Matching support program and matching support system

Abstract

We provide a matching support program that enables realistic and efficient matching. [Solution] The method includes: a question presentation step of presenting a selector with a pair of questions consisting of negative expressions corresponding to evaluation factors used in matching; an evaluation score addition / subtraction step of adding points to the evaluation score for evaluation factors related to the questions selected by the selector and subtracting points for evaluation factors related to questions not selected by the selector; a score leveling step of leveling the evaluation scores for all evaluation factors so that the average of the evaluation scores for multiple evaluation factors becomes zero; an evaluation factor extraction step of extracting evaluation factors whose evaluation scores are lower than a predetermined threshold as rejection factors and evaluation factors whose evaluation scores are above a predetermined threshold as tolerance factors; and a matching determination step of determining matching based on the rejection factors and tolerance factors.

Description

Technical Field

[0001] The present invention relates to a matching support program and a matching support system.

Background Art

[0002] In job hunting activities, marriage activities, etc., a person matching support system that supports the determination of the degree of compatibility between people is known. In the conventional person matching support system, a method of narrowing down the selected person centering on the desired conditions (for example, work experience, area of expertise, physical characteristics, annual income, personality diagnosis results, etc.) presented by the selector has been adopted.

[0003] In the conventional person matching support system, various devices for effectively realizing matching are known. For example, a system that visualizes the process when hiring workers to improve the efficiency of selection is known (see Patent Document 1).

Prior Art Documents

Patent Documents

[0004]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0005] In the prior art exemplified in Patent Document 1, a method of quantifying the answers input by the selected person to the questions prepared by the selector as information indicating psychological characteristics and personality tendencies is used. These quantified informations are mainly for analyzing the tendency of "preference". Therefore, the tendency of the quantified information based on the answers to the desired conditions may deviate greatly from the reality.

[0006] This invention has been made in view of the problems of the prior art described above, and aims to provide a matching support program that uses rejection factors in the factors used by selectors to select those to be selected, thereby enabling realistic and efficient matching. [Means for solving the problem]

[0007] To solve the above problems, one aspect of the present invention is a matching support program that causes a computer to perform information processing to support the matching of a selector and a selected person, comprising: a question presentation step of presenting a pair of questions consisting of negative expressions corresponding to evaluation factors used in the matching to the selector; an evaluation score addition / subtraction step of adding points to the evaluation score of the evaluation factor related to the question selected by the selector from the presented questions, and subtracting points from the evaluation score of the evaluation factor related to the question not selected by the selector; a score leveling step of equalizing the number of times the questions related to all the evaluation factors used in the matching are presented so that the average of the evaluation scores for a plurality of evaluation factors becomes zero, thereby leveling the evaluation scores; an evaluation factor extraction step of extracting evaluation factors from the leveled evaluation factors whose evaluation scores are lower than a predetermined threshold as rejection factors, and evaluation factors whose evaluation scores are above a predetermined threshold as tolerance factors; and a matching determination step of determining the matching of the selected person to the selector based on the rejection factors and the tolerance factors. [Effects of the Invention]

[0008] According to the present invention, realistic and efficient matching can be achieved by using rejection factors as indicators for selectors to choose non-selectors. [Brief explanation of the drawing]

[0009] [Figure 1] A block diagram showing the overall configuration of a matching system according to one embodiment of the present invention. [Figure 2] A block diagram showing an example of the hardware configuration of the matching system according to this embodiment. [Figure 3] A block diagram showing the functional configuration of the matching system according to this embodiment. [Figure 4] A flowchart illustrating the overall flow of the matching process in this embodiment. [Figure 5] A flowchart illustrating the process for extracting rejection factors / tolerance factors in this embodiment. [Figure 6] A diagram illustrating the process of extracting rejection factors / tolerance factors in this embodiment. [Figure 7] This figure shows an example of a screen displaying questions to the selector in this embodiment. [Figure 8] This figure shows an example of a display screen for the diagnostic results in this embodiment. [Modes for carrying out the invention]

[0010] [Summary of the present invention] This invention is based on the known personality assessment method "Big Five" or similar theories with a psychological or statistical background. The person being selected (the person being selected) has their temperament measured according to the theory, and the person doing the selecting (the selector) is presented with elements that are negative versions of the temperament they wish to measure. This allows the selector to be asked, "If you had to choose, which would you accept?", and by quantifying the answer, the invention analyzes the degree of match between the selector's desired tendencies and the selected person's tendencies, providing a solution for mutual matching.

[0011] For example, in a job-hunting scenario, the "job seeker" is the one being selected, and the "company" that wants to hire talent is the one doing the selecting. It should be noted that the company itself does not determine the criteria for selecting individuals; in reality, the hiring manager, taking into account the company's objectives, will formulate the selection criteria. This hiring manager will then create the questions presented to determine whether an individual matches the company's desired profile.

[0012] Furthermore, considering the scenario of job hunting, it's conceivable that the job seeker becomes the "selector" and the company becomes the "selected." In this case, the job seeker would formulate the criteria for selecting a company. That is, they would create questions to help the job seeker determine which companies match their "ideal company profile."

[0013] In the context of marriage, both parties can be either the selector or the selected. For convenience, the following explanation will use the terms "selected" for the party being selected and "selector" for the party doing the selecting.

