Occupational evaluation method based on large model

Abstract

The invention is suitable for the technical field of occupational evaluation, and particularly relates to an occupational evaluation method based on a large model, and the method comprises the steps: S100, obtaining the self-evaluation filled by an evaluation object for an occupational evaluation questionnaire, extracting questions in the occupational evaluation questionnaire, obtaining the chat record of the evaluation object based on a preset grabbing authority, and obtaining the chat record of the evaluation object; and inputting the chat record into a pre-constructed semantic understanding model, outputting and obtaining a fragment with the same semantics as the question in the chat record, and defining the fragment as a clear fragment. According to the method, the answer of the scene question can be determined by determining the clear fragment, meanwhile, the evaluation flexibility is enhanced, the self-recognition of the evaluation object is promoted, the evaluation process is simplified, the answer of the associated question can be determined by determining the target fragment, the evaluation accuracy is further improved, the deviation in self-evaluation can be compared by determining the distinguishing item, and the evaluation accuracy is improved. The self-recognition of the evaluation object can be improved, and the reliability of the evaluation result is enhanced.

Description

Technical Field

[0001] The present invention relates to the technical field of occupational assessment, and in particular to an occupational assessment method based on a large model. Background Art

[0002] Career assessment is an important means to help the assessed person understand his or her career inclinations, interests, abilities and values. Career assessment is usually conducted through questionnaires and tests. By analyzing the assessment results, the assessed person can identify a career direction that suits him or her. At the same time, companies can also provide personalized career development suggestions to employees or customers based on the assessment results to optimize talent allocation.

[0003] However, since career assessment needs to be conducted through questionnaires and tests, there are relatively large subjective factors. When filling out questionnaires or tests, the assessment subjects are easily affected by personal emotions and environmental factors, which may cause the results to deviate from the actual career orientation or ability level. Therefore, "how to use the chat records between the assessment subjects and the evaluators to correct the assessment results" is the technical problem that the present invention needs to solve. Summary of the invention

[0004] The purpose of the present invention is to provide a career assessment method based on a large model to solve the problem of "how to use the chat records between the assessment object and the assessor to correct the assessment results" raised in the above background technology.

[0005] To achieve the above object, the present invention provides the following technical solutions: A career assessment method based on a large model, the method comprising: S100: Obtain the self-evaluation filled in by the assessment subject on the career assessment questionnaire, extract the questions in the career assessment questionnaire, obtain the chat records of the assessment subject based on the preset crawling authority, and input the chat records into a pre-built semantic understanding model, output the segments in the chat records with the same semantics as the questions, define them as explicit segments, divide the questions into scenario questions and related questions, and traverse the answers to the scenario questions from the explicit segments; S200: creating a queue corresponding to the related questions one by one, inputting the chat records and the related questions into the semantic understanding model, outputting a segment with similar semantics to the related questions, defining it as a similar segment and determining the similarity of the similar segment, placing the similar segment into a queue, and arranging them in descending order of similarity, judging whether the similarity of the similar segment ranked first in the queue is greater than a preset threshold; if so, defining the similar segment ranked first as a target segment, and determining the answer to the related question using the target segment; if not, finding the evaluator corresponding to the target segment, sending the related question to the evaluator, and receiving the social evaluation uploaded by the evaluator; S300: Synchronize the answers and social evaluations to corresponding queues, integrate all queues and answers to situational questions to obtain social evaluations, compare the self-evaluations and social evaluations, determine the difference items, and push the difference items to a preset terminal.

[0006] Furthermore, the S100 includes: Obtaining behavioral data of the evaluation subject, wherein the behavioral data at least includes: historical evaluation results and work experience; The behavioral data were normalized and the self-assessments were corrected.

[0007] Furthermore, the S100 further includes: Obtaining a friend list of the evaluation subject, and dividing the friend list into several levels; The time and scope of the crawling authority are configured, and a corresponding relationship between the crawling authority and different levels is established.

