Intelligent matching method based on talent information

Abstract

The invention provides an intelligent matching method based on talent information, and belongs to the field of intelligent matching. The problem that talent information is difficult to search is solved; the method specifically comprises the following steps: S1, obtaining talent information and employment demand information of an enterprise; s2, acquiring a human demand standard, and dividing the human demand standard into a rigid requirement and an elastic requirement; necessary items are obtained according to the hardness requirement; according to the elasticity requirement, obtaining score adding items; s3, screening the talent information through necessary items to obtain optimized talent information, and performing similarity matching on the optimized talent information and an elastic requirement to obtain a similarity value; s4, setting a weight for the elasticity requirement; constructing a matching scoring model according to the similarity values and the weights, and scoring the talents; and S5, sorting the scoring results, and performing talent recommendation on the enterprise in combination with the employment demand information of the enterprise. According to the invention, through intelligent matching of talent information, data reference is provided for enterprises, the cost of recruiting people is reduced, and the efficiency of recruiting people is improved.

Description

Technical Field

[0001] The present invention discloses an intelligent matching method based on talent information, which relates to the field of intelligent matching. Background Art

[0002] The existing methods for intelligent matching of talent information have the following shortcomings:

[0003] Lack of personalized user needs: Different companies (such as startups and large enterprises) have significantly different talent requirements, but the system often matches talent based on existing matching rules, making it difficult to meet the needs of different companies.

[0004] Insufficient matching accuracy: Existing intelligent talent matching methods match talent based on a single dimension (such as work experience or skills), resulting in one-sided matching results. Most matching methods rely on semantic text matching and directly output matching results, resulting in poor matching accuracy.

[0005] High interaction complexity: Intelligent talent information matching methods capable of multi-dimensional matching usually require users to independently set weight parameters for different dimensions. They have high operational thresholds, are complex to use, and are difficult to use widely. Summary of the Invention

[0006] In view of the shortcomings of the existing technology, the purpose of the present invention is to provide a method based on intelligent matching of talent information, aiming to solve the problem of difficulty in finding talent information.

[0007] In order to achieve the above-mentioned object, the present invention is implemented through the following technical solution: a method for intelligent matching based on talent information, the matching method comprising:

[0008] Step S1: Obtain talent information and the company's employment demand information;

[0009] Step S2: Process the enterprise's employment demand information to obtain employment demand standards, which are divided into hard requirements and flexible requirements; obtain necessary items based on hard requirements; and obtain bonus items based on flexible requirements;

[0010] Step S3: Process the talent information, filter the talent information by necessary items, and obtain optimized talent information; perform similarity matching between the optimized talent information and the bonus items to obtain similarity values; sort the similarity values, and assign initial scores based on the sorting results;

[0011] Step S4: weights are set based on the numerical distribution of similarity values; a matching scoring model is constructed based on the similarity values, initial scores, and weights to score the talent;

[0012] Step S5: Sort the scoring results and recommend talents to the enterprise based on the enterprise's employment demand information.

[0013] Furthermore, the specific steps of step S2 are as follows:

[0014] Step S21: Acquire the enterprise's employment demand information; perform structural field conversion on the enterprise's employment demand information to obtain employment demand standards, which are represented by standard dimensions and standard descriptions;

[0015] Step S22: Extracting the hard requirements from the hiring demand standards, obtaining the standard dimensions and standard descriptions corresponding to the hard requirements, integrating them, recording the integrated standard dimensions as necessary dimensions, and recording the integrated standard descriptions as necessary descriptions, to obtain necessary items; necessary items: {necessary dimensions: necessary descriptions};

[0016] Extract the flexibility requirements in the employment demand standards, obtain the standard dimensions and standard descriptions corresponding to the flexibility requirements, integrate them, record the integrated standard dimensions as bonus dimensions, and record the integrated standard descriptions as bonus descriptions to obtain bonus items; bonus items: {bonus dimensions: bonus descriptions}.

[0017] Furthermore, the specific steps of step S21 are as follows:

[0018] Step S211: Preprocess the text of the enterprise's employment demand information, extract the text of the employment demand information, perform cyclic judgment on the text, and delete the text content if it contains irrelevant characters, formatting marks, or stop words to obtain a clean text;

[0019] Step S212: Perform word segmentation on the cleaned text, identify key entities in the text, classify the key entities, obtain classification dimensions, and record the key entities under the same classification dimension as dimension content; based on the classification dimension and dimension content, obtain the employment demand standard: {standard dimension: standard description}.