[0014] [Summary of Embodiments According to the Present Invention] The following outlines embodiments of the present invention. The present invention is mainly composed of a combination of technical ideas such as "measurement of factors of those being selected," "measurement of rejection tendencies of those who select," "weighting and equalization of factor scores," "extraction of rejection and tolerance factors," "updating through feedback," "factor-based search," "combination with positive preferences," "dynamic question generation," "non-psychological definition of factors," "extraction without testing," and "display and application."

[0015] [Regarding the measurement of factors in the selected individuals] "Factor measurement of selected individuals" means analyzing data such as speech, responses, behavior, documents, or physiological indicators of selected individuals using conventionally used psychological theories such as the Big Five or HEXACO, or using unique factors extracted through statistical analysis comparing high-performing and low-performing groups in companies or organizations, machine learning, natural language processing, generative AI, or other mathematical or conceptual methods, and conducting a personality trait diagnostic test.

[0016] For example, the question format could be a multiple-choice format on a 5-point scale, or it could be a free-response format where AI (Artificial Intelligence) analyzes and extracts semantic vectors. Alternatively, the AI ​​could transcribe utterances into text, and this text could be analyzed as free-response text to obtain semantic vectors. The results would then be quantized and stored in a database.

[0017] [Regarding the measurement of the selector's rejection tendency] The options presented to the selector are in the form of questions. In these questions, the expressions of each evaluation factor are converted into negative expressions (hereinafter referred to as "negative expressions"). Also, the negative expressions are changed to form a pair of options. For example, a pair of negative expressions for a certain factor are "shyness" and "selfishness". Thus, a pair of options converted into negative expressions corresponding to a certain inch are presented, and the question is something like "If forced to say, which one would you tolerate?" In this way, questions in a form where the presented pair of negative expressions are selected in a contextual expression or story form are generated and presented. The value of the evaluation score of the factor selected by the selector is increased, and the score of the factor not selected is decreased.

[0018] Therefore, if the pair of negative expressions used in the question for evaluating a certain factor is "shyness" or "selfishness", etc., only the evaluation score of the selected side needs to be increased, and the evaluation score of the unselected side is decreased, but it is not limited to this. <0​​​​​​​​​​​​The weighting and equalization of factor scores involves several steps, including, for example, "weighting and score design using core and peripheral questions," "scoring using binary negative questions," "question structure and trial control," "score processing and equalization of distribution of selection results," "diversification and rephrasing of negative descriptions," and "weighting and score design using core and peripheral questions."

[0022] In the "Weighting and Score Design using Core and Peripheral Questions" method, multiple questions are prepared for each factor using different expressions and perspectives, and scoring is performed. Questions that evaluate the core characteristics of a factor are given a higher weight, while peripheral questions are given a lighter weight. This weighting is designed to make the evaluation of the factors clearer.

[0023] Furthermore, in the "scoring using negative questions in a two-choice format," the questions are transformed into negative statements that represent the deficiency of each factor. For example, short sentences such as "shy" are presented for the "extroversion" factor, and "selfish" for the "agreeableness" factor. The score is then adjusted by selecting which deficiency is tolerable. Pairs are displayed randomly or according to a pre-designed pattern, and all factors are presented equally.

[0024] Furthermore, in "Question Structure and Trial Control," the question structure is designed so that all factors have the same number of trials. By always keeping the number of questions even, the question structure is generated so that the final addition and subtraction result of the score approaches zero.

[0025] Furthermore, in "Score Processing and Distribution Leveling of Selection Results," two negative factors are presented, and "+1" is added to the score of the factor that was chosen, while "-1" is added to the score of the factor that was not chosen. By performing the same processing on all factors, the true rejection and acceptance tendencies of the selectors are statistically extracted.

[0026] Furthermore, in "Diversification and Rephrasing of Negative Descriptions," using multiple expressions such as "failure to keep promises," "shifting blame," and "not taking responsibility for one's words" for the same factor makes it more difficult for respondents to discern the intent, while also providing more accurate criteria for judgment.

[0027] Furthermore, in the "Weighting and Score Design using Core and Peripheral Questions," multiple questions using different expressions and perspectives are prepared for each factor, and scoring is performed. Questions that evaluate the core characteristics of a factor are given a higher weight, while peripheral questions are given a lighter weight. This weighting is designed to make the evaluation of the factors clearer.

[0028] [Regarding the extraction of rejection / tolerance factors] After the test is completed, the factor with the lowest aggregate score (including up to two ties) is extracted as the rejection factor, and the factors with a score statistically greater than +1σ (including up to two ties) are extracted as the tolerance factor. If the score distribution is even, all are positive, or all are negative, or if there are three or more ties, it is determined to be "no preference" (no rejection or tolerance).

[0029] [Regarding updates based on feedback] The system dynamically retrains the selector's thresholds for rejection and tolerance factors by evaluating the selected individuals on a 5-point scale after actual interviews or dates. This process may be automatically executed by AI. The automatic execution (automatic updating) by AI may include extracting rejection and tolerance factors and adjusting parameters such as thresholds from the free speech of the selected individuals interacting with the AI ​​verbally.

[0030] [About factor-based searches] Based on the extracted rejection and tolerance factors, targets are searched from the target database using numerical conditions or semantic vectors. Threshold settings may also be based on absolute values ​​or standard deviations. Furthermore, this filtering may include dynamic filtering using machine learning based on feedback from some users (selectors and selected) or accumulated knowledge.

[0031] [Regarding use in conjunction with positive preference] In addition to rejection factors, the system also supports matching based on four factors ("reject," "tolerance," "desirable," and "undesirable") by using a "positive preference test" that allows users to actively select elements they like.

