database.

[0070] It should be noted that the basic label information of the present invention is not limited to the above-mentioned label information, but can be adaptively modified and supplemented according to the talent data information in the talent pool or the search information input by the user. For example, the information on the overseas work and study experience such as the number of years studied abroad, the number of years lived abroad, and the number of years worked abroad in the talent data information can be used to form a basic label "overseas background label".

[0071] In addition, at least one of the gender label, nationality label, health label, education label, job title label, industry label, age label, and achievement label may be selected to generate upper-level label information.

[0072] In some embodiments, the upper label information includes: scientific and technological innovation leaders, group company technical experts, international talents, young scientific and technological talents, high-level talents, etc. High-level overseas students and overseas scientific and technological experts are collectively referred to as high-level talents, among which "high-level overseas students" refer to those who study abroad on government or self-funded by my country, and after graduation, they engage in scientific research, teaching, engineering technology, finance, management and other work overseas and have achieved remarkable results, and are senior management talents, senior professional and technical talents, academic and technical leaders urgently needed in China, and talents with patents, inventions or professional technologies with good industrial development prospects.

[0073] For example: Age label: between 35 and 40 years old, Education label: studying abroad on government sponsorship or at one’s own expense, Professional title label: academic and technical leaders, senior professional and technical personnel, Achievement label: personnel with patents, inventions or professional and technical skills with good industrial development prospects; Based on the above basic label information, the upper-level label information is generated as young high-level talents.

[0074] In some embodiments, the educational label is a bachelor's degree or above, the professional title label is an intermediate professional title or above, the industry label is clinical medicine, the age label is 40 to 50 years old, the overseas background label is studying abroad for more than two years, the achievement label is winning multiple international honors, and the higher-level label that can be obtained is: the company's middle-aged overseas medical expert.

[0075] In any of the above embodiments, any talent data includes one or more of reward information, patent achievement information and scientific research achievement information. Based on M basic label information including achievement labels, the M basic label information possessed by any talent data in the talent database is determined, including: using the textRank keyword extraction algorithm of the word graph model to extract keywords from the reward information, patent achievement information and scientific research achievement information to obtain achievement labels.

[0076] In this embodiment, when any talent data is extracted from the talent database, and when the talent data includes one or more of reward information, patent achievement information, and scientific research achievement information, and the M basic label information corresponding to the talent data includes an achievement label, the textRank keyword extraction algorithm of the word graph model is used to extract keywords related to the reward information, patent achievement information, and scientific research achievement information in the talent data, and the extracted keywords are used as achievement labels. Thus, it is determined that for any talent data from the talent database, the above-mentioned method of generating achievement labels does not require learning multiple documents in advance, and is very suitable for extracting keywords such as reward information, patent information, and scientific research achievement information, so as to ensure that when determining the basic label for any talent data, the information related to the achievement involved in any talent data can be accurately converted into an achievement label, thereby ensuring that the talent database can form achievement labels more accurately, and then ensuring the accuracy of the basic label information in the talent database.

[0077] In any of the above embodiments, any talent data includes date of birth information, and based on M basic label information including an age label, determining the M basic label information possessed by any talent data in the talent database includes: obtaining the division relationship between the date of birth information and the age group; and determining the age group corresponding to the date of birth information based on the division relationship to obtain the age label.

[0078] In this embodiment, the talent database pre-stores relevant data on the division of age groups. When any talent data extracted from the talent database includes date of birth information, and the M basic label information corresponding to the any talent data includes an age label, the division is performed according to the correspondence between the date of birth information and the age group, and then the age group corresponding to the date of birth information is determined based on the above division result, and the age group is used as the age label of the date of birth information in the above any talent data, thereby determining the method for generating an age label for any talent data from the talent database, so as to ensure that when determining the basic label for any talent data, the age-related information involved in the above any talent data can be accurately converted into the age label, thereby ensuring that the talent database can form the age label more accurately and quickly, so as to ensure the accuracy of the basic label information in the talent database.

