Intelligent disambiguation method for talent information database

Abstract

The present invention relates to the field of data processing technology, and specifically to an intelligent disambiguation method for a talent information database, comprising: obtaining all data records of all data types of each talent entity information; obtaining the information mark degree of each data type according to the information association difference of different data records in the same data type; obtaining the necessity of information disambiguation of each talent entity information according to the information difference between each talent entity information and the overall talent entity information in the same data type and all data types, combined with the information mark degree; obtaining a number of clusters according to the necessity of information disambiguation; obtaining information similarity consistency according to the distribution trend of all data types of each talent entity information in the same cluster; and disambiguating according to the information similarity of each talent entity information. The present invention aims to solve the problem of missing and incorrect entity information, so as to improve the integrity and accuracy of the talent information database.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing, and in particular to an intelligent disambiguation method for a talent information database. Background Art

[0002] The talent information database stores a large amount of talent entity information. There is a possibility of missing information and entry errors in the process of entering and updating this information. Therefore, it is necessary to disambiguate the information in the talent information database, including identifying and correcting duplicate data, solving the problem of different expressions of the same entity, eliminating ambiguity, etc., to improve the data quality of the database and ensure that the information queried by talents is accurate.

[0003] Existing disambiguation of talent information databases mainly groups similar entity information together to help distinguish different individuals to achieve disambiguation. For example, talents with the same or similar educational background and work experience are grouped into the same group, and the educational background and work experience of the talents in the group are similar and closely related. However, due to the lack of some talent entity information in the talent information database, and the entity information of some talents is different from that of others, the completeness and accuracy of the data disambiguation of the entity information is affected, so that when performing disambiguation, erroneous and specific data causes errors in the disambiguation results, and the optimal disambiguation results cannot be achieved. Summary of the invention

[0004] The present invention provides an intelligent disambiguation method for a talent information database to solve the existing problem that due to the possibility of missing and erroneous entity information of some talents in the talent information database, the integrity and accuracy of data clustering of the entity information are affected, so that when performing cluster disambiguation, the erroneous and missing data cause errors in the clustering results.

[0005] The intelligent disambiguation method for talent information database of the present invention adopts the following technical scheme:

[0006] An embodiment of the present invention provides an intelligent disambiguation method for a talent information database, the method comprising the following steps:

[0007] Get all data records of all data types of each talent entity information;

[0008] According to the information association differences of different data records in the same data type, the information marking degree of each data type is obtained;

[0009] Based on the information difference between each talent entity information and the overall talent entity information in the same data type, as well as the information difference between each talent entity information and the overall talent entity information in all data types, combined with the degree of information marking, the necessity of information disambiguation for each talent entity information is obtained;

[0010] All talent entity information is classified according to the necessity of information disambiguation to obtain several clusters; based on the consistency of distribution trends of all data types of each talent entity information in the same cluster, the information similarity and consistency of each talent entity information is obtained;

[0011] Data records in the talent information database are disambiguated based on the information similarity of each talent entity information.

[0012] Preferably, obtaining the information mark level of each data type includes:

[0013] The specificity of all data types is inversely normalized using an exponential function with a natural constant as the base, and the normalized result of the specificity of each data type is recorded as the information marker degree of each data type.

[0014] Preferably, the specificity of each data type includes:

[0015] Count all data records in each data type. Among all data records in each data type, calculate the cosine similarity of the text vector of each data record with each other data record and the absolute value of the text vector length difference. Calculate the cumulative sum of the ratio of the cosine similarity to the absolute value of each data record in each data type, and record it as the specificity of each data type.

[0016] Preferably, the necessity of obtaining information disambiguation of each talent entity information includes:

[0017] The ratio of the sum of the information difference coefficients of each data type of each talent entity information to the information difference coefficients of the overall data type of each talent entity information is linearly normalized and recorded as the necessity of information disambiguation for each talent entity information.