[0032] [Regarding dynamic question generation] AI may be used to dynamically generate negative questions. This could involve including multiple factors in a single question, creating a story from which the user selects unacceptable behaviors, or creating and displaying "selfish" behaviors as sentences by making the factors negative. Furthermore, a user interface (UI) that allows users to select these displays in a highlighting format is also included in the embodiments of this invention.

[0033] [Regarding the non-psychological definition of factors] Factors are not necessarily limited to psychological models; they may be defined independently based on performance indicators, observational data, and cultural standards within the company.

[0034] [Regarding extraction without testing] Alternatively, the group of candidates could be classified as "favorable / neutral / unfavorable," and then rejection / tolerance factors could be extracted by working backward from that classification.

[0035] [Regarding display and application] The diagnostic results will be displayed, sorted, and labeled with warnings on the UI, and notifications can also be sent via email or paper. To reduce the burden on the user, instead of having them take the test all at once, questions could be distributed daily, and the filters could be automatically updated periodically by performing aggregated analysis once sufficient data has been collected.

[0036] [Details of embodiments according to the present invention] Next, embodiments of the present invention will be described in detail with reference to the drawings. In this embodiment, a matching support system provided as an application that runs on an information terminal such as a smartphone or PC will be described as an example.

[0037] [System Configuration] Figure 1 is an overall configuration diagram of a matching system 1 according to one embodiment of the present invention. As shown in Figure 1, the matching system 1 mainly consists of a support server 100, an information processing terminal 200 owned by a user 300, and a communication network 400. The support server 100 and the information processing terminal 200 are configured to communicate with each other via the communication network 400.

[0038] Users 300 are divided into selectors 310, who are users 300 who select personnel, and selected users 320, who are users 300 who are selected. Figure 1 illustrates a situation where there is one selector 310 and two selected users 320, but the number of users 300 who can use the matching system 1 simultaneously, and the ratio of selectors 310 to selected users 320, are not limited to the example in Figure 1.

[0039] The support server 100 performs information communication processing between the information processing terminals 200 owned by each user 300 via the communication network 400. As will be described in detail later, the support server 100 presents the selector 310 with multiple questions in order to determine the selector 310's desired conditions for narrowing down the list of candidates to be selected 320, and performs a matching process with the candidates to be selected 320 based on the selector 310's answers.

[0040] The questions presented here consist of short sentences containing evaluation factors that enable the evaluation of the character of the selected person 320, in order to identify a selected person 320 who fits the desired character profile of the selector 310. The questions are generated so that the evaluation factors used in the short sentences are negative. The questions are generated so that they contain a pair of short sentences containing negative expressions corresponding to the evaluation factors. That is, the question contains short sentence A containing a negative expression X1 for evaluation factor X, and short sentence B containing another negative expression X2. The questions are generated in a format that asks which of short sentence A or short sentence B is "acceptable".

[0041] Then, the selector 310 selects either "the expression that includes short sentence A" or "the expression that includes short sentence B" on the information processing terminal 200 in response to the presented question. This selection result is notified from the information processing terminal 200 to the support server 100. Based on the notified selection result, the support server 100 calculates an evaluation score based on the selector 310's desired conditions, identifies the desired conditions based on "rejection factors" and "tolerance factors" based on the evaluation core, and searches for the selected person 320.

[0042] As described above, the support server 100, acting as a matching support device, presents the selector 310 with questions containing pairs of negative short sentences corresponding to multiple evaluation factors via the information processing terminal 200. Then, by having the selector 310 choose "which is more unacceptable" for each question, the support server 100 performs a process to obtain information to identify the desired person profile.

[0043] The support server 100 then uses the selection results of the selector 310 to determine the ideal person profile desired by the selector 310. This includes questions containing pairs of negative short sentences corresponding to multiple evaluation factors. Based on the selector 310's selections, the server adds points to the selected factor and deducts points from the unselected factor. The points added or deducted here are the evaluation scores corresponding to the factors (evaluation factors).

[0044] The support server 100 then controls the number of times it presents questions to the selector 310 so that the number of times each question is presented is equal for all evaluation factors. It also performs a leveling process to bring the average evaluation score closer to zero by adding or subtracting evaluation scores based on core and peripheral weights for each question.

[0045] The support server 100 then extracts and labels one or two evaluation factors that have an evaluation score below a predetermined threshold, or that have the lowest evaluation score, as "rejection factors," based on the standardized evaluation score. It also extracts and labels evaluation factors that have an evaluation score above a predetermined threshold as "tolerance factors."

[0046] The support server 100 then searches a database (selected person DB) containing evaluation factor information for selected persons 320 based on "rejection factors" and "tolerance factors," and performs a matching process to extract selected persons 320 that meet the desired conditions of the selector 310.

[0047] The support server 100 then receives input from the selector 310 regarding the interview results with the selected person 320 extracted through the matching process, and executes a process to dynamically update the selector 310's score or filter conditions based on that feedback. This process may also be executed by so-called artificial intelligence using a learning model constructed through a machine learning process.

[0048] [Hardware configuration of support server 100] The support server 100, which constitutes the matching system 1 described in Figure 1, will now be explained in more detail. Figure 2 is a block diagram showing an example of the hardware configuration of the support server 100, which is a computer. The support server 100 is equipped with a CPU 101, RAM 102, ROM 103, storage device 104, communication interface 105, etc., and these are interconnected via a bus 106. The storage device 104 stores programs and data for realizing this system, and the CPU 101 loads these into RAM 102 and executes them, thereby realizing the various functions described later.