[0008] Furthermore, the S100 further includes: Numbering the questions, and inputting the questions into the semantic understanding model in sequence according to the numbers, outputting segments with the same semantics as the questions, and synchronizing the numbers; The outputted segment is determined as a clear segment, and the mapping between the title, number and clear segment is edited.

[0009] Further, the S200 includes: Creating a semantic embedding model and linking it with the semantic understanding model, inputting the chat records and related questions into the semantic understanding model, and outputting similar segments, wherein one related question corresponds to a plurality of similar segments; Using the semantic embedding model, similar segments with strong sentiment values ​​are removed.

[0010] Furthermore, the S200 further includes: Determining the confidence value of the self-assessment, and dividing the confidence value into the related questions; Based on the confidence value, the threshold is deviated.

[0011] Furthermore, the S200 further includes: A multi-friend evaluation mechanism is embedded in the queue, and when the similarity of the target segment is less than a set value, a preset number of similar segments are selected from the front of the queue; Based on the similar segments, a multi-friend dialogue is established, a group evaluation of the related questions is extracted, and the social evaluation is adjusted using the group evaluation.

[0012] Further, the S300 includes: Using the existing personality analysis model, the chat records are integrated to determine the personality label of the subject of evaluation, and the label is inserted into the distinguishing item; A preset item comparison table is queried to determine the evaluation suggestion, wherein the item comparison table is composed of personality label items and evaluation suggestion items.

[0013] Furthermore, the S300 further includes: The chat record or target segment corresponding to the distinguishing item is found, a reference is generated, and the reference is sent to a preset terminal.

[0014] Furthermore, the method further comprises: receiving feedback from the evaluation subject on the distinguishing items, correcting the self-evaluation, and generating a career evaluation form; Construct a career assessment model, input the career assessment form into the career assessment model, identify career preferences, and configure career development strategies.

[0015] Compared with the prior art, the present invention has the following beneficial effects: By splitting the questions, the evaluation errors can be greatly reduced and the accuracy of social evaluation can be improved. By determining clear segments, the answers to situational questions can be determined, the flexibility of the evaluation can be enhanced, the self-awareness of the evaluation subjects can be promoted, and the evaluation process can be simplified. By determining the target segments, the answers to related questions can be determined, further improving the accuracy of the evaluation. By determining the distinguishing items, the deviations in self-evaluation can be compared, which greatly improves the self-awareness of the evaluation subjects and enhances the reliability of the evaluation results. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 A flowchart of a career assessment method based on a large model provided by an embodiment of the present invention; Figure 2 A first sub-flow chart of a career assessment method based on a large model provided by an embodiment of the present invention; Figure 3 A second sub-flow chart of the career assessment method based on a large model provided by an embodiment of the present invention; Figure 4 A third sub-process flowchart of the large model-based career assessment method provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0017] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0018] In Example 1, Figure 1 The implementation process of the career assessment method based on a large model provided by an embodiment of the present invention is shown and described in detail below: S100: Obtain the self-evaluation filled out by the assessment subject on the career assessment questionnaire, extract the questions in the career assessment questionnaire, obtain the chat history of the assessment subject based on the preset crawling authority, and input the chat history into a pre-built semantic understanding model, and output the fragments in the chat history with the same semantics as the questions, which are defined as explicit fragments, and divide the questions into situational questions and related questions, and traverse the answers to the situational questions from the explicit fragments.

[0019] Before the implementation of this application, it is necessary to organize the test takers to chat with multiple evaluators. The content of the chat has no clear direction, and the chat content is recorded with the permission of both parties.

[0020] Through testing, the test results of the subject's career assessment questionnaire are determined, and the test results are determined as self-evaluation; the questions in the career assessment questionnaire are extracted, the capture permission of the subject's chat history is obtained, and the captured chat history is input into the semantic understanding model, wherein the semantic understanding model is a natural language processing model in the prior art, which is mainly used to calculate the similarity of two text fragments. The specific calculation method may be cosine similarity, Euclidean distance, etc.