[0018] Preferably, the information difference coefficient of each data type includes:

[0019] Calculate the mean vector of the text vectors of the data records of each data type of all talent entity information, take the information mark degree of each data type as the weight, calculate the product of the weight and the vector modulus of the difference between the text vector of the data record of each data type of each talent entity information and the mean vector of the data records of the corresponding data type of all talent entity information, and record it as the information difference coefficient of each data type of each talent entity information.

[0020] Preferably, the information difference coefficient of the overall data type of each talent entity information includes:

[0021] The vector modulus of the difference between the sequence consisting of text vectors of all data types of each talent entity information and the sequence consisting of the vector mean of the text vectors of all data types of each talent entity information in the talent information database is recorded as the information difference coefficient of the overall data type of each talent entity information.

[0022] Preferably, the step of classifying all talent entity information according to the necessity of information disambiguation to obtain a number of clusters includes:

[0023] A two-dimensional clustering space is constructed. The horizontal axis of the clustering space is the serial number of the talent entity information, and the vertical axis is the value of the necessity of information disambiguation. The necessity of information disambiguation of the talent entity information is used as the clustering metric of the DBSCAN clustering algorithm. Clustering is performed according to the preset number of samples and the preset sample radius to obtain several clusters.

[0024] Preferably, obtaining the information similarity and consistency of each talent entity information includes:

[0025] In the same cluster, the average of the absolute values ​​of the information disambiguation necessity differences between each talent entity information and all other talent entity information is recorded as the information similarity measure of each talent entity information. The information similarity measure of each talent entity information in all clusters is normalized, and the normalized result of the information similarity measure of each talent entity information is recorded as the information similarity consistency of each talent entity information.

[0026] Preferably, performing data record disambiguation in the talent information database according to the information similarity of each talent entity information includes:

[0027] All high-trust talent entity information is packaged to obtain packaged information, and the packaged information is integrated and verified through the talent information database to achieve disambiguation of the talent information database; for each talent entity information to be disambiguated, the data record of the data type of each talent entity information to be disambiguated is compared and verified with the source data collected from the talent information database, and the data records that fail the verification are disambiguated through difference processing or average filling of the corresponding data type.

[0028] Preferably, the specific method of obtaining the high-trust talent entity information and the talent entity information to be disambiguated includes:

[0029] A disambiguation threshold is preset; talent entity information whose information similarity consistency is greater than the disambiguation threshold is recorded as high-trust talent entity information, and talent entity information whose information similarity consistency is greater than the disambiguation threshold is recorded as talent entity information to be disambiguated.

[0030] The technical solution of the present invention has the following beneficial effects: the present invention obtains all data records of all data types of each talent entity information; obtains the information mark degree of each data type according to the information association difference of different data records in the same data type, reflects the influence of the information difference of different data types on the overall similarity, thereby ensuring that the overall feature loss of the data after disambiguation is minimal compared with the original data, and realizes the refined disambiguation of the talent entity information; obtains the necessity of information disambiguation of each talent entity information according to the information difference between each talent entity information and the overall talent entity information in the same data type, and the information difference between each talent entity information and the overall talent entity information in all data types, combined with the information mark degree, reflects the significance of the information features of each data type through the information mark degree, and obtains the disambiguation requirement of each talent entity information according to the difference between the information meaning of each talent entity information and the overall information meaning; and then obtains the information similarity consistency of each talent entity information according to the distribution trend consistency of all data types of each talent entity information in the same cluster, realizes the retention of normal information that is specific to other talent entity information, and avoids the normal information being misjudged as information that needs to be disambiguated; and performs data record disambiguation of the talent information database according to the information similarity of each talent entity information, thereby improving the integrity and accuracy of the talent information database. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0032] Figure 1 The figure is a flowchart of the steps of the intelligent disambiguation method for a talent information database according to the present invention. DETAILED DESCRIPTION

[0033] In order to further explain the technical means and effects adopted by the present invention to achieve the predetermined invention purpose, the specific implementation method, structure, features and effects of the intelligent disambiguation method for talent information database proposed by the present invention are described in detail below in combination with the accompanying drawings and preferred embodiments. In the following description, different "one embodiment" or "another embodiment" does not necessarily refer to the same embodiment. In addition, specific features, structures or characteristics in one or more embodiments may be combined in any suitable form.