[0049] [Functional Configuration of Support Server 100] Figure 3 is a block diagram showing the functional configuration realized by the support server 100. The support server 100 includes a control unit 110, a selector DB 211, a selected person DB 212, and a question DB 213.

[0050] The control unit 110 functions as a factor acquisition unit 201, a question presentation unit 202 as a selection acquisition unit, a score processing unit 203, a factor extraction unit 204, a selected person search unit 205 as a matching unit, and an AI processing unit 206.

[0051] The factor acquisition unit 201 executes a factor acquisition test on the selected person 320. More specifically, it executes a factor acquisition test to measure the factors (evaluation factors) related to the selected person 320, and then performs a process to identify the factors of the selected person 320, which will be used as search conditions when the selector 310 selects evaluation factors included in the questions and then performs matching with the selected person 320. The factor acquisition test is transmitted from the support server 100 to the information processing terminal 200 held by the selected person 320. The selected person 320 answers the questions presented in the factor acquisition test via the information processing terminal 200, and records their own factors in the selected person DB 212 held by the support server 100.

[0052] The question presentation unit 202, acting as a selection acquisition unit, presents the selector 310 with questions in the question presentation step that include short sentences using pairs of negative expressions corresponding to multiple evaluation factors, and asks the selector to choose "which is less acceptable" for each question. This allows the unit to acquire the more acceptable evaluation factor and notify the score processing unit 203. The question presentation unit 202 constitutes the question presentation means.

[0053] In the score processing unit 203, during the evaluation score addition / subtraction step, "+1" is added to the score (evaluation score) of the evaluation factor related to the question selected by the selector 310. The score processing unit 203 also performs a process of adding "-1" to the evaluation score of the evaluation factor related to the question not selected by the selector 310. In other words, it adds to the evaluation score of the selected evaluation factor and subtracts from the evaluation score of the unselected evaluation factor. Furthermore, in the score standardization step, after all factors used to match the selected person 320 and the selector 310 have been asked an equal number of questions, the score processing unit 203 adds and subtracts evaluation scores based on core and peripheral weights for each question. In this way, the score processing unit 203 performs a process of leveling the scores so that the average of the evaluation scores approaches zero. The score processing unit 203 constitutes an evaluation score addition / subtraction means and a score leveling means.

[0054] In the evaluation factor extraction step, the factor extraction unit 204 extracts one or two factors with the lowest scores as rejection factors based on the standardized evaluation scores. It also performs a process to extract evaluation factors with evaluation scores above a threshold as tolerance factors. The factor extraction unit 204 constitutes the evaluation factor extraction means.

[0055] The selected person search unit 205, acting as a matching unit, searches the selected person DB 212, which contains factor information of selected persons 320, based on the extracted rejection factors and tolerance factors in the matching determination step, and performs the process of extracting selected persons 320 that match the selector 310. The selected person search unit 205 constitutes the matching determination means.

[0056] In the AI ​​processing step, the AI ​​processing unit 206 performs a process to dynamically update the score or filter conditions of the corresponding selector 310 using the feedback information provided by the selector 310 as training data, when the selector 310 conducts an interview or other interaction with the selector 320 extracted as a result of the matching determination by the selector search unit 205, and the selector 310 inputs feedback information regarding the extraction results.

[0057] In other words, the AI ​​processing unit 206 collects the input feedback information and uses it as training data to analyze the correlation between the feedback scores of the selected individuals 320 who were highly rated by the selector 310 and the factor scores of those selected individuals 320. It also retrains and automatically updates the thresholds for extracting rejection factors. As a result, the more the matching system 1 is used, the more accurately the selector 310's potential tolerance range and rejection tendencies are reflected. The AI ​​processing unit 206 then performs a threshold update step.

[0058] Selector DB211 stores information about selector 310 (ID, password, past selection history, extracted rejection factors and tolerance factors, etc.). Selected person DB212 stores information about selected person 320 (ID, profile, results of factor measurement described later, etc.).

[0059] [Matching support processing] Next, the flow of the matching support process in this embodiment will be described. Figure 4 is a flowchart showing the overall flow of the matching support process in this embodiment.

[0060] First, the selected participant 320 accesses the support server 100 using their information processing terminal 200 and registers as a member. At that time, the support server 100, specifically the factor acquisition unit 201, conducts a factor measurement test on the selected participant 320 (S401). The evaluation factors used in this factor measurement test are based on psychologically established theoretical models such as the "Big Five" (extraversion, agreeableness, conscientiousness, neuroticism, openness) or "HEXACO." Alternatively, evaluation factors defined by statistical comparison of high-performing and low-performing groups within an organization may be used. Or, not limited to psychological models, evaluation factors unique to that organization may be designed based on corporate culture, work characteristics, organizational behavior indicators, etc. That is, evaluation factors unique to that organization (e.g., "autonomy," "customer orientation," etc.) extracted by statistically comparing and analyzing the characteristics of high-performing and low-performing groups within a particular company or organization may be used. The scores for each factor obtained as a result of this evaluation test are associated with the selected person's information and stored in the selected person DB212.

[0061] Next, the selector 310 registers as a member with the support server 100. The control unit 110 of the support server 100 conducts an evaluation factor test on the selector 310 to extract rejection factors / tolerance factors (S402). Details of this evaluation factor test will be described later.