[0020] Furthermore, the specific steps of step S211 are as follows:

[0021] Remove emoticons: obtain the Unicode code bq1 of the first emoticon in the computer, obtain the Unicode code bq2 of the last emoticon in the computer, and obtain the emoticon code range BQJ according to the emoticon code range [bq1, bq2]; convert the employment demand information into character codes, count the number of character codes bm, and record the character codes as ZF(1), ZF(2), ..., ZF(bm). If the character codes ZF(1) to ZF(bm)∈BQJ, replace the character codes with null;

[0022] Remove formatting tags: Get the type of formatting tag z and record the formatting tags as gs1, gs2, ..., gsz; build a formatting list, use the formatting tags gs1 to gsz as list elements, fill the formatting list, match the employment demand information with gs1 to gsz according to the formatting list, delete the corresponding content of gs1 to gsz in the employment demand information, and complete the cleaning of the formatting list;

[0023] The details are as follows:

[0024] The format mark is binary-encoded to obtain the length r of the binary code; the code position is recorded as x, and the value at the code position is recorded as y, a plane rectangular coordinate system is constructed, and the binary code of the format mark is mapped in the plane rectangular coordinate system; the format code point set {(x1, y1), (x2, y2), ..., (xr, yr)} is obtained. According to the point set {(x1, y1) to (xr, yr)}, the points with y values ​​of 0 and 1 are connected respectively, and the length of the connection is calculated:

[0025]

[0026] Where: cd0 represents the connection length when the y value is 0; x w and y w Indicates the w-th x value and y value;

[0027]

[0028] Where: cd1 represents the connection length when the y value is 1;

[0029] According to cd0 and cd1, the employment demand information is matched and deleted to complete the cleaning of the format list;

[0030] Remove stop words: Get the stop word type t and record the stop words as TY1, TY2, ..., TYt; use stop words TY1 to TYt as elements to construct a stop word list, express the stop word list using regular expressions, search for stop words in the job demand information, and delete the searched stop words.

[0031] Furthermore, the specific steps of step S3 are as follows:

[0032] Step S31: Obtain talent information, obtain necessary dimensions and necessary descriptions based on necessary items, match and locate the talent information based on the necessary dimensions, match the necessary descriptions with the talent information, extract talent information that meets the necessary items, and delete talent information that does not meet the necessary items, thereby obtaining optimized talent information;

[0033] Step S32: According to the bonus items, obtain the bonus dimensions and bonus descriptions; match the bonus dimensions with the optimized talent information, calculate the similarity in combination with the bonus descriptions, obtain similarity values, sort the similar values ​​of the same dimension according to the bonus dimensions, and perform initial scoring based on the sorting results.

[0034] Furthermore, the specific steps of step S31 are as follows:

[0035] Step S311: Obtain the number a of necessary dimensions, record the necessary dimensions as bwd1, bwd2, ..., bwda; record the necessary descriptions corresponding to the necessary dimension items as bms1, bms2, ..., bmsa; match and locate talent information in sequence according to the necessary dimensions; and match the necessary descriptions with the talent information based on the matching results;

[0036] Using the necessary dimensions as the judgment criteria, the talent information is traversed and the necessary dimensions are matched with the talent information through character comparison. If the talent information does not contain a description of the necessary dimensions, the talent information is deleted.

[0037] If there is a description of the necessary dimension in the talent information, the corresponding content is extracted and matched with the necessary description, and the talent information is deleted based on the matching result;

[0038] Match the necessary dimensions bwd1 to bwda with the necessary descriptions bms1 to bmsa;

[0039] Step S312: If corresponding matching content can be obtained for the necessary dimensions bwd1 to bwda and the necessary descriptions bms1 to bmsa in the talent information, the talent information is retained; and the retained talent information is integrated to obtain optimized talent information.

[0040] Furthermore, the specific steps of step S312 are as follows:

[0041] Build an optimized talent information list, obtain the necessary dimensions bwd1, bwd2, ..., bwda and the necessary descriptions bms1, bms2, ..., bmsa, and obtain corresponding matching talent information in the talent information. Delete the necessary dimensions and necessary descriptions in the talent information.

[0042] The talent information is simplified by performing an XOR operation on the necessary dimensions bwd1 to bwda and the necessary descriptions bms1 to bmsa with the matching talent information, and the simplified talent information is used as a list element to fill the optimized talent information list;

[0043] Sort the elements in the optimized talent information list; obtain the data storage value of the simplified talent information, and sort the talent information according to the size of the data storage value;

[0044] Obtain the number of elements ysl in the optimized talent information list, and record the elements in the optimized talent information list as ylb[ys]; use the first talent information in the optimized talent information list as the comparison value bj; and perform a cyclic comparison on the optimized talent information list based on the comparison value;

[0045] If bj>ylb[ys], replace ylb[ys] with bj and compare ylb[ys+1] until ys=ysl;

[0046] If bj≤ylb[ys], then compare ylb[ys+1] until ys=ysl;

[0047] When the loop is completed, the optimized talent information is sorted in descending order of the talent information data storage value; a sorted optimized talent information list is obtained, and the optimized talent information is called according to the sorted optimized talent information list.

[0048] Furthermore, the specific steps of step S32 are as follows:

[0049] Step S321: Obtain the number b of bonus dimensions, record the bonus dimensions as jwd1, jwd2, ..., jwdb; record the bonus descriptions corresponding to the bonus dimensions as jms1, jms2, ..., jmsb; use the bonus dimensions as search targets to search for optimized talent information; obtain search content, match the search content with the bonus descriptions, and obtain similarity values;

[0050] Step S322: Obtain the similarity values ​​of the b bonus dimensions to obtain a similarity value list xsl, where xsl = [xsz1, xsz2, ..., xszb]; obtain the number f of optimized talent information; sort the similarity values ​​of the b bonus dimensions in descending order based on the number f of optimized talent information and the similarity value list xsl; count the sorting positions px to obtain a sorted list bxl, where bxl = [px1, px2, ..., pxb];

[0051] Step S323: Initially assign points to the optimized talent information based on the sorted list; map the sorted positions to the interval [1, 2] for scoring. The specific scoring is as follows:

[0052] Obtain the ranking px, the number of optimized talent information f, and the mapping interval [1,2]. Perform a mapping calculation on the ranking px according to the mapping interval [1,2] to obtain the initial score cff.