[0021] The semantic understanding model compares the question with the chat record, outputs the fragments in the chat record with the same semantics as the question, and determines the obtained fragments as explicit fragments, determines the question corresponding to the explicit fragment as a situational question, and determines the part of the question other than the situational question as a related question; and uses the explicit fragments to determine the answer to the situational question.

[0022] S200: Create a queue corresponding to the related questions one by one, input the chat records and related questions into the semantic understanding model, output the fragments with similar semantics to the related questions, define them as similar fragments and determine the similarity of the similar fragments, put the similar fragments into the queue, and arrange them in order of similarity from large to small, and judge whether the similarity of the similar fragment ranked first in the queue is greater than a preset threshold; if so, define the similar fragment ranked first as the target fragment, and use the target fragment to determine the answer to the related question; if not, find out the evaluator corresponding to the target fragment, send the related question to the evaluator, and receive the social evaluation uploaded by the evaluator.

[0023] A queue is created for each related question, and the chat records and related questions are input into the semantic understanding model again. The output is the fragments with similar semantics to the related questions, namely the similar fragments, and the similarity between each similar fragment and the related question is output at the same time. The greater the similarity, the higher the consistency between the question and the similar fragment in terms of semantic themes, keywords and structures, and the lower the difficulty of using similar fragments to determine the answer to the related question. According to the similar fragments, the arrangement order of the similar fragments in the queue is adjusted in descending order of similarity.

[0024] If the similarity of the first-ranked similar segment is greater than a preset threshold, the first-ranked similar segment is defined as a target segment, and the target segment and the related question are input into a pre-built machine learning model to output the answer to the related question.

[0025] If the similarity of the first similar segment is less than or equal to a preset threshold, the evaluator corresponding to the target segment is found, and the associated question is sent to the evaluator for the evaluator to perform social evaluation on the associated question.

[0026] For example, a certain career assessment test paper is divided into 10 questions, of which four questions numbered 2, 3, 7 and 9 are related questions. Taking the related question numbered 2 as an example, find out the fragments in the chat record whose similarity is greater than a threshold, that is, similar fragments. It should be noted that there may be one or more similar fragments; sort the similar fragments in order of similarity from large to small, and form a queue, and determine the similar fragment ranked first as the target fragment. If the similarity of the target fragment is greater than the threshold, it means that the answer to related question 2 can be determined based on this target fragment; conversely, if the target fragment is less than or equal to the threshold, it means that the answer to related question 2 cannot be determined based on the chat record at this time, then it is necessary to determine the evaluator of the target fragment in the chat record, send related question 2 to the evaluator, and determine the social evaluation.

[0027] S300: Synchronize the answers and social evaluations to corresponding queues, integrate all queues and answers to situational questions to obtain social evaluations, compare the self-evaluations and social evaluations, determine the difference items, and push the difference items to a preset terminal.

[0028] Determine the answers or social evaluations of all related questions, integrate the answers to the situational questions, and compare them with the self-evaluation of the subject to find out the questions with differences between the self-evaluation and the social evaluation, that is, the distinguishing items; the distinguishing items also mean that there is a difference between the self-evaluation of the subject on the corresponding question in the career assessment questionnaire and the social evaluation given by the evaluator; send the distinguishing items to the preset terminal, and re-determine the answers to the distinguishing items, wherein the preset terminal can be the terminal of the subject.

[0029] In Example 2, Figure 2 The implementation process of the career assessment method based on a large model provided by an embodiment of the present invention is shown, and S100 is described in detail below, as follows: S101: Obtaining behavioral data of the evaluation subject, wherein the behavioral data at least includes: historical evaluation results and work experience.