[0062] The process loops until all evaluation factor tests have been performed (S403: No). Once all evaluation factor tests have been performed (S403: YES), the rejection and tolerance factors for each selector 310 are extracted and stored in the selector DB211.

[0063] Next, the selected person search unit 205 of the support server 100 performs a search for selected persons 320 by the selector 310 (S404). In step S404, the selected person DB 212, which contains factor information of selected persons 320, is searched based on rejection factors and tolerance factors. As a search result, selected persons 320 that match the selector 310 are extracted, and the matching results are displayed on the display device of the information processing terminal 200 held by the selector 310.

[0064] Next, the selector 310 conducts interviews with the selected person 320 displayed in the matching results and inputs their evaluation of the matching results to the support server 100 via the information processing terminal 200 (S405).

[0065] Next, the AI ​​processing unit 206 records the evaluation input by the selector 310 in step S405 in the selector DB 211, and dynamically updates the threshold consisting of the selector 310's score or filter (search condition) previously recorded in the selector DB 211 (S406).

[0066] [Rejection Factors / Tolerance Factor Processing] Next, the details of the "rejection factor / tolerance factor processing" in step S402 will be explained using the flowchart in Figure 5. First, the question presentation unit 202 presents the selector 310 with a question containing pairs of negative short sentences corresponding to multiple evaluation factors (S501). The display screen of the information processing terminal 200 held by the selector 310 displays the question screen G1 exemplified in Figure 7. On the question screen G1, the selector 310 selects the question they consider acceptable and presses "Confirm Answer," and the selection result is obtained (S502).

[0067] Based on the selection results in step S502, points are added to the selected factors and deducted from the unselected factors (S503). In step S503, evaluation scores are added and subtracted for each question based on core and peripheral weights. This addition and subtraction is used to equalize the evaluation scores so that they approach an average of 0.

[0068] Since the processing in step S503 requires that all evaluation factors be asked an equal number of questions, it is determined in step S504 whether all questions have been completed based on the questions presented in step S502. Steps S502 and S503 are executed until all questions have been completed (S504: NO).

[0069] When evaluation factors for all questions are obtained and the score processing in step S503 is performed (S504: YES), one or two evaluation scores with the lowest values ​​are extracted based on the averaged evaluation scores and designated as "rejection factors". In addition, evaluation factors with evaluation scores above a predetermined threshold are designated as "tolerance factors" (S505). Figure 6 shows the processing image for step S503.

[0070] The extracted "rejection factors" and "tolerance factors" are associated with the selector 310 and recorded in the selector DB 211 (S506).

[0071] Based on the rejection factor / tolerance factor processing described above, the selected person search unit 205 configures a lower limit filter and executes the selected person search process described in step S404, based on the rejection and tolerance factors of the selected person 310 identified through this process. In this process, the selected person stored in the selected person DB 212 is narrowed down by combining the lower limit filter based on rejection factors with the upper limit filter or relaxation conditions based on tolerance factors.

[0072] For example, if the rejection factor is "lack of integrity," then candidates whose "integrity" score in the factor measurement results falls below a predetermined threshold (e.g., the bottom 20%) will be excluded from the search.

[0073] On the other hand, tolerance factors are used to construct upper limit filters or relaxation conditions. For example, if the tolerance factor is "low extraversion," candidates with low extraversion scores are not excluded, or candidates are displayed even if other search criteria (e.g., desired annual salary) do not quite match, thus relaxing the filtering conditions. This ensures that the line that selectors absolutely cannot tolerate is maintained, while preventing missed opportunities for traits with a wide range of tolerance.

[0074] Furthermore, the selected person search unit 205 may also use semantic vector similarity search in addition to numerical filtering based on factor scores. In this case, step S404 is performed using semantic vector similarity search in addition to numerical filtering of factor scores.

[0075] For example, the self-introduction text entered by the selected candidate and the text describing the "ideal candidate profile" entered by the selector are converted into semantic vectors using natural language processing technology. Then, the cosine similarity between the selected candidate's semantic vector and the selector's semantic vector is calculated, and candidates with high similarity are displayed higher in the search results. This enables more nuanced matching that cannot be captured by numerical factor scores alone.

[0076] Thus, the narrowed-down list of selected individuals 320 is presented to the information processing terminal 200 held by the selector 310. At this time, the selected individual search unit 205 may visualize the search results not only as a list but also with factor names. At this time, warnings or highlight labels may also be displayed simultaneously.

[0077] For example, the results display screen G2 shown in Figure 8 is an example of a user interface (UI) according to this embodiment. In this UI, next to each candidate's profile, a warning label such as "!Caution: Concerns about planning ability" is displayed for candidates with characteristics that may violate rejection factors, and a highlight label such as "★Highly suitable: Emphasizes cooperativeness" is displayed for candidates who are a good match for tolerance factors. This allows the selector to intuitively grasp the characteristics of each candidate and efficiently decide whether or not to proceed to the next step, such as an interview.

[0078] The selector 310 conducts in-person interviews or online conversations with selected candidates 320 who are of interest from the presented list. After a match is made or after the conversation, the selector 310 inputs their evaluation of the selected candidate 320 into the support server 100.

[0079] [Rejection / Tolerance Factor Extraction Process] Next, the extraction process of rejection / tolerance factors (step S402 in Figure 4), which is the core part of the present invention, will be explained in detail using the flowchart in Figure 5.