[0079] Embodiment 2

[0080] Figure 2 FIG2 is a second flow chart of a talent information processing method according to an embodiment of the present invention. Figure 2 As shown, the talent information processing method includes:

[0081] Step 202, determining M basic label information of any talent data in the talent database;

[0082] Step 204, determining upper-level label information for summarizing N basic label information, where N is less than or equal to M;

[0083] Step 206, associating any talent data with upper-level label information;

[0084] Step 208: using the Bert model, encode the M basic label information and the N upper label information to obtain a vector distribution corresponding to any talent data;

[0085] Step 210, associating the vector distribution with any talent data;

[0086] Step 212, and receiving talent search information, performing unsupervised text matching between the talent search information and the vector distribution corresponding to any talent data in the talent database to output target talent data.

[0087] In this embodiment, the M basic label information and N upper label information determined above are encoded using the Bert model to obtain a vector distribution corresponding to any talent data, and the above vector distribution is associated with any talent data, so that the talent database can convert any talent data into a corresponding vector distribution, so that the talent database can realize vector-based search, avoiding the problems in related technologies such as low search recall rate due to the ambiguity and diversity of language expression when searching based on keywords, and low search accuracy due to different semantics of keywords in different sentences. After receiving the talent search information, the talent search information is matched with the vector distribution corresponding to any talent data in the talent database by unsupervised text, so as to adapt to the randomness of the search information input by the user, and output more accurate target talent data for the user, thereby improving the user's satisfaction with the searched information and the user's search recall rate.

[0088] Embodiment 3

[0089] Figure 3 FIG3 is a flow chart showing a method for processing talent information according to an embodiment of the present invention. Figure 3 As shown, the talent information processing method includes:

[0090] Step 302, determining M basic label information of any talent data in the talent database;

[0091] Step 304, determining upper-level label information for summarizing N basic label information, where N is less than or equal to M;

[0092] Step 306, associating any talent data with upper-level label information;

[0093] Step 308: using the Bert model, encode the M basic label information and the N upper label information to obtain the vector distribution corresponding to any talent data;

[0094] Step 310, using a standardized flow method to perform Gaussian transformation on the vector distribution corresponding to any talent data;

[0095] Step 312, associating the vector distribution with any talent data;

[0096] Step 314, and receiving talent search information, performing unsupervised text matching between the talent search information and the vector distribution corresponding to any talent data in the talent database to output target talent data.

[0097] In this embodiment, after using the Bert model to encode M basic label information and N upper-level label information to obtain the vector distribution corresponding to any talent data, further, using a standardized flow method, the vector distribution corresponding to any of the above talent data is converted into a smooth, isotropic standard Gaussian vector distribution, and the above Gaussian vector distribution is associated with any of the above talent data to improve the representation capability of the vector semantic similarity in the vector distribution, thereby improving the speed and accuracy of the search, and further enhancing the search effect.

[0098] Specifically, taking the user input "I want to find an unmarried male with rich overseas experience in the oil field" as an example, the most matching person searched in the talent database with the talent information processing method of the present invention is: FX0027: China, single, Xinjiang, experienced, University of Aberdeen, UK, Beijing Jiaotong University, national talent, oil and gas engineering, master's degree, proficient in English, proficient in spoken English, mining industry, oil and gas engineering department, supervisor, outstanding employee of 2014. It can be seen that this is exactly the tag that the user wants to search. Specific embodiment:

[0100] In the related art, the human resources system stores basic tags, such as name, gender, age, awards received, university, foreign language, etc. Searches based on these tags are usually based on keyword searches. However, with the diversification of talent tags and the complexity and personalization of search requirements, keyword searches based on basic subordinate information are no longer applicable. For example, a human resources specialist searches in the talent system: I want to find senior IT talents with overseas experience. There are two problems: (1) Senior tags such as overseas experience and IT talent do not exist in the human resources system, and the overseas experience super-tag requires multiple subordinate tags to be extracted together; (2) The recall rate of keyword-based searches is low due to the ambiguity and diversity of language expressions. At the same time, keywords have different semantics in different sentences, resulting in reduced search accuracy.