[0030] In addition to determining the self-evaluation based on the subjects' completed questionnaires, the self-evaluation should also be revised based on the subjects' behavioral data. This is because there is a high degree of randomness in the completion of career assessment questionnaires, and revising the self-evaluation can greatly improve the accuracy of the assessment results.

[0031] S102: Standardizing the behavior data and correcting the self-evaluation.

[0032] Before using behavioral data to correct self-evaluation, behavioral data from multiple sources need to be standardized. The specific standardization process should be determined based on the specific source.

[0033] In Example 3, Figure 2 The implementation process of the career assessment method based on a large model provided by an embodiment of the present invention is shown, and S100 is further described in detail below, as follows: S103: Editing the friend list of the evaluation object, and dividing the friend list into several levels.

[0034] The friend list here refers to: a collection of evaluators who chat with the evaluation object, and the friend list of the evaluation object is divided into several levels.

[0035] S104: configuring the time and scope of the crawling authority, and establishing a corresponding relationship between the crawling authority and different levels.

[0036] Determine the time and scope of the crawling permission for each level; for example, only the chat records within a fixed time period within a certain level in the friend list can be crawled.

[0037] In Example 4, Figure 2 The implementation process of the career assessment method based on a large model provided by an embodiment of the present invention is shown, and S100 is further described in detail below, as follows: S105: Numbering the questions, and inputting the questions into the semantic understanding model in sequence according to the numbers, outputting segments with the same semantics as the questions, and synchronizing the numbers.

[0038] The questions are numbered and input into the semantic understanding model in the order of the numbers. Each question is compared with the chat record in turn, and the fragments with the same semantics as the question are output. At the same time, the number of each fragment is determined, so as to establish the correspondence between the fragment and the question.

[0039] S106: Determine the output segment as a clear segment, and edit the mapping between the title, number and clear segment.

[0040] In the above process, the segments output by the semantic understanding model are identified as clear segments; the number of each question and the number of each clear segment are recorded.

[0041] In Example 5, Figure 3 The implementation process of the career assessment method based on a large model provided by an embodiment of the present invention is shown, and S200 is described in detail below, as follows: S201: Create a semantic embedding model and link it with the semantic understanding model, input the chat records and related questions into the semantic understanding model, and output similar segments, where one related question corresponds to several similar segments.

[0042] Using the semantic understanding model, find the segments with similar semantics to the related questions in the chat records and determine them as similar segments. According to common sense, one related question may correspond to one similar segment or multiple similar segments.

[0043] S202: Utilizing the semantic embedding model, deleting similar segments with strong sentiment values.

[0044] A semantic embedding model is established, and the sentiment value of each similar fragment is calculated using the semantic embedding model. If the sentiment value is greater than a preset threshold, it is called a strong sentiment value. The step of calculating the sentiment value of similar fragments at least includes: querying a preset sentiment dictionary to determine the sentiment value of a single word in the similar fragment, and performing weighted summation to determine the total sentiment value of all words; then using the semantic embedding model to determine the sentiment value of the sentence in the similar fragment, superimposing the sentiment value of the similar fragment, and deleting the similar fragments with strong sentiment values ​​from the queue.

[0045] In Example 6, Figure 3 The implementation process of the career assessment method based on a large model provided by an embodiment of the present invention is shown, and S200 is further described in detail below, as follows: S203: Determine the confidence value of the self-assessment, and divide the confidence value into the related questions.

[0046] During the self-evaluation process of the test subject, the test subject's confidence value for each question is determined, where the confidence value refers to the degree of confidence the test subject has about each question. The higher the confidence value, the more certain the test subject is about the answer to the corresponding related question.

[0047] S204: Deviate from the threshold based on the confidence value.

[0048] Based on the confidence value, the answers to the corresponding related questions are deviated; for example, if the confidence value of the test subject for a certain related question is high, the threshold of the corresponding queue should be lowered.