[0080] First, the question presentation unit 202 presents a question to the selector as the step corresponding to claim 1(a) and as the selection acquisition unit of claim 11 (S501). When the selector 310 makes a selection, the score processing unit 203 acquires the selection result (S502).

[0081] In steps S501 and S502, participants are presented with questions containing pairs of negative short sentences corresponding to multiple evaluation factors, and their selection is obtained for each question regarding which of the two is more acceptable.

[0082] Next, as a score processing step, based on the selection results obtained in step S502, the system adds points to the evaluation of the selected side and subtracts points from the evaluation of the unselected side. Subsequently, the score processing unit 203 asks all factors an equal number of questions and performs a process to equalize the scores so that they approach the mean of 0 by adding or subtracting scores based on the core and peripheral weights for each question (S503).

[0083] Specifically, the combination of questions is controlled so that all evaluation factors are presented as pairs the same number of times throughout the entire set of questions, and the total number of questions is always an even number to correct for statistical bias. In other words, in the score leveling step performed in step S502, the number of questions is controlled so that each factor is presented the same number of times, and the number of questions is even, thereby correcting for statistical deviation.

[0084] Furthermore, for each factor, a "core question" that asks about the core characteristics of that factor and a "peripheral question" that asks about peripheral characteristics may be prepared, and scoring may be performed by assigning different weights to each (e.g., ±2 for the score variation of the core question and ±1 for the peripheral question). This will enable the extraction of rejection tendencies with greater accuracy.

[0085] The control unit 110 determines whether all of the specified number of questions have been completed (S504). If all questions have been completed (S504: YES), the factor extraction unit 204 selects one or two evaluation factors with the lowest values ​​based on the evaluation score as "rejection factors". Furthermore, the factor extraction unit 204 selects evaluation factors with values ​​above a predetermined threshold as "tolerance factors" (S505).

[0086] Specifically, based on the final scores of each factor, the factor with the lowest final score (up to two in case of a tie) is extracted as the "rejection factor." On the other hand, factors with statistically significantly higher scores, such as those with a score of +1σ (standard deviation) or higher (up to two in case of a tie), are extracted as the "tolerance factor." In addition, if the score distribution is uniform, all scores are either positive or negative, or there are three or more factors with the same score, and no particular trend is observed, it can be determined as "no preference" (neither rejection nor tolerance).

[0087] The extracted rejection and tolerance factors are stored in the selector DB211 in association with the selector's identification information (step S506).

[0088] [Differentiation] Although embodiments of the present invention have been described above, the present invention is not limited to the above embodiments. For example, the following modifications are possible.

[0089] The questions presented by the question presentation unit 202 are not limited to a pre-prepared, fixed list. The AI ​​processing unit 206 may dynamically generate questions that include negative short sentences. For example, even for the same evaluation factor of "integrity," the AI ​​can generate and use diverse expressions such as "does not keep promises," "shifts blame," and "does not take responsibility for their words," making it more difficult for the respondent 320 to discern the intention and allowing for the extraction of more essential tendencies. Furthermore, it is also possible to include multiple evaluation factors in a single question, or to present questions in the form of specific contextual expressions or stories, such as "In a situation where a team project is delayed, which of the following member actions would you find most unacceptable?" and ask the respondent to select the most unacceptable action.

[0090] Furthermore, matching system 1 can be used in conjunction with a "positive preference test" in which selectors 310 actively choose elements they like, in addition to negative preferences based on rejection factors. In this case, matching can be performed along four axes: "rejection factors," "tolerance factors," "desired factors," and "undesired factors," resulting in a more multifaceted and accurate matching process.

[0091] Furthermore, the factor extraction unit 204 may also use a method for extracting factors that does not rely on a binary question format. For example, the selector 310 is presented with profiles of numerous candidates 320 who have been interviewed in the past, and is asked to classify them into three categories: "Want to hire," "Want to consider," and "Will not hire." The classification results (labels) and the factor scores of each candidate 320 are then used as training data to construct a machine learning model (e.g., a decision tree or a support vector machine). By analyzing this model, it is also possible to employ a method that extracts rejection factors and tolerance factors inversely from the selector 310's classification tendencies. This method is particularly effective for selectors 310 who do not have time to answer questions.

[0092] [Examples of applying the present invention to matching services] The following describes a specific example of how the matching system 1 according to this embodiment can be used when applied to a job-hunting support application for smartphones.

[0093] Corporate HR personnel (310 participants) register for this application and first take a test to diagnose what characteristics of students are unacceptable in light of their company's hiring policies. The screen displays a series of questions such as, "If you had to choose between a student who 'waits for instructions' and a student who 'acts independently,' which would you hire?" and "If you had to choose between a student with 'low stress tolerance' and a student who 'dislikes trying new things,' which would you hire?"

[0094] Once the person in charge, designated as Selector 310, has finished answering dozens of questions, Matching System 1 compiles the evaluation scores and presents an analysis result such as, "Diagnosis: The trait your company finds most unacceptable (rejection factor) is 'lack of integrity.' On the other hand, a trait that is relatively tolerant (tolerance factor) is 'low extroversion.'"

[0095] On the other hand, students currently job hunting (320 selected individuals) take the Big Five personality assessment when registering for the application. Their scores for each of the "extraversion," "agreeableness," "conscientiousness," "neuroticism," and "openness" factors are calculated and registered in the selected individuals database (212) along with their profile information.