[0053]

[0054] Furthermore, the specific steps of step S321 are as follows:

[0055] Obtain search content, perform word segmentation on the search content, and obtain a search word list; obtain the number of search words c, and record the search words as js1, js2, ..., jsc; perform word segmentation on the bonus description to obtain a description word list; obtain the number of description words d, and record the description words as ms1, ms2, ..., msd;

[0056] The search word js1 is matched with the description words ms1 to msd respectively to obtain the meaning matching value, and the maximum meaning matching value is obtained as the matching value xfz of the search word; the matching value of the search word is judged,

[0057] If xfz>0.5, the search term and the description term are judged to be matched, and the matching value of the search term is saved;

[0058] If xfz≤0.5, the search term and the description term are judged to be unmatched and the search value is deleted;

[0059] Similarly, match js2 to jsc with the description words ms1 to msd to obtain the matching values ​​of the search words; count the matching values ​​that can match the search words and description words to obtain a matching value list;

[0060] Get the number of matching values ​​e in the matching value list; add up the matching values ​​to get the similarity value xsz:

[0061]

[0062] Among them: xfz i Indicates the i-th matching value.

[0063] Furthermore, the specific steps of step S4 are as follows:

[0064] Step S41: Obtain the similarity values ​​of b bonus dimensions, calculate the variance of the similarity values ​​of the b bonus dimensions, and set weights for the score calculation of the bonus dimensions based on the variance of the similarity values, as follows:

[0065] Obtain the number f of optimized talent information, and based on the number f of optimized talent information, record the similarity values ​​of the b bonus dimensions as A(U,V);

[0066] Calculate the mean of similarity values ​​under the same dimension to obtain the mean of similarity values ​​xjz;

[0067]

[0068] Where: xjz(U) represents the mean similarity value under the U-th dimension, A(U,V) represents the similarity value of the U-th bonus dimension and the V-th optimized talent information;

[0069] According to the similarity mean xjz(U) and the similarity value A(U,V), the variance fc is calculated:

[0070]

[0071] Where: fc(U) represents the variance in the U-th dimension;

[0072] Set the variance as the calculation weight qz of the bonus dimension, that is, fc(U) = qz(U);

[0073] Step S42: Obtain the initial score cff(U), and build a matching score model pmx by combining the similarity value xszU and the weight qz(U):

[0074]

[0075] Where: b is the number of bonus dimensions, xszU represents the similarity value under the U-th dimension;

[0076] Step S43: Substitute the similarity value, initial score and weight of the optimized talent information into the matching score model for calculation to obtain the score of the optimized talent information, and recommend the optimized talent information based on the score.

[0077] Compared with the prior art, the present invention has the following beneficial effects:

[0078] Improve the personalization of user needs: Obtain user needs, perform text processing based on user needs, and formulate matching rules based on user needs, thereby improving the personalization of user needs and providing different matching rules for different users to meet the matching needs of different users;

[0079] Refine matching results: Analyze users' individual needs to obtain specific matching dimensions, and then perform matching based on these dimensions to make matching results more accurate. Perform multi-dimensional talent scoring on matching results to provide users with high-quality talent information.

[0080] Reduce interaction complexity: Collect statistics on talent information matching results, analyze the data distribution of talent information in different matching dimensions, set weights for matching dimensions based on the data distribution of different matching dimensions, reduce user operations, and lower the operational threshold; based on the weights, detail the matching process and optimize the matching results. BRIEF DESCRIPTION OF THE DRAWINGS

[0081] Other features, objects and advantages of the present invention will become more apparent upon reading the detailed description of non-limiting embodiments with reference to the following drawings:

[0082] Figure 1 Schematic diagram of the method of the present invention;

[0083] Figure 2 Schematic diagram of the data processing flow of the present invention;

[0084] Figure 3 This is a schematic diagram of talent information of the present invention. DETAILED DESCRIPTION

[0085] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0086] Example 1

[0087] See also Figure 1 , a talent information-based intelligent matching method includes:

[0088] Step S1: Obtain talent information and the company's employment demand information;

[0089] See also Figure 2 Step S2: Process the company's employment demand information, obtain employment demand standards, and divide the employment demand standards into hard requirements and flexible requirements; obtain necessary items based on hard requirements; obtain bonus items based on flexible requirements;

[0090] Step S21: Acquire the enterprise's employment demand information; perform structural field conversion on the enterprise's employment demand information to obtain employment demand standards, which are represented by standard dimensions and standard descriptions;

[0091] Step S211: Preprocess the text of the enterprise's employment demand information, extract the text of the employment demand information, perform cyclic judgment on the text, and delete the text content if it contains irrelevant characters, formatting marks, or stop words to obtain a clean text;