[0049] In Example 7, Figure 3 The implementation process of the career assessment method based on a large model provided by an embodiment of the present invention is shown, and S200 is further described in detail below, as follows: S205: embedding a multi-friend evaluation mechanism into the queue, and when the similarity of the target segment is less than a set value, selecting a preset number of similar segments from the front row of the queue.

[0050] A multi-friend evaluation mechanism is embedded in the queue, where if the similarity of the target fragment is less than a set value, it will be evaluated by multiple evaluators; the set value is pre-determined by professionals, then the selection range is expanded, and multiple similar fragments are selected from the front of the queue.

[0051] S206: Based on the similar segments, a multi-friend dialogue is established, a group evaluation of the related questions is extracted, and the social evaluation is adjusted using the group evaluation.

[0052] The evaluators corresponding to multiple similar fragments are traced back, and a multi-friend dialogue is established. The corresponding related questions are sent to the multi-friend dialogue, the collected feedback is determined as a group evaluation, and the social evaluation is adjusted.

[0053] In actual use, if the related questions are sent to a certain evaluator, the subjective factors of the social evaluation will be relatively large, which will also cause the social evaluation to deviate from the true value; therefore, sending the related questions to multiple evaluators and letting the evaluators who receive the related questions conduct the evaluation can improve the authenticity of the social evaluation.

[0054] In Example 8, Figure 4 The implementation process of the career assessment method based on a large model provided by an embodiment of the present invention is shown, and S300 is described in detail below, as follows: S301: Using the existing personality analysis model, the chat records are integrated to determine the personality label of the evaluation object and insert it into the distinguishing item.

[0055] By using the personality analysis model in the prior art, the personality label of the evaluation subject is determined according to the chat records of the evaluation subject, and the self-evaluation of the evaluation subject is adjusted.

[0056] S302: Query a preset item comparison table to determine evaluation suggestions, wherein the item comparison table consists of personality label items and evaluation suggestion items.

[0057] After determining the personality label of the assessment subject, query the item comparison table, determine the assessment suggestion, and give suggestions on the assessment subject's career planning. This suggestion is also the assessment suggestion.

[0058] In Example 9, Figure 4 The implementation process of the career assessment method based on a large model provided by an embodiment of the present invention is shown, and S300 is further described in detail below, as follows: S303: Find the chat record or target segment corresponding to the distinguishing item, generate a reference, and send the reference to a preset terminal.

[0059] By comparing the self-evaluation of the subject, the social evaluation given by the evaluator and the answers to the related questions determined by the chat records, the distinguishing items can be obtained, and the chat records or social evaluations corresponding to the distinguishing items can be found and sent to the preset terminal, where the preset terminal is the terminal of the subject.

[0060] In Example 10, different from Example 1, in this embodiment of the present invention, the method further includes: receiving feedback from the evaluation subject on the distinguishing items, correcting the self-evaluation, and generating a career evaluation form; Construct a career assessment model, input the career assessment form into the career assessment model, identify misunderstandings, and configure career development strategies.

[0061] After sending the distinguishing items to the subject of the assessment, the subject will adjust his or her self-evaluation, and use the changes in the self-evaluation before and after the adjustment to generate a career assessment form, and input the career assessment form into the career assessment model to identify the parts of the subject's misunderstanding of his or her own evaluation; use the self-evaluation to generate a career development strategy, and use the misunderstanding to adjust the career development strategy.

[0062] The technical features of the above-described embodiments may be arbitrarily combined. To make the description concise, not all possible combinations of the technical features in the above-described embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0063] The above-mentioned embodiments only express several implementation methods of the present invention, and the description thereof is relatively specific and detailed, but it cannot be understood as limiting the scope of the patent of the present invention. It should be pointed out that, for ordinary technicians in this field, several variations and improvements can be made without departing from the concept of the present invention, which all belong to the protection scope of the present invention. Therefore, the protection scope of the patent of the present invention shall be subject to the attached claims.