[0096] When HR personnel search for students, matching system 1 applies these rejection and acceptance factors as filters in the background. Specifically, students whose "conscientiousness" score falls below a certain threshold in the Big Five personality assessment are automatically excluded from the search results, even if they have high academic qualifications and skills. On the other hand, introverted students with low "extraversion" scores are actively displayed as candidates if they meet other criteria.

[0097] This allows HR personnel, acting as selectors (310), to avoid the hassle of reviewing student profiles that fundamentally don't align with their company's culture and values, allowing them to focus on more meaningful matching. It significantly reduces the inefficiencies of traditional methods, such as searching using vague and idealized criteria like "students with strong communication skills," and then discovering mismatches after interviewing numerous candidates.

[0098] Furthermore, after a company interviews multiple students, it inputs each student's evaluation (e.g., a 5-point rating and comments) into the application. The AI ​​processing unit 206 analyzes this feedback data and learns patterns such as, "This company was tolerant of 'low extroversion,' but in reality, students with high interview evaluations tend to be highly extroverted." Then, in subsequent matchings, it automatically fine-tunes the threshold for this company's tolerance regarding "extroversion." Through this dynamic retraining, the system learns potential hiring criteria that even HR personnel themselves may not be aware of, and can continuously improve matching accuracy.

[0099] As described above, the matching support program and matching support system according to the present invention enable practical and realistic matching by narrowing down the candidates 320 based on the "unacceptable characteristics (rejection factors)" and "factors that do not matter whether the characteristics are high or low (tolerance factors)" of the selector 310, in relation to the technology of person matching.

[0100] According to the present invention, by clearly identifying the unacceptable "rejection factors" rather than the desired conditions of the selector 310, it becomes possible to suppress ideal inflation and efficiently extract more realistic and less mismatched candidates 320.

[0101] Furthermore, compared to matching based solely on objective conditions as in the prior art, the present invention has the superior effect of preventing internal mismatches such as personality and values ​​in advance. Moreover, compared to profiling techniques that pre-classify the personalities of the selected individuals 320, it has the effect of enabling a more satisfying match that better matches the deeper needs of the selector 310 by clearly defining the "non-negotiable line" on the selector's side 310.

[0102] It should be noted that the present invention is not limited to the embodiments exemplified above. Therefore, the present invention can be modified by adding or changing components without departing from the technical spirit. Thus, all technical matters included in the technical concept described in the claims are covered by the present invention. The embodiments exemplified above are specific examples that are suitable for implementation. Furthermore, those skilled in the art can implement various modifications from the disclosed content, and such modifications are included in the technical scope described in the claims.

[0103] [Aspects of the present invention] The contents of this invention are, for example, as follows: <1> A matching support program that causes a computer to perform information processing to assist in matching selectors with selected individuals, A question presentation step in which a pair of questions including negative expressions corresponding to the evaluation factors used in the matching are presented to the selector, The evaluation score addition / subtraction step involves adding points to the evaluation score for the evaluation factors related to the questions selected by the selected person from the presented questions, and subtracting points for the evaluation factors related to the questions not selected by the selected person. A score equalization step in which the number of times questions related to all evaluation factors used in the matching are presented is made equal so that the average of the evaluation scores for multiple evaluation factors is zero, thereby equalizing the evaluation scores, A step of extracting evaluation factors from the standardized evaluation factors, wherein evaluation factors whose evaluation score is lower than a predetermined threshold are selected as rejection factors, and evaluation factors whose evaluation score is equal to or greater than a predetermined threshold are selected as tolerance factors, A matching determination step in which the matching of the selected person to the selector is determined based on the rejection factor and the tolerance factor, This is a matching support program characterized by including [specific features]. <2> The aforementioned evaluation factors are defined by the Big Five personality traits, HEXACO, or by statistical comparisons between high-performing and low-performing groups within an organization. The aforementioned <1> This is the matching support program described in [the relevant section]. <3> In the step of presenting the question, The aforementioned question includes a pair of negative short sentences corresponding to multiple evaluation factors, The short text is dynamically generated using a machine learning model and is in the form of a contextual expression or story that includes multiple evaluation factors. The aforementioned <1> or <2> This is the matching support program described in [the relevant section]. <4> In the score leveling step, The number of times the above question is presented is even. The aforementioned <1> or the above <3> This is a matching support program listed in one of the following categories. <5> In the matching determination step, At a minimum, the matching is determined by a combination of a lower limit filter related to the rejection factor and an upper limit filter related to the tolerance factor. The aforementioned <1> or the above <4> This is a matching support program listed in one of the following categories. <6> In the matching determination step, Further numerical filtering of the aforementioned evaluation score is used to determine the matching based on the similarity using semantic vectors. The aforementioned <1> or the above <5> This is a matching support program listed in one of the following categories. <7> Based on the matching determination result, a threshold update step is further executed to update the threshold according to the interview results between the selector and the selected person. The aforementioned <1> or the above <6> This is a matching support program listed in one of the following categories. <8> In the threshold update step, Based on the feedback score for the selected person evaluated by the selector, the thresholds for the rejection factor and the tolerance factor are relearned. The aforementioned <7> This is the matching support program described in [the relevant section]. <9> In the step of presenting the question, Along with the aforementioned question for the selector, information indicating the evaluation factors related to the question is output in a format that the selector can recognize. The aforementioned <1> or the above <8> This is a matching support program listed in one of the following categories. <10> In the evaluation factor extraction step, Rather than extracting the rejection factor and tolerance factor using the threshold, Regardless of the aforementioned questions, the rejection factors and tolerance factors are extracted in reverse from the tendencies of multiple previously classified individuals. The aforementioned <1> or the above <9> This is a matching support program listed in one of the following categories. <11> A matching support system that performs processing to assist in matching selectors and selected individuals, A question presentation means that presents the selector with a pair of questions containing negative expressions corresponding to the evaluation factors used in the matching, A means for adding or subtracting evaluation scores, which adds points to the evaluation score for the evaluation factor corresponding to the question selected by the selector from the presented questions, and subtracts points for the evaluation score for the evaluation factor corresponding to the question not selected by the selector, A score leveling means that equalizes the number of times the questions corresponding to all the evaluation factors used in the matching are presented, so that the average of the evaluation scores for multiple evaluation factors becomes zero, and levels out the evaluation scores, An evaluation factor extraction means that extracts evaluation factors from the standardized evaluation factors whose evaluation score is lower than a predetermined threshold as rejection factors, and evaluation factors whose evaluation score is equal to or greater than a predetermined threshold as tolerance factors, A matching determination means that determines the matching of the selected person to the selector based on the rejection factor and the tolerance factor, This is a matching support system characterized by having [a certain feature]. [Explanation of symbols]