[0101] In order to solve the above-mentioned technical problems, the present invention innovatively extracts labels organically from complex original personnel information through rules, keyword extraction, supervised upper-level label extraction, etc., adopts different methods to obtain labels for different personnel information, and establishes a complete system from the bottom-level personnel information to the formation of enterprise employee portrait labels, and automatically extracts upper-level label information from the basic label information; and based on the use of Bert encoding, standard Gaussian distribution of label vectors and unsupervised text matching, talent search is carried out to improve the talent search capability.

[0102] Specifically, Figure 4 FIG4 is a flow chart of a talent information processing method according to an embodiment of the present invention. Figure 4 As shown, the talent information processing method includes:

[0103] Step 402, extracting and cleaning the talent data in the original talent database to obtain basic label information;

[0104] Step 404, generating upper-level label information according to the basic label information;

[0105] Step 406, performing Bert model encoding on the basic label information and the upper label information to obtain a vector distribution;

[0106] Step 408, performing Gaussian transformation on the vector distribution according to the standardized flow to obtain a standard Gaussian vector distribution;

[0107] Step 410 , performing unsupervised text matching between the talent search information input by the user and the standard Gaussian vector distribution corresponding to the talent data.

[0108] In this embodiment, first, data extraction and data cleaning are performed on the talent data in the original talent database. Specifically, labels are extracted from the original talent database and data cleaning is performed. The labels in the original talent database are manually entered and stored in the MySQL database. The above-mentioned label extraction and data cleaning are as follows: (1) Gender labels, nationality labels, health status labels and other labels remain unchanged; (2) For labels such as date of birth, they are converted into age according to rules, corresponding to: young people, middle-aged people, and people about to retire; (3) For labels such as award information, patent information, and scientific research results, keywords are extracted, and the textRank keyword extraction algorithm based on the word graph model is adopted. It does not require prior learning and training of multiple documents, and is very suitable for keyword extraction of award information, patent information, scientific research results, etc.

[0109] Secondly, the upper-level label information is classified. The upper-level label information classification is mainly based on the basic label to obtain the upper-level label. For example: according to the extracted basic label, the label information is as follows: bottom-level label a, bottom-level label b, bottom-level label c, bottom-level label d, bottom-level label e, among which bottom-level label a can be an academic degree label, bottom-level label b can be a professional title label, bottom-level label c can be an industry label, bottom-level label d can be an age label, and bottom-level label e can be an achievement label. When bottom-level label a is a bachelor's degree or above, bottom-level label b is an intermediate or above professional title, bottom-level label c is engaged in exploration, development and mining, and marine engineering in the major professional category, bottom-level label d is no more than 35 years old, and bottom-level label e is obtained when one or more related patents or group-level honors are obtained, the upper-level label that can be obtained is: young technical talents of the company.

[0110] In addition, this method adopts a multi-classification method based on a pre-trained model, which can classify high-level labels such as leading talents in scientific and technological innovation, technical experts of group companies, international talents, young scientific and technological talents, and "double-high" talents.

[0111] Secondly, the Bert model is used to encode and form a standard Gaussian vector distribution. Specifically, the Bert model is used for encoding, and the vector distribution is transformed into a smooth, isotropic standard Gaussian distribution using a normalized flow, and the above vector distribution is stored to improve the representation ability of the semantic similarity of the sentence vector and enhance the search effect.

[0112] Finally, Bert is used for unsupervised text matching. Specifically, since the input is natural language, containing the semantics of labels, and there are a large number of labels and the search language is arbitrary, the talent search information input by the user is matched with the standard Gaussian vector distribution corresponding to the talent data for unsupervised text matching, so as to search for the target talent data that matches the above user input information, so as to achieve natural language semantic matching for the search personnel and search for the corresponding labels.

[0113] Figure 5 FIG. 1 is a schematic diagram showing a data processing flow in a talent database according to an embodiment of the present invention. Figure 5 As shown in the figure, the specific process of data processing in the talent database is as follows:

[0114] Step 502: Acquire talent data in a talent database;

[0115] Step 504: Talent data extraction and data cleaning;

[0116] Step 506: Classify the upper level information;

[0117] Step 508: Bert model encoding;

[0118] Step 510: standard Gaussian vector distribution;

[0119] Step 512: Perform space vector search.