[0092] Step S2111: remove emoticons: obtain the Unicode code bq1 of the first emoticon in the computer, obtain the Unicode code bq2 of the last emoticon in the computer, and obtain the coding interval BQJ of the emoticons according to the coding range [bq1, bq2] of the emoticons; convert the employment demand information into character codes, count the number of character codes bm, and record the character codes as ZF(1), ZF(2), ..., ZF(bm); if the character codes ZF(1), ZF(2), ..., ZF(bm)∈BQJ, then replace the character codes with null;

[0093] The details are as follows:

[0094] Create an array cleaned_text based on the character codes ZF(1), ZF(2), ..., ZF(bm), cleaned_text = [ZF(1), ZF(2), ..., ZF(bm)]

[0095] Traverse the array:

[0096] for char in text:

[0097] When it is determined that the traversed element char belongs to an expression, it is set to empty;

[0098] if is_emoji(char):

[0099] cleaned_text.append("null")

[0100] else:

[0101] cleaned_text.append(char)

[0102] It should be noted that emoticons have a specific range in Unicode encoding (such as U+1F600 to U+1F64F). The encoding in this range is deleted to complete the deletion of emoticons.

[0103] Step S2112: Format tag removal: Format tag type z is obtained, and format tags are recorded as gs1, gs2, ..., gsz; a format list is constructed, and format tags gs1, gs2, ..., gsz are used as list elements to fill the format list. Based on the format list, the employment demand information is matched with gs1, gs2, ..., gsz, and the corresponding content of gs1, gs2, ..., gsz in the employment demand information is deleted to complete the cleaning of the format list;

[0104] The details are as follows:

[0105] The format mark is binary-encoded to obtain the length r of the binary code. The code position is recorded as x, and the value at the code position is recorded as y. A plane rectangular coordinate system is constructed, and the binary code of the format mark is mapped in the plane rectangular coordinate system. The point set {(x1, y1), (x2, y2), ..., (xr, yr)} of the format code is obtained. According to the point set {(x1, y1), (x2, y2), ..., (xr, yr)}, the cases of y = 0 and y = 1 are connected respectively, and their lengths are calculated:

[0106]

[0107] Where: cd0 represents the length when y=0; y w , x w Indicates the w-th x value and y value; through y w The -1 method retains the coordinate length with a y value of 0 and discards the length with a y value of 1;

[0108]

[0109] Among them: cd1 represents the length when y=0, through w The multiplication method retains the length of the coordinates with a y value of 1 and discards the length of 0 and 1.

[0110] According to cd0 and cd1, the employment demand information is matched and deleted to complete the cleaning of the format list.

[0111] Step S2113: Remove stop words: Obtain stop word type t, record the stop words as TY1, TY2, ..., TYt; construct a stop word list using the stop words TY1, TY2, ..., TYt as elements, express the stop word list using regular expressions, search for stop words in the job demand information, and delete the searched stop words;

[0112] It's important to note that stop words, in natural language processing (NLP) or text analysis, are words that have no real semantic value, appear frequently, and are unhelpful for analysis. Removing stop words can reduce noise and improve efficiency in keyword extraction, text classification, and information retrieval tasks.

[0113] Step S212: The cleaned text is segmented using the Jieba word segmentation tool. Combined with natural language processing technology, key entities in the text are identified and classified to obtain classification dimensions. Key entities under the same classification dimension are recorded as dimension content. Based on the classification dimension and dimension content, the employment requirement standard is obtained: {standard dimension: standard description}.

[0114] For example, a company's job requirements are: "We need a software engineer with more than three years of Java development experience, familiarity with the Spring framework, and good communication and teamwork skills."

[0115] After structural transformation, we get the employment requirement standards: {Position: "Software Engineer", Skills: ["Java", "Spring Framework"], Experience: "More than 3 years", Soft skills: ["Communication skills", "Teamwork spirit"]}.

[0116] Step S22: Extracting the hard requirements from the hiring demand standards, obtaining the standard dimensions and standard descriptions corresponding to the hard requirements, integrating them, recording the integrated standard dimensions as necessary dimensions, and recording the integrated standard descriptions as necessary descriptions, to obtain necessary items; necessary items: {necessary dimensions: necessary descriptions};

[0117] Extract the flexibility requirements in the employment demand standards, obtain the standard dimensions and standard descriptions corresponding to the flexibility requirements, integrate them, record the integrated standard dimensions as bonus dimensions, and record the integrated standard descriptions as bonus descriptions to obtain bonus items; bonus items: {bonus dimensions: bonus descriptions};

[0118] It should be noted that: rigid requirements refer to conditions that must be strictly met and cannot be changed, while flexible requirements refer to requirements that have a certain degree of flexibility and adjustability while meeting basic needs or goals, allowing appropriate changes and optimizations based on actual conditions;

[0119] Step S3: Process the talent information, filter the talent information by necessary items, and obtain optimized talent information; perform similarity matching between the optimized talent information and the bonus items to obtain similarity values; sort the similarity values, and assign initial scores based on the sorting results;

[0120] See also Figure 3 Step S31: Obtain talent information, obtain necessary dimensions and necessary descriptions based on necessary items, match and locate talent information based on necessary dimensions, match necessary descriptions with talent information, extract talent information that meets necessary items, and delete talent information that does not meet necessary items; thus, optimizing talent information is obtained.