[0034] Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.

[0035] The specific scheme of the intelligent disambiguation method for a talent information database provided by the present invention is described in detail below with reference to the accompanying drawings.

[0036] See also Figure 1 , which shows a flowchart of a method for intelligent disambiguation of a talent information database provided by an embodiment of the present invention, the method comprising the following steps:

[0037] Step S001: Obtain all data records of all data types of each talent entity information.

[0038] The purpose of this embodiment is to perform difference analysis on the data records in the talent information database, and then perform disambiguation processing on the data records with different disambiguation requirements, so as to ensure the integrity and accuracy of the data records in the talent information database. Therefore, it is first necessary to collect all data records of all data types of each talent entity information in the talent information database.

[0039] Specifically, this embodiment collects all entity information of each person in the talent information database, and records each entity information category collected as a data type. The entity information categories include each person's name, contact information, intellectual property achievements, paper projects, award information and other categories; the collected entity information mainly exists in the form of text and characters. Since text and characters cannot directly reflect the characteristics between entity information and entity information, this embodiment will perform word segmentation on all the collected talent entity information, and use the word segmentation results as a data set to construct a text information vectorization model, and obtain the text vector of each data type of each talent entity information as a data record of each data type of each talent entity information.

[0040] Preferably, in one embodiment of the present invention, all collected talent entity information is segmented, and the segmentation results are used as a data set to construct a text information vectorization model, and the text vector of each data type of each talent entity information is obtained and recorded as a data record of each data type of each talent entity information, including the specific method of:

[0041] Use the Jieba word segmentation algorithm to segment the entity information of all people to obtain several word segments, use all the word segments as corpus to train the word2vec model, obtain the text vector combination of all entity information of each data type of each person, and record the text vector of each data type of each person as the data record of each data type of each talent entity information. It should be noted that the Jieba word segmentation algorithm described in this embodiment and the use of the word2vec model to train the corpus to obtain the text vector are existing well-known technologies, and will not be repeated in this embodiment.

[0042] Step S002: Obtain the information marking degree of each data type according to the information association difference of different data records in the same data type.

[0043] It should be noted that different data types have different disambiguation requirements because the information they represent has different information reference significance to the talent information database. As an example, the information characteristics of a name as an entity information have weak similarity and correlation, and it is impossible to obtain similar characteristics of the talent entity information by analyzing all the names in the talent information database, that is, there is no reference significance of the information. Therefore, the disambiguation requirement for the entity information of the name is extremely small, and disambiguation can be completed only by interpolation or re-verification. Therefore, this embodiment obtains the information marking degree of each data type by the information differences and correlation differences represented by all data records of each data type in the talent information database.

[0044] It should be further explained that information differences of different data types have different impacts on overall similarity. Therefore, it is necessary to achieve refined disambiguation of talent entity information while ensuring that the overall feature loss of the data after disambiguation is minimal compared to the original data. For data records of the same data type, the greater the degree of repetition and similarity in meaning between the data records, the stronger the feature strength of the data type. That is, when disambiguating, the greater the necessity of disambiguation that needs to be set, eliminating data that differs from the overall data characteristics to ensure the integrity and accuracy of the data in the talent database. At the same time, the more missing data there is in each data record, the less significance this data type has for the overall analysis of talent entity information. Therefore, the greater the necessity of disambiguation that needs to be set.