[0120] Specifically, the talent data in the original talent database is first extracted and cleaned, and then the upper-level label information is classified (the upper-level label information classification is to generate the upper-level label information according to the basic label information), and then the Bert model is used to encode and form a standard Gaussian vector distribution. Finally, the vector space is used for search, and Bert is used for unsupervised text matching.

[0121] The second aspect of the embodiment of the present invention proposes a talent information processing device 600, including: a first determination unit 602, used to determine M basic label information possessed by any talent data in a talent database; a second determination unit 604, used to determine the upper-level label information used to summarize the N basic label information, wherein N is less than or equal to M; an association unit 606, used to associate any talent data with the upper-level label information.

[0122] In this embodiment, if Figure 6 As shown, a schematic diagram of the structure of a talent information processing device according to an embodiment of the present invention is shown, and the talent information processing device 600 includes: a first determination unit 602, a second determination unit 604, and an association unit 606. The first determination unit 602 extracts any talent data from the talent database, determines the M basic labels corresponding to the talent data, and relabels the talent data in the talent database to form basic labels. The second determination unit 604 extracts N basic label information from the M basic label information determined above, and summarizes the N basic label information, wherein N is a positive integer less than or equal to M, and N is a positive integer greater than or equal to 1. And any talent data mentioned above is associated with the upper label information through the association unit 606, and the upper label corresponding to the talent data is added to the talent database, thereby avoiding the problem in the human resource system of the related technology that when searching for the upper label, multiple basic labels need to be extracted together due to the lack of upper labels in the human resource system, which not only improves the accuracy of the searched information, but also provides more satisfactory talent data when searching for talent data. It should be noted that the numbers M and N of the above basic tag information are not fixed values, but can be changed with the talent data in the talent database and the received user search information.

[0123] In the above embodiment, the second determining unit 604 is used to summarize the upper label information of the N basic label information, and is specifically used to: input the N basic label information into the pre-trained target classification model to obtain the upper label information.

[0124] In this embodiment, a pre-trained target classification model is pre-stored in the talent database, and the target classification model is used to summarize N basic label information to obtain the corresponding upper label information through the second determination unit 604. In the embodiment of the present invention, the above-mentioned method of summarizing the basic label information not only ensures the matching degree between the acquired upper label information and the basic label information to ensure the accuracy of the upper label information, but also ensures that the acquired upper label information can be adaptively modified based on the basic label information and the language environment, thereby ensuring the formation of the diversity of the upper label information, thereby improving the diversity of labels in the talent database.

[0125] In any of the above embodiments, the basic label information includes at least one of the following labels: a gender label, a nationality label, a health label, an education label, a professional title label, an industry label, an age label, and an achievement label.

[0126] In this embodiment, the basic label information in the talent information processing device is limited to one or more of the following labels: gender label, nationality label, health label, education label, professional title label, industry label, age label, and achievement label. This provides a basis for label content for determining the basic label corresponding to any talent data in the talent database, so as to ensure the accuracy and reliability of the talent data cleansing process in the talent database.

[0127] It should be noted that the basic label information of the present invention is not limited to the above-mentioned label information, but can be adaptively modified and supplemented according to the talent data information in the talent pool or the search information input by the user. For example, the information on the overseas work and study experience such as the number of years studied abroad, the number of years lived abroad, and the number of years worked abroad in the talent data information can be used to form a basic label "overseas background label".

[0128] In any of the above embodiments, any talent data includes one or more of reward information, patent achievement information and scientific research achievement information. Based on M basic label information including achievement labels, the M basic label information possessed by any talent data in the talent database is determined, including: using the textRank keyword extraction algorithm of the word graph model to extract keywords from the reward information, patent achievement information and scientific research achievement information to obtain achievement labels.