[0121] Step S311: Obtain the number a of necessary dimensions, record the necessary dimensions as bwd1, bwd2, ..., bwda; record the necessary descriptions corresponding to the necessary dimension items as bms1, bms2, ..., bmsa; match and locate talent information in sequence according to the necessary dimensions; and match the necessary descriptions with the talent information based on the matching results;

[0122] Step S3111: Using the necessary dimension as a judgment condition, the talent information is traversed and the necessary dimension is matched with the talent information by character comparison. If the necessary dimension does not exist in the talent information, the talent information is deleted.

[0123] Step S3112: If the talent information contains a description of the necessary dimension, extract the corresponding content and match it with the necessary description, and delete the talent information based on the matching result;

[0124] Step S3113: According to steps S3111-S3112, the necessary dimensions bwd1, bwd2, ..., bwda are matched with the necessary descriptions bms1, bms2, ..., bmsa;

[0125] Step S312: If the necessary dimensions bwd1, bwd2, ..., bwda and the necessary descriptions bms1, bms2, ..., bmsa all have corresponding matching contents in the talent information, the talent information is retained; the retained talent information is integrated to obtain optimized talent information;

[0126] Step S3121: Build an optimized talent information list, obtain necessary dimensions bwd1, bwd2, ..., bwda and necessary descriptions bms1, bms2, ..., bmsa, and find corresponding matching talent information in the talent information. Delete the necessary dimensions and necessary descriptions in the talent information.

[0127] Step S3122: The talent information is simplified by performing an XOR operation on the necessary dimensions bwd1, bwd2, ..., bwda and the necessary descriptions bms1, bms2, ..., bmsa with the matching talent information to reduce data storage costs. The simplified talent information is used as a list element to populate the optimized talent information list.

[0128] Step S3123: sorting the elements in the optimized talent information list; obtaining the data storage value of the simplified talent information, and sorting the talent information according to the size of the data storage value;

[0129] It should be noted that the data storage value reflects the richness of talent information. Rich talent information can better reflect whether it is compatible with enterprise needs.

[0130] Obtain the number of elements ysl in the optimized talent information list, and record the elements in the optimized talent information list as ylb[ys]; use the first talent information in the optimized talent information list as the comparison value bj; and perform a cyclic comparison on the optimized talent information list based on the comparison value;

[0131] If bj>ylb[ys], replace ylb[ys] with bj and compare ylb[ys+1] until ys=ysl;

[0132] If bj≤ylb[ys], then compare ylb[ys+1] until ys=ysl;

[0133] When the loop is completed, the optimized talent information is sorted in descending order of the talent information data storage value; a sorted optimized talent information list is obtained, and the optimized talent information is called according to the sorted optimized talent information list;

[0134] Step S32: Based on the bonus items, obtain bonus dimensions and bonus descriptions; match the bonus dimensions with the optimized talent information, calculate similarity based on the bonus descriptions, and obtain similarity values; sort similar values ​​of the same dimension based on the bonus dimensions, and assign initial scores based on the sorting results;

[0135] Step S321: Obtain the number b of bonus dimensions, record the bonus dimensions as jwd1, jwd2, ..., jwdb; record the bonus descriptions corresponding to the bonus dimensions as jms1, jms2, ..., jmsb; use the bonus dimensions as search targets to search for optimized talent information; obtain search content, match the search content with the bonus descriptions, and obtain similarity values; the details are as follows:

[0136] Step S3211: Obtain search content, perform word segmentation on the search content, and obtain a search word list; obtain the number of search words c, and record the search words as js1, js2, ..., jsc; perform word segmentation on the bonus description, and obtain a description word list; obtain the number of description words d, and record the description words as ms1, ms2, ..., msd;

[0137] Step S3212: perform semantic matching on the search term js1 and the description terms ms1, ms2, ..., msd respectively to obtain semantic matching values, and obtain the maximum semantic matching value as the matching value xfz of the search term; judge the matching value of the search term,

[0138] If xfz>0.5, the search term and the description term are judged to be matched, and the matching value of the search term is saved;

[0139] If xfz≤0.5, the search term and the description term are judged to be unmatched and the search value is deleted;

[0140] Step S3213: Similarly, according to step S3213, js2, ..., jsc are matched with the description words ms1, ms2, ..., msd to obtain matching values ​​of the search words; the matching values ​​that can match the search words and the description words are counted to obtain a matching value list;

[0141] Step S3214: Obtain the number e of matching values ​​in the matching value list; perform cumulative calculation on the matching values ​​to obtain a similarity value xsz:

[0142]

[0143] Among them: xfz i represents the i-th matching value;

[0144] Step S322: According to steps S3211 to S3214, similarity values ​​of the b bonus dimensions are obtained to obtain a similarity value list xsl, where xsl = [xsz1, xsz2, ..., xszb]. The number f of optimized talent information is obtained. According to the number f of optimized talent information and the similarity value list xsl, the similarity values ​​of the b bonus dimensions are sorted in descending order. The sorting positions px are counted to obtain a sorted list bxl, where bxl = [px1, px2, ..., pxb].