[0045] Preferably, in one embodiment of the present invention, the information mark degree of each data type is obtained according to the information association difference of different data records in the same data type, and the specific method includes:

[0046] Count all data records in each data type, and in all data records of each data type, calculate the cosine similarity of the text vector of each data record and each other data record and the absolute value of the text vector length difference, calculate the cumulative sum of the ratio of the cosine similarity to the absolute value of each data record in each data type and record it as the specificity of each data type, and use an exponential function with a natural constant as the base to inversely normalize the specificity of all data types, and record the normalized result of the specificity of each data type as the information mark degree of each data type. As an example, the information mark degree of the i-th data type is calculated as follows:

[0047]

[0048] In the formula, ρ iis the specificity of the i-th data type, R i is the number of data records of the i-th data type, a i,r is the text vector of the rth data record in the i-th data type, a i,l is the text vector of the lth data record except the rth data record in the i-th data type; n i,r is the vector modulus of the text vector of the rth data record in the i-th data type, n i,l is the vector modulus of the text vector of the lth data record excluding the rth data record in the ith data type; || represents the absolute value function, coss() represents the calculation of the cosine similarity function, ε is a preset hyperparameter used to avoid the denominator being 0. In this embodiment, the preset hyperparameter is described as ε=0.01, and other embodiments may set other values, which are not specifically limited in this embodiment.

[0049] It should be noted that coss(a i,r ,a i,l ) represents the cosine similarity between the text vector of the rth data record in the ith data type and the text vector of the lth data record excluding the rth data record. The larger the value, the higher the similarity of the two text vectors, and the more similar the meanings are, that is, the higher the correlation of the features expressed, the more likely the ith data type is to be an important data type with a high degree of disambiguation demand, so the greater the value of the specificity of the ith data type; |n i,r -n i,l | represents the difference in text vector module length between the text vectors of the rth data record in the ith data type and the lth data record excluding the rth data record. The smaller the value, the more similar the two text vectors are, and the greater the specificity of the ith data type.

[0050] Similarly, the specificity of all data types is obtained, and the specificity of all data types is inversely normalized using an exponential function with a natural constant as the base. The normalized result of the specificity of each data type is recorded as the information mark degree of each data type.

[0051] Step S003: Obtain the necessity of information disambiguation for each talent entity information based on the information difference between each talent entity information and the overall talent entity information in the same data type, and the information difference between each talent entity information and the overall talent entity information in all data types, combined with the information identification degree.

[0052] It should be noted that the disambiguation of the talent information database is performed on the data records of all data types of each talent entity information, and the data records of each data type of different talent entity information exist in two states: normal and need to be disambiguated. Therefore, it is necessary to analyze the data records of each data type of each talent entity information and the data records of the data type corresponding to all talent entity information, so as to evaluate the necessity of information disambiguation for each talent entity information based on the difference between the data records of each talent entity information and all talent entity information.

[0053] It should be further explained that the sequence formed by the data records of all data types of each talent entity information represents the overall information characteristics of the talent entity information. Therefore, for all talent entity information, the greater the difference between the data records of the data type with the greater information mark degree in the sequence of all talent entity information, the greater the difference. This means that even if some data types of the talent entity information are similar to the data types of all talent entity information, the talent entity information also needs to be disambiguated because of the existence of data types with great disambiguation requirements. Therefore, this embodiment uses the information mark degree as the reference weight of each data type based on the information difference between each talent entity information and the overall talent entity information in the same data type, combined with the information difference between each talent entity information and the overall talent entity information in all data types, so as to obtain the necessity of information disambiguation for each talent entity information.