[0129] In this embodiment, any talent data is extracted from the talent database of the talent information processing device. When the talent data includes one or more of reward information, patent achievement information and scientific research achievement information, and the M basic label information corresponding to the talent data includes an achievement label, the textRank keyword extraction algorithm of the word graph model is used to extract keywords related to the reward information, patent achievement information and scientific research achievement information in the above talent data, and the extracted keywords are used as achievement labels. Thus, it is determined that for any talent data from the talent database, the above-mentioned method of generating achievement labels does not require learning multiple documents in advance, and is very suitable for extracting keywords such as reward information, patent information, scientific research achievement information, etc., so as to ensure that when determining the basic label for any talent data, the information related to the achievement involved in any talent data can be accurately converted into an achievement label, thereby ensuring that the talent database can form achievement labels more accurately, and then ensuring the accuracy of the basic label information in the talent database.

[0130] In any of the above embodiments, any talent data includes date of birth information, and based on M basic label information including an age label, determining the M basic label information possessed by any talent data in the talent database includes: obtaining the division relationship between the date of birth information and the age group; and determining the age group corresponding to the date of birth information based on the division relationship to obtain the age label.

[0131] In this embodiment, the talent database pre-stores relevant data on the division of age groups. When any talent data extracted from the talent database includes date of birth information, and the M basic label information corresponding to the any talent data includes an age label, the division is performed according to the correspondence between the date of birth information and the age group, and then the age group corresponding to the date of birth information is determined based on the above division result, and the age group is used as the age label of the date of birth information in the above any talent data, thereby determining the method for generating an age label for any talent data from the talent database, so as to ensure that when determining the basic label for any talent data, the age-related information involved in the above any talent data can be accurately converted into the age label, thereby ensuring that the talent database can form the age label more accurately and quickly, so as to ensure the accuracy of the basic label information in the talent database.

[0132] In any of the above embodiments, the talent information processing device also includes: an encoding unit 608, which is used to use the Bert model to encode M basic label information and N upper-level label information to obtain a vector distribution corresponding to any talent data; an association unit 606, which is also used to associate the vector distribution with any talent data; a matching unit 610, which is used to receive talent search information, and perform unsupervised text matching between the talent search information and the vector distribution corresponding to any talent data in the talent database to output the target talent data.

[0133] In this embodiment, the M basic label information and N upper label information determined above are encoded by the Bert model through the encoding unit 608, so as to obtain a vector distribution corresponding to any talent data. Through the association unit 606, the above vector distribution is associated with any talent data, so that the talent database can convert any talent data into a corresponding vector distribution, so that the talent database can realize vector-based search, avoiding the problems in the related technology that the recall rate of the search is low due to the ambiguity and diversity of the language expression when searching based on keywords, and the keywords have different semantics in different sentences, resulting in low search accuracy. After receiving the talent search information, the matching unit 610 performs unsupervised text matching on the talent search information and the vector distribution corresponding to any talent data in the talent database to adapt to the randomness of the search information input by the user, and output more accurate target talent data for the user, thereby improving the user's satisfaction with the searched information and the user's search recall rate.

[0134] In any of the above embodiments, the talent information processing device further includes: a vector conversion unit 612, which is used to perform Gaussian transformation on the vector distribution corresponding to any talent data in a standardized flow manner.

[0135] In this embodiment, through the vector conversion unit 612, after using the Bert model to encode M basic label information and N upper label information to obtain the vector distribution corresponding to any talent data, the vector distribution corresponding to the above-mentioned any talent data is converted into a smooth, isotropic standard Gaussian vector distribution by using a standardized flow method, and the above-mentioned Gaussian vector distribution is associated with the above-mentioned any talent data to improve the representation ability of the vector semantic similarity in the vector distribution, thereby improving the speed and accuracy of the search, and further enhancing the search effect.

[0136] A third aspect of the present invention is as follows Figure 7 As shown, a computer device 700 is proposed, including: a memory 702, the memory 702 stores a computer program; a controller 704, the controller 704 executes the computer program to implement the steps of the talent information processing method in any of the above embodiments.