[0145] Step S323: Initially assign points to the optimized talent information based on the sorted list; map the sorted positions to the interval [1, 2] for scoring. The specific scoring is as follows:

[0146] Obtain the ranking px, the number of optimized talent information f, and the mapping interval [1,2]. Perform a mapping calculation on the ranking px according to the mapping interval [1,2] to obtain the initial score cff.

[0147]

[0148] It should be noted that the initial score is limited by mapping the ranking position px to the interval [1, 2] to prevent the data from being too large;

[0149] For example, for the ranking position 1, its similarity value is the largest, so its initial score is close to 2; for the ranking position f, its similarity value is the smallest, so its initial score is 1; the initial score is limited to between 1 and 2 through mapping.

[0150] Step S4: Set weights based on the numerical distribution of similarity values; construct a matching scoring model based on the similarity values, initial scores, and weights to score the talents;

[0151] Step S41: Obtain the similarity values ​​of b bonus dimensions, calculate the variance of the similarity values ​​of the b bonus dimensions, and set weights for the score calculation of the bonus dimensions based on the variance of the similarity values, as follows:

[0152] Obtain the number f of optimized talent information. Based on the number f of optimized talent information, record the similarity values ​​of the b bonus dimensions as A(U,V). Where A(U,V) represents the similarity value of the U-th bonus dimension and the V-th optimized talent information.

[0153] Calculate the mean of similarity values ​​under the same dimension to obtain the mean of similarity values ​​xjz;

[0154]

[0155] Where: xjz(U) represents the mean similarity value under the U-th dimension, A(U,V) represents the similarity value of the U-th bonus dimension and the V-th optimized talent information;

[0156] According to the similarity mean xjz(U) and the similarity value A(U,V), the variance fc is calculated:

[0157]

[0158] Where: fc(U) represents the variance in the U-th dimension;

[0159] Set the variance as the calculation weight qz of the bonus dimension, that is, fc(U) = qz(U);

[0160] It's important to note that variance is a key metric used in statistics to measure the degree of dispersion in a set of data. It represents the average squared distance between data points and the mean. By calculating variance, we can understand the magnitude of data fluctuations, providing a basis for data analysis and model building.

[0161] Step S42: Obtain the initial score cff(U), and build a matching score model pmx by combining the similarity value xszU and the weight qz(U):

[0162]

[0163] Where: b is the number of bonus dimensions;

[0164] It should be noted that the data is restricted by the initial scoring method to prevent the data from being inaccurate due to some similarity values ​​being 0. At the same time, it is summed with the similarity value xszU to better express the difference, and the score is scaled by weight; the weight is obtained based on the variance. A large variance means that the difference between the data is large and more representative; it is amplified;

[0165] Step S43: Substitute the similarity value, initial score, and weight of the optimized talent information into the matching scoring model for calculation to obtain the score of the optimized talent information. Recommend the optimized talent information based on the score, and proceed to step S5;

[0166] Step S5: Sort the scoring results and recommend talents to the enterprise based on the enterprise's employment demand information;

[0167] Step S51: Obtain the scoring results and sort the scores in descending order to obtain a scoring ranking; obtain the actual number of talent needs s and the company's recruitment ratio h based on the company's recruitment demand information; calculate the recommended number tjs of the company based on the actual number of talent needs s and the company's recruitment ratio h;

[0168] tjs=s×(1+h);

[0169] It should be noted that the enterprise's demand ratio means that if the enterprise needs 10 people, but during the actual interview or notification process, the number of people will be increased by a certain ratio. This ratio is defined as the expansion ratio in the present invention.

[0170] Step S52: Obtain the number f of optimized talent information and compare the number f of optimized talent information with the recommended number tjs:

[0171] If f≤tjs, all optimized talent information will be recommended;

[0172] If f>tjs, obtain the score ranking and recommend the optimized talent information with scores ranging from 1 to tjs.

[0173] The above formulas are all dimensionless and calculated by taking their numerical values. The formula is a formula for the most recent real situation obtained by collecting a large amount of data and performing software simulation. The preset parameters in the formula are set by technicians in this field according to actual conditions. For example, if there are weight coefficients and proportional coefficients, the size of the settings is to quantify each parameter to obtain a specific value, which is convenient for subsequent comparison. Regarding the size of the weight coefficient and the proportional coefficient, as long as it does not affect the proportional relationship between the parameter and the quantized value, it is fine.

[0174] Finally, it should be noted that the above-described embodiments are only specific implementations of the present invention, which are used to illustrate the technical solutions of the present invention, rather than to limit them. The scope of protection of the present invention is not limited thereto. Although the present invention has been described in detail with reference to the above-described embodiments, those skilled in the art should understand that any person skilled in the art can modify or easily conceive of changes to the technical solutions described in the above-described embodiments within the technical scope disclosed by the present invention, or replace some of the technical features therein with equivalents. Such modifications, changes, or replacements do not deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should be included in the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be based on the scope of protection of the claims.