[0054] Preferably, in an embodiment of the present invention, according to the information difference between each talent entity information and the overall talent entity information in the same data type, the information mark degree is used as the reference weight of each data type, and the information difference between each talent entity information and the overall talent entity information in all data types is combined to obtain the necessity of information disambiguation for each talent entity information, including the specific method of:

[0055] Calculate the mean vector of the text vectors of the data records of each data type of all talent entity information, take the information mark degree of each data type as the weight, calculate the product of the weight and the vector modulus of the difference between the text vector of each data type of each talent entity information and the mean vector of the data records of the corresponding data type of all talent entity information, and record it as the information difference coefficient of each data type of each talent entity information; record the vector modulus of the difference between the sequence composed of the text vectors of all data types of each talent entity information and the vector mean of the text vectors of all data types of each talent entity information in the talent information database as the information difference coefficient of the overall data type of each talent entity information; linearly normalize the cumulative sum of the information difference coefficients of each data type of each talent entity information and the ratio of the information difference coefficients of the overall data type of each talent entity information, and record it as the necessity of information disambiguation of each talent entity information. As an example, record the information mark degree of the i-th data type as WB i , the calculation method of the necessity of information disambiguation of the j-th talent entity information is:

[0056]

[0057] In the formula, XB j is the information disambiguation requirement of the j-th talent entity information, I j is the number of data types of the j-th talent entity information, a i,j is the text vector of the data record of the i-th data type of the j-th talent entity information, A is the mean vector of data records of the ith data type of all talent entity information; j A sequence of text vectors of all data types for each talent entity information, It is a sequence consisting of the vector means of text vectors of all data types of each talent entity information in the talent information database; ‖‖ represents the vector modulus symbol, ε is a preset hyperparameter used to avoid the denominator being 0. In this embodiment, the preset hyperparameter is described as ε=0.01, and other embodiments may set other values, which are not specifically limited in this embodiment.

[0058] It should be noted that The cumulative sum of the information difference coefficients of all data types representing the j-th talent entity information. The information mark degree of each data type is used as the weight. The larger the value, the greater the difference between the j-th talent entity information and the overall information in the talent database. And they all appear in the data type with a high information mark degree. Therefore, the greater the value of the information disambiguation demand degree of the j-th talent entity information; The information difference coefficient representing the overall data type of the j-th talent entity information. The larger the value, the greater the difference between the overall information represented by all data types of the j-th talent entity information and the overall information of all talents in the talent database, and the greater the degree of information disambiguation requirement for the j-th talent entity information.

[0059] Similarly, the information disambiguation requirement of all talent entity information is obtained, and the linear normalization model is used to normalize the information disambiguation requirement of all talent entity information. The normalized result of the information disambiguation requirement of the j-th talent entity information is recorded as the necessity of information disambiguation of the j-th talent entity information.

[0060] Step S004: classify all talent entity information according to the necessity of information disambiguation to obtain a number of clusters; obtain the information similarity consistency of each talent entity information according to the consistency of distribution trends of all data types of each talent entity information in the same cluster.

[0061] It should be noted that after obtaining the necessity of information disambiguation of all talent entity information, the necessity of information disambiguation represents the information characteristics of each talent entity information. However, due to the existence of different data types, the information of some talent entities is specific to other talent entities. Therefore, the talent entities specific to other talent entities present a clustered distribution of small clusters in the necessity of information disambiguation of all talent entity information, and their values ​​of information disambiguation necessity are larger. Talent entities specific to other talent entities do not need to be disambiguated. Therefore, this embodiment clusters the necessity of information disambiguation of all talent entity information to obtain several clustering clusters, each of which contains several human entity information with similar necessity of information disambiguation.

[0062] It should be further explained that the trend distribution of the values ​​of the necessity of information disambiguation of talent entity information in each cluster cluster represents the information characteristics of the talent entity information in the cluster cluster. The more uniform the distribution of the necessity of information disambiguation of talent entity information in the cluster cluster, the more normal the difference in information characteristics, that is, the same type of talent entity information, which also indicates that the cluster only contains information that is unique to other talent entities or only contains normal talent entity information. When the distribution of the necessity of information disambiguation in the cluster cluster is uneven, it indicates that there is abnormal talent entity information that needs to be disambiguated in the cluster cluster, and the distribution of abnormal talent entity information and other normal talent entity information in the cluster cluster has no obvious regularity due to different characteristics. Therefore, this embodiment obtains the information similarity and consistency of each talent entity information according to the difference in the location of each talent entity information and other talent entity information in the same cluster cluster.