[0137] Specifically, the memory 702 may include a large capacity memory 702 for data or instructions. By way of example and not limitation, the memory 702 may include a hard disk drive (HDD), a floppy disk drive, a flash memory, an optical disk, a magneto-optical disk, a magnetic tape, or a universal serial bus (USB) drive, or a combination of two or more of these. Where appropriate, the memory 702 may include a removable or non-removable (or fixed) medium. Where appropriate, the memory 702 may be inside or outside the integrated gateway disaster recovery device. In a specific embodiment, the memory 702 is a non-volatile solid-state memory. In a specific embodiment, the memory 702 includes a read-only memory 702 (ROM). Where appropriate, the ROM may be a mask-programmed ROM, a programmable ROM (PROM), an erasable PROM (EPROM), an electrically erasable PROM (EEPROM), an electrically rewritable ROM (EAROM), or a flash memory, or a combination of two or more of these. The above-mentioned processor may include a central processing unit (CPU), or an application specific integrated circuit (ASIC), or may be configured to implement one or more integrated circuits of an embodiment of the present invention.

[0138] The fourth aspect of the present invention provides a readable storage medium, which stores a program or instruction. When the program or instruction is executed by a processor, the steps of the talent information processing method in any of the above embodiments are implemented.

[0139] The readable storage medium provided by the present invention implements the steps of the talent information processing method of any of the above embodiments when the processor executes the computer program. Therefore, the readable storage medium includes all the beneficial effects of the talent information processing method of any of the above embodiments.

[0140] In the description of this specification, the terms "first" and "second" are used for descriptive purposes only and cannot be understood as indicating or implying relative importance unless otherwise clearly specified and limited; the terms "connection", "installation", "fixation" and the like should be understood in a broad sense, for example, "connection" can be a fixed connection, a detachable connection, or an integral connection; it can be directly connected or indirectly connected through an intermediate medium. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0141] In the description of this specification, the description of the terms "one embodiment", "some embodiments", "specific embodiments", etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representation of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any one or more embodiments or examples in a suitable manner.

[0142] The above are only preferred embodiments of the present invention and are not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A talent information processing method, characterized in that: include: Determine M basic label information of any talent data in the talent database; Determine upper-level label information for summarizing N basic label information, where N is less than or equal to M; Associating any talent data with the upper-level label information; The determining of the upper-level label information for summarizing the N basic label information specifically includes: Inputting the N basic label information into a pre-trained target classification model to obtain the upper-level label information; The basic tag information includes at least one of the following tags: Gender label, nationality label, health label, education label, job title label, industry label, age label, achievement label; Any talent data includes one or more of reward information, patent achievement information and scientific research achievement information. Based on the M basic label information including the achievement label, the determining of the M basic label information of any talent data in the talent database includes: The textRank keyword extraction algorithm of the word graph model is used to extract keywords from the reward information, the patent achievement information and the scientific research achievement information to obtain the achievement label.

2. The talent information processing method according to claim 1, characterized in that: Any talent data includes date of birth information, Based on the M basic label information including the age label, the determining of the M basic label information of any talent data in the talent database includes: Get the relationship between date of birth information and age groups; According to the division relationship, the age group corresponding to the birth date information is determined to obtain the age label.

3. The talent information processing method according to claim 1 or 2, characterized in that: Also includes: Using the Bert model, encoding the M basic label information and the N upper label information to obtain a vector distribution corresponding to any talent data; Associating the vector distribution with any talent data; as well as Receive talent search information, perform unsupervised text matching between the talent search information and a vector distribution corresponding to any talent data in the talent database, and output target talent data.

4. The talent information processing method according to claim 3, characterized in that: Also includes: The standardized flow method is used to perform Gaussian transformation on the vector distribution corresponding to any talent data.

5. A talent information processing device, using the talent information processing method according to any one of claims 1 to 4, characterized in that: include: A first determining unit, used to determine M basic label information of any talent data in the talent database; A second determining unit, configured to determine upper-level label information for summarizing N pieces of the basic label information, wherein N is less than or equal to M; An associating unit is used to associate any talent data with the upper-level label information.

6. A computer device, characterized in that: include: a memory storing a computer program; A controller, wherein the controller executes the computer program to implement the steps of the talent information processing method according to any one of claims 1 to 4.

7. A readable storage medium, characterized in that: The readable storage medium stores a program or instruction, and when the program or instruction is executed by the processor, the steps of the talent information processing method according to any one of claims 1 to 4 are implemented.

Images