Claims

1. A method for intelligent matching based on talent information, characterized in that: The matching method includes: Step S1: Obtain talent information and the company's employment demand information; Step S2: Process the enterprise's employment demand information to obtain employment demand standards, which are divided into hard requirements and flexible requirements; obtain necessary items based on hard requirements; and obtain bonus items based on flexible requirements; Step S3: Process the talent information, filter the talent information by necessary items, and obtain optimized talent information; perform similarity matching between the optimized talent information and the bonus items to obtain similarity values; sort the similarity values, and assign initial scores based on the sorting results; Step S4: weights are set based on the numerical distribution of similarity values; a matching scoring model is constructed based on the similarity values, initial scores, and weights to score the talent; Step S5: Sort the scoring results and recommend talents to the enterprise based on the enterprise's employment demand information.

2. The method according to claim 1, wherein: The specific steps of step S2 are as follows: Step S21: Acquire the enterprise's employment demand information; perform structural field conversion on the enterprise's employment demand information to obtain employment demand standards, which are represented by standard dimensions and standard descriptions; Step S22: Extracting the hard requirements from the hiring demand standards, obtaining the standard dimensions and standard descriptions corresponding to the hard requirements, integrating them, recording the integrated standard dimensions as necessary dimensions, and recording the integrated standard descriptions as necessary descriptions, to obtain necessary items; necessary items: {necessary dimensions: necessary descriptions}; Extract the flexibility requirements in the employment demand standards, obtain the standard dimensions and standard descriptions corresponding to the flexibility requirements, integrate them, record the integrated standard dimensions as bonus dimensions, and record the integrated standard descriptions as bonus descriptions to obtain bonus items; bonus items: {bonus dimensions: bonus descriptions}.

3. The method according to claim 2, wherein: The specific steps of step S21 are as follows: Step S211: Preprocess the text of the enterprise's employment demand information, extract the text of the employment demand information, perform cyclic judgment on the text, and delete the text content if it contains irrelevant characters, formatting marks, or stop words to obtain a clean text; Step S212: Perform word segmentation on the cleaned text, identify key entities in the text, classify the key entities, obtain classification dimensions, and record the key entities under the same classification dimension as dimension content; based on the classification dimension and dimension content, obtain the employment demand standard: {standard dimension: standard description}.

4. The method according to claim 3, wherein: The specific steps of step S211 are as follows: Remove emoticons: obtain the Unicode code bq1 of the first emoticon in the computer, obtain the Unicode code bq2 of the last emoticon in the computer, and obtain the emoticon code range BQJ according to the emoticon code range [bq1, bq2]; convert the employment demand information into character codes, count the number of character codes bm, and record the character codes as ZF(1), ZF(2), ..., ZF(bm). If the character codes ZF(1) to ZF(bm)∈BQJ, replace the character codes with null; Remove formatting tags: Get the type of formatting tag z and record the formatting tags as gs1, gs2, ..., gsz; build a formatting list, use the formatting tags gs1 to gsz as list elements, fill the formatting list, match the employment demand information with gs1 to gsz according to the formatting list, delete the corresponding content of gs1 to gsz in the employment demand information, and complete the cleaning of the formatting list; The details are as follows: The format mark is binary-encoded to obtain the length r of the binary code; the code position is recorded as x, and the value at the code position is recorded as y, a plane rectangular coordinate system is constructed, and the binary code of the format mark is mapped in the plane rectangular coordinate system; the format code point set {(x1, y1), (x2, y2), ..., (xr, yr)} is obtained. According to the point set {(x1, y1) to (xr, yr)}, the points with y values ​​of 0 and 1 are connected respectively, and the length of the connection is calculated: Where: cd0 represents the connection length when the y value is 0; x w and y w Indicates the w-th x value and y value; Where: cd1 represents the connection length when the y value is 1; According to cd0 and cd1, the employment demand information is matched and deleted to complete the cleaning of the format list; Remove stop words: Get the stop word type t and record the stop words as TY1, TY2, ..., TYt; use stop words TY1 to TYt as elements to construct a stop word list, express the stop word list using regular expressions, search for stop words in the job demand information, and delete the searched stop words.

5. The method according to claim 1, wherein: The specific steps of step S3 are as follows: Step S31: Obtain talent information, obtain necessary dimensions and necessary descriptions based on necessary items, match and locate the talent information based on the necessary dimensions, match the necessary descriptions with the talent information, extract talent information that meets the necessary items, and delete talent information that does not meet the necessary items, thereby obtaining optimized talent information; Step S32: According to the bonus items, obtain the bonus dimensions and bonus descriptions; match the bonus dimensions with the optimized talent information, calculate the similarity in combination with the bonus descriptions, obtain similarity values, sort the similar values ​​of the same dimensions according to the bonus dimensions, and perform initial scoring based on the sorting results.

6. The method according to claim 5, characterized in that: The specific steps of step S31 are as follows: Step S311: Obtain the number a of necessary dimensions, record the necessary dimensions as bwd1, bwd2, ..., bwda; record the necessary descriptions corresponding to the necessary dimension items as bms1, bms2, ..., bmsa; match and locate talent information in sequence according to the necessary dimensions; and match the necessary descriptions with the talent information based on the matching results; Using the necessary dimensions as the judgment criteria, the talent information is traversed and the necessary dimensions are matched with the talent information through character comparison. If the talent information does not contain a description of the necessary dimensions, the talent information is deleted. If there is a description of the necessary dimension in the talent information, the corresponding content is extracted and matched with the necessary description, and the talent information is deleted based on the matching result; Match the necessary dimensions bwd1 to bwda with the necessary descriptions bms1 to bmsa; Step S312: If the necessary dimensions bwd1 to bwda and the necessary descriptions bms1 to bmsa can all be matched in the talent information, the talent information is retained; Integrate the retained talent information to obtain optimized talent information.