[0064] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A career assessment method based on a large model, characterized in that: The method comprises: S100: Obtain the self-evaluation filled in by the assessment subject on the career assessment questionnaire, extract the questions in the career assessment questionnaire, obtain the chat records of the assessment subject based on the preset crawling authority, and input the chat records into a pre-built semantic understanding model, output the segments in the chat records with the same semantics as the questions, define them as explicit segments, divide the questions into scenario questions and related questions, and traverse the answers to the scenario questions from the explicit segments; S200: creating a queue corresponding to the related questions one by one, inputting the chat records and the related questions into the semantic understanding model, outputting a segment with similar semantics to the related questions, defining it as a similar segment and determining the similarity of the similar segment, placing the similar segment into a queue, and arranging them in descending order of similarity, judging whether the similarity of the similar segment ranked first in the queue is greater than a preset threshold; if so, defining the similar segment ranked first as a target segment, and determining the answer to the related question using the target segment; if not, finding the evaluator corresponding to the target segment, sending the related question to the evaluator, and receiving the social evaluation uploaded by the evaluator; S300: Synchronize the answers and social evaluations to corresponding queues, integrate all queues and answers to situational questions to obtain social evaluations, compare the self-evaluations and social evaluations, determine the difference items, and push the difference items to a preset terminal.

2. The career assessment method based on a large model according to claim 1 is characterized in that: The S100 includes: Obtaining behavioral data of the evaluation subject, wherein the behavioral data at least includes: historical evaluation results and work experience; The behavioral data were normalized and the self-assessments were corrected.

3. The career assessment method based on a large model according to claim 1 is characterized in that: The S100 further includes: Obtaining a friend list of the evaluation subject, and dividing the friend list into several levels; The time and scope of the crawling authority are configured, and a corresponding relationship between the crawling authority and different levels is established.

4. The career assessment method based on a large model according to claim 1 is characterized in that: The S100 further includes: Numbering the questions, and inputting the questions into the semantic understanding model in sequence according to the numbers, outputting segments with the same semantics as the questions, and synchronizing the numbers; The outputted segment is determined as a clear segment, and the mapping between the title, number and clear segment is edited.

5. The career assessment method based on a large model according to claim 1 is characterized in that: The S200 includes: Creating a semantic embedding model and linking it with the semantic understanding model, inputting the chat records and related questions into the semantic understanding model, and outputting similar segments, wherein one related question corresponds to a plurality of similar segments; Using the semantic embedding model, similar segments with strong sentiment values ​​are removed.

6. The career assessment method based on a large model according to claim 5 is characterized in that: The S200 further includes: Determining the confidence value of the self-assessment, and dividing the confidence value into the related questions; Based on the confidence value, the threshold is deviated.

7. The career assessment method based on a large model according to claim 5 is characterized in that: The S200 further includes: A multi-friend evaluation mechanism is embedded in the queue, and when the similarity of the target segment is less than a set value, a preset number of similar segments are selected from the front of the queue; Based on the similar segments, a multi-friend dialogue is established, a group evaluation of the related questions is extracted, and the social evaluation is adjusted using the group evaluation.

8. The career assessment method based on a large model according to claim 1 is characterized in that: The S300 includes: Using the existing personality analysis model, the chat records are integrated to determine the personality label of the subject of evaluation, and the label is inserted into the distinguishing item; A preset item comparison table is queried to determine the evaluation suggestion, wherein the item comparison table is composed of personality label items and evaluation suggestion items.

9. The career assessment method based on a large model according to claim 8 is characterized in that: The S300 further includes: The chat record or target segment corresponding to the distinguishing item is found, a reference is generated, and the reference is sent to a preset terminal.

10. The career assessment method based on a large model according to claim 9 is characterized in that: The method further comprises: receiving feedback from the evaluation subject on the distinguishing items, correcting the self-evaluation, and generating a career evaluation form; Construct a career assessment model, input the career assessment form into the career assessment model, identify career preferences, and configure career development strategies.

Images