[0104] 1: Matching System 100: Support Server 110: Control Unit 200: Information processing terminal 201:Factor acquisition part 202:Question presentation part 203: Score Processing Unit 204:Factor extraction part 205: Selected Person Search Section 206: AI Processing Unit 310: Selector 320: Selected person DB211: Selector DB212: Selected person DB213: Question

Claims

1. A matching support program that causes a computer to perform information processing to assist in matching selectors with selected individuals, A question presentation step in which a pair of questions consisting of negative expressions corresponding to the evaluation factors used in the matching are presented to the selector, The evaluation score addition / subtraction step involves adding points to the evaluation score for the evaluation factors related to the questions selected by the selected person from the presented questions, and subtracting points for the evaluation factors related to the questions not selected by the selected person. A score equalization step in which the number of times questions related to all evaluation factors used in the matching are presented is made equal so that the average of the evaluation scores for multiple evaluation factors is zero, thereby equalizing the evaluation scores, A step of extracting evaluation factors from the standardized evaluation factors, wherein evaluation factors whose evaluation score is lower than a predetermined threshold are selected as rejection factors, and evaluation factors whose evaluation score is equal to or greater than a predetermined threshold are selected as tolerance factors, A matching determination step in which the matching of the selected person to the selector is determined based on the rejection factor and the tolerance factor, A matching support program characterized by including the following.

2. The aforementioned evaluation factors are factors defined by the Big Five personality traits, HEXACO, or by statistical comparisons between high-performing and low-performing groups within an organization. The matching support program according to claim 1.

3. In the step of presenting the question, The aforementioned question consists of a pair of negative contextual expressions or short sentences in story format that correspond to multiple evaluation factors. The short sentence is dynamically generated using a machine learning model. The matching support program according to claim 1 or 2.

4. In the score leveling step, The number of times the above question is presented is even. The matching support program according to claim 1 or 2.

5. In the matching determination step, At a minimum, the matching is determined by a combination of a lower limit filter related to the rejection factor and an upper limit filter related to the tolerance factor. The matching support program according to claim 1 or 2.

6. In the matching determination step, Further numerical filtering of the aforementioned evaluation score is used to determine the matching based on the similarity using semantic vectors. The matching support program according to claim 1 or 2.

7. Based on the matching determination result, a threshold update step is further executed to update the threshold according to the interview results between the selector and the selected person. The matching support program according to claim 1 or 2.

8. In the threshold update step, Based on the feedback score for the selected person evaluated by the selector, the thresholds for the rejection factor and the tolerance factor are relearned. The matching support program according to claim 7.

9. In the step of presenting the question, Along with the aforementioned question for the selector, information indicating the evaluation factors related to the question is output in a format that the selector can recognize. The matching support program according to claim 1 or 2.

10. In the evaluation factor extraction step, Rather than extracting the rejection factor and tolerance factor using the threshold, Regardless of the aforementioned questions, the rejection factors and tolerance factors are extracted in reverse from the tendencies of multiple previously classified individuals. The matching support program according to claim 1 or 2.

11. A matching support system that performs processing to assist in matching selectors and selected individuals, A question presentation means for presenting the selector with a pair of questions consisting of negative expressions corresponding to the evaluation factors used in the matching, A means for adding or subtracting evaluation scores, which adds points to the evaluation score for the evaluation factor corresponding to the question selected by the selector from the presented questions, and subtracts points for the evaluation score for the evaluation factor corresponding to the question not selected by the selector, A score leveling means that equalizes the number of times the questions corresponding to all the evaluation factors used in the matching are presented, so that the average of the evaluation scores for multiple evaluation factors becomes zero, and levels out the evaluation scores, An evaluation factor extraction means that extracts evaluation factors from the standardized evaluation factors whose evaluation score is lower than a predetermined threshold as rejection factors, and evaluation factors whose evaluation score is equal to or greater than a predetermined threshold as tolerance factors, A matching determination means that determines the matching of the selected person to the selector based on the rejection factor and the tolerance factor, A matching support system characterized by having the following features.

Images