7. The method according to claim 6, characterized in that: The specific steps of step S312 are as follows: Build an optimized talent information list, obtain the necessary dimensions bwd1, bwd2, ..., bwda and the necessary descriptions bms1, bms2, ..., bmsa, and obtain corresponding matching talent information in the talent information. Delete the necessary dimensions and necessary descriptions in the talent information. The talent information is simplified by performing an XOR operation on the necessary dimensions bwd1 to bwda and the necessary descriptions bms1 to bmsa with the matching talent information, and the simplified talent information is used as a list element to fill the optimized talent information list; Sorting the elements in the optimized talent information list; Obtain the data storage value of the simplified talent information, and sort the talent information according to the size of the data storage value; Obtain the number of elements ysl in the optimized talent information list, and record the elements in the optimized talent information list as ylb[ys]; use the first talent information in the optimized talent information list as the comparison value bj; and perform a cyclic comparison on the optimized talent information list based on the comparison value; If bj>ylb[ys], replace ylb[ys] with bj and compare ylb[ys+1] until ys=ysl; If bj≤ylb[ys], then compare ylb[ys+1] until ys=ysl; When the loop is completed, the optimized talent information is sorted in descending order of the talent information data storage value; a sorted optimized talent information list is obtained, and the optimized talent information is called according to the sorted optimized talent information list.

8. The method according to claim 5, wherein: The specific steps of step S32 are as follows: Step S321: Obtain the number b of bonus dimensions, record the bonus dimensions as jwd1, jwd2, ..., jwdb; record the bonus descriptions corresponding to the bonus dimensions as jms1, jms2, ..., jmsb; use the bonus dimensions as search targets to search for optimized talent information; obtain search content, match the search content with the bonus descriptions, and obtain similarity values; Step S322: Obtain the similarity values ​​of the b bonus dimensions to obtain a similarity value list xsl, where xsl = [xsz1, xsz2, ..., xszb]; obtain the number f of optimized talent information; sort the similarity values ​​of the b bonus dimensions in descending order based on the number f of optimized talent information and the similarity value list xsl; count the sorting positions px to obtain a sorted list bxl, where bxl = [px1, px2, ..., pxb]; Step S323: Initially assign points to the optimized talent information based on the sorted list; map the sorted positions to the interval [1, 2] for scoring. The specific scoring is as follows: Obtain the ranking px, the number of optimized talent information f, and the mapping interval [1,2]. Perform a mapping calculation on the ranking px according to the mapping interval [1,2] to obtain the initial score cff.

9. The method according to claim 8, characterized in that: The specific steps of step S321 are as follows: Obtain search content, perform word segmentation on the search content, and obtain a search word list; obtain the number of search words c, and record the search words as js1, js2, ..., jsc; perform word segmentation on the bonus description to obtain a description word list; obtain the number of description words d, and record the description words as ms1, ms2, ..., msd; The search word js1 is matched with the description words ms1 to msd respectively to obtain the meaning matching value, and the maximum meaning matching value is obtained as the matching value xfz of the search word; the matching value of the search word is judged, If xfz>0.5, the search term and the description term are judged to be matched, and the matching value of the search term is saved; If xfz≤0.5, the search term and the description term are judged to be unmatched and the search value is deleted; Similarly, match js2 to jsc with the description words ms1 to msd to obtain the matching values ​​of the search words; count the matching values ​​that can match the search words and description words to obtain a matching value list; Get the number of matching values ​​e in the matching value list; add up the matching values ​​to get the similarity value xsz: Among them: xfz i Indicates the i-th matching value.

10. The method according to claim 1, wherein: The specific steps of step S4 are as follows: Step S41: Obtain the similarity values ​​of b bonus dimensions, calculate the variance of the similarity values ​​of the b bonus dimensions, and set weights for the score calculation of the bonus dimensions based on the variance of the similarity values, as follows: Obtain the number f of optimized talent information, and based on the number f of optimized talent information, record the similarity values ​​of the b bonus dimensions as A(U,V); Calculate the mean of similarity values ​​under the same dimension to obtain the mean of similarity values ​​xjz; Where: xjz(U) represents the mean similarity value under the U-th dimension, A(U,V) represents the similarity value of the U-th bonus dimension and the V-th optimized talent information; According to the similarity mean xjz(U) and the similarity value A(U,V), the variance fc is calculated: Where: fc(U) represents the variance in the U-th dimension; Set the variance as the calculation weight qz of the bonus dimension, that is, fc(U) = qz(U); Step S42: Obtain the initial score cff(U), and build a matching score model pmx by combining the similarity value xszU and the weight qz(U): Where: b is the number of bonus dimensions, xszU represents the similarity value under the U-th dimension; Step S43: Substitute the similarity value, initial score and weight of the optimized talent information into the matching score model for calculation to obtain the score of the optimized talent information, and recommend the optimized talent information based on the score.

Images