[0063] Preferably, in one embodiment of the present invention, the information similarity consistency of each talent entity information is obtained according to the location difference between each talent entity information and other talent entity information in the same cluster, including the specific method of:

[0064] The necessity of information disambiguation of talent entity information is used as a similarity measure to construct a two-dimensional clustering space, the horizontal axis of the clustering space is the sequence number of the talent entity information, and the vertical axis is the value of the necessity of information disambiguation. The necessity of information disambiguation of talent entity information is used as the clustering measure of the DBSCAN clustering algorithm, and clustering is performed according to the preset number of samples and the preset sample radius to obtain a number of clusters, each of which contains a number of talent entity information. It should be noted that this embodiment is described by taking the preset number of samples as 5 and the preset sample radius as 0.05 as an example, and other embodiments may set other values, which will not be described in detail in this embodiment.

[0065] Furthermore, in the same cluster, the average of the absolute values ​​of the information disambiguation necessity differences between each talent entity information and all other talent entity information is recorded as the information similarity measure of each talent entity information, and the information similarity measure of each talent entity information in all clusters is normalized, and the normalized result of the information similarity measure of each talent entity information is recorded as the information similarity consistency of each talent entity information. As an example, the steps for obtaining the information similarity consistency of the mth talent entity information in the kth cluster are:

[0066] The necessity of information disambiguation for the mth talent entity in the kth cluster is denoted as XB k,m , the necessity of information disambiguation of the nth talent entity information in the kth cluster is recorded as XB k,n , then the information similarity measure of the m-th talent entity information in the k-th cluster is calculated as:

[0067]

[0068] Among them, QL k,m K is the information similarity measure of the mth talent entity information in the kth cluster. k is the number of talent entity information in the kth cluster, and || is the absolute value function.

[0069] Furthermore, in all clusters, the information similarity measure of all talent entity information is calculated, and the sigmoid function is used as a normalization model to normalize the information similarity measure of all talent entity information to [0,1], and the normalized result of the information similarity measure of the m-th talent entity information is recorded as the information similarity consistency of the m-th talent entity information.

[0070] Step S005: disambiguate data records in the talent information database according to the information similarity of each talent entity information.

[0071] Preferably, in one embodiment of the present invention, data records in the talent information database are disambiguated according to the information similarity of each talent entity information, including the specific method of:

[0072] A disambiguation threshold is preset. The disambiguation threshold of this embodiment is described by taking 0.6 as an example. Other embodiments may set other values, which are not specifically limited in this embodiment. Talent entity information whose information similarity consistency is greater than the disambiguation threshold is recorded as high-trust talent entity information, and talent entity information whose information similarity consistency is greater than the disambiguation threshold is recorded as talent entity information to be disambiguated. All high-trust talent entity information is packaged to obtain packaged information, and the packaged information is integrated and verified through the talent information database to achieve disambiguation of the talent information database. For each talent entity information to be disambiguated, the data record of the data type of each talent entity information to be disambiguated is compared and verified with the source data collected by the talent information database, and the data records that fail the verification are disambiguated by difference processing or average filling of the corresponding data type.

[0073] This is the end of the present invention.

[0074] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the principles of the present invention should be included in the protection scope of the present invention.

Claims

1. An intelligent disambiguation method for a talent information database, characterized in that: The method comprises the following steps: Get all data records of all data types of each talent entity information; According to the information association differences of different data records in the same data type, the information marking degree of each data type is obtained; Based on the information difference between each talent entity information and the overall talent entity information in the same data type, as well as the information difference between each talent entity information and the overall talent entity information in all data types, combined with the degree of information marking, the necessity of information disambiguation for each talent entity information is obtained; All talent entity information is classified according to the necessity of information disambiguation to obtain several clusters; based on the consistency of distribution trends of all data types of each talent entity information in the same cluster, the information similarity and consistency of each talent entity information is obtained; Disambiguate data records in the talent information database based on the information similarity of each talent entity information; The necessity of obtaining information disambiguation for each talent entity information includes: The accumulated sum of the information difference coefficients of each data type of each talent entity information and the ratio of the information difference coefficients of the overall data type of each talent entity information are linearly normalized and recorded as the necessity of information disambiguation of each talent entity information; The information difference coefficient of each data type includes: Calculate the mean vector of the text vectors of the data records of each data type of all talent entity information, take the information mark degree of each data type as the weight, calculate the product of the weight and the vector modulus of the difference between the text vector of the data record of each data type of each talent entity information and the mean vector of the data records of the corresponding data type of all talent entity information, and record it as the information difference coefficient of each data type of each talent entity information; The information difference coefficient of the overall data type of each talent entity information includes: The vector modulus of the difference between the sequence consisting of text vectors of all data types of each talent entity information and the sequence consisting of the vector mean of the text vectors of all data types of each talent entity information in the talent information database is recorded as the information difference coefficient of the overall data type of each talent entity information.

2. The intelligent disambiguation method for a talent information database according to claim 1, characterized in that: The information mark level of each data type is obtained as follows: The specificity of all data types is inversely normalized using an exponential function with a natural constant as the base, and the normalized result of the specificity of each data type is recorded as the information marker degree of each data type.

3. The intelligent disambiguation method for talent information database according to claim 1, characterized in that: The specificity of each data type includes: Count all data records in each data type. Among all data records in each data type, calculate the cosine similarity of the text vector of each data record with each other data record and the absolute value of the text vector length difference. Calculate the cumulative sum of the ratio of the cosine similarity to the absolute value of each data record in each data type, and record it as the specificity of each data type.

4. The intelligent disambiguation method for a talent information database according to claim 1, characterized in that: The information of all talent entities is classified according to the necessity of information disambiguation to obtain several clusters including: A two-dimensional clustering space is constructed. The horizontal axis of the clustering space is the serial number of the talent entity information, and the vertical axis is the value of the necessity of information disambiguation. The necessity of information disambiguation of the talent entity information is used as the clustering metric of the DBSCAN clustering algorithm. Clustering is performed according to the preset number of samples and the preset sample radius to obtain several clusters.

5. The intelligent disambiguation method for talent information database according to claim 1, characterized in that: The information similarity and consistency of each talent entity information obtained includes: In the same cluster, the average of the absolute values ​​of the information disambiguation necessity differences between each talent entity information and all other talent entity information is recorded as the information similarity measure of each talent entity information. The information similarity measure of each talent entity information in all clusters is normalized, and the normalized result of the information similarity measure of each talent entity information is recorded as the information similarity consistency of each talent entity information.

6. The intelligent disambiguation method for a talent information database according to claim 1, characterized in that: The disambiguating data records in the talent information database according to the information similarity of each talent entity information includes: All high-trust talent entity information is packaged to obtain packaged information, and the packaged information is integrated and verified through the talent information database to achieve disambiguation of the talent information database; for each talent entity information to be disambiguated, the data record of the data type of each talent entity information to be disambiguated is compared and verified with the source data collected from the talent information database, and the data records that fail the verification are disambiguated through difference processing or average filling of the corresponding data type.

7. The intelligent disambiguation method for a talent information database according to claim 6, characterized in that: The specific methods for obtaining the high-trust talent entity information and the talent entity information to be disambiguated include: A disambiguation threshold is preset; talent entity information whose information similarity consistency is greater than the disambiguation threshold is recorded as high-trust talent entity information, and talent entity information whose information similarity consistency is greater than the disambiguation threshold is recorded as talent entity information to be disambiguated.

Images