Talent evaluation management method and system based on artificial intelligence

Abstract

The present invention relates to the field of personnel management technology, and in particular to a talent evaluation management method based on artificial intelligence, which mainly realizes sentiment analysis in the talent evaluation management process by analyzing historical chat records. Specifically, the present invention maps the vocabulary in the historical chat records to the vector space by word embedding, and obtains the sentiment analysis result of the target evaluation talent according to the position distribution relationship. The preset sentiment evaluation words are artificially set words that can evaluate emotions. By analyzing the position distribution relationship between the ontology point set and the sentiment evaluation anchor point, the emotional tendency of the target evaluation talent can be identified. In addition, the present invention also analyzes the position distribution relationship between the ontology point set and the object point set to reflect the commonality of the target ontology chat record and the target object chat record, thereby obtaining a more accurate analysis and evaluation result, solving the problem of how to combine sentiment analysis technology with talent evaluation.

Description

Technical Field

[0001] The present invention relates to the technical field of personnel management, and in particular to a talent evaluation management method and system based on artificial intelligence. Background Art

[0002] Talent evaluation management is a vital part of organizational development. It not only helps companies effectively identify, attract and retain outstanding talents, but also promotes the growth and development of employees, thereby improving overall performance and competitiveness. By scientifically and comprehensively evaluating employees' skills, experience and performance, companies can better carry out human resource planning, ensure the matching degree between talents and positions, and improve team cohesion and execution.

[0003] Emotions have gradually become an important indicator in talent evaluation management, and their value cannot be underestimated. The emotional state of employees at work is crucial to improving the work efficiency of the team and the job satisfaction of employees. The emotional state and emotional experience of employees are directly related to their work performance and teamwork level. Effectively capturing, analyzing and managing employees' emotions can arrange employees to suitable positions and assign appropriate tasks, which is also very helpful in evaluating employees' abilities and performance.

[0004] Therefore, there is an urgent need for a method that can combine sentiment analysis technology and talent evaluation. Summary of the invention

[0005] Therefore, the present invention provides a talent evaluation management method and system based on artificial intelligence to solve the problem of how to combine sentiment analysis technology with talent evaluation.

[0006] The present invention provides a talent evaluation management method based on artificial intelligence, comprising:

[0007] Obtain the historical chat records of the target evaluation talent within a preset time period, wherein the historical chat records include the target entity chat records and the target object chat records;

[0008] Perform word segmentation and word embedding processing on the historical chat records to obtain a point set in the vector space corresponding to the word embedding, wherein the point set includes an ontology point set corresponding to the target ontology chat records and an object point set corresponding to the target object chat records;

[0009] Obtaining sentiment evaluation anchor points corresponding to multiple preset sentiment evaluation words in the vector space;

[0010] According to the position distribution relationship between the ontology point set and the emotional evaluation anchor point, as well as the position distribution relationship between the ontology point set and the object point set, the emotional analysis results of the target evaluation talent are obtained;

[0011] Manage the target evaluation talents based on the sentiment analysis results of the target evaluation talents.

[0012] In a further solution: the word segmentation and word embedding processing of the historical chat records to obtain a point set in the vector space corresponding to the word embedding includes:

[0013] Segment the historical chat records to obtain multiple record words, each of which has two to four characters.

[0014] Based on the preset word embedding matrix, each record word is embedded, and each record word is mapped to a point in the vector space to obtain a point set.

[0015] In a further solution: the historical chat records are segmented to obtain multiple record words, including:

[0016] Get the target chat record, which is a historical chat record to be segmented;

[0017] From the beginning to the end of the sentence, four consecutive characters in the target chat record can be overlapped and taken as the first word to be matched;

[0018] From the beginning to the end of the sentence, three consecutive characters in the target chat record can be taken in an overlapping manner as the second words to be matched;

[0019] From the beginning to the end of the sentence, two consecutive words in the target chat record can be taken in an overlapping manner as the third word to be matched;

[0020] The first to-be-matched word, the second to-be-matched word and the third to-be-matched word are matched with words in a preset dictionary, and the words that are successfully matched are used as recorded words.

[0021] In a further solution: the sentiment analysis result of the target evaluation talent is obtained according to the position distribution relationship between the ontology point set and the sentiment evaluation anchor point, and the position distribution relationship between the ontology point set and the object point set, including:

[0022] According to the position distribution relationship between the ontology point set and the emotional evaluation anchor point, the emotional analysis value of the target evaluation talent is obtained. The emotional analysis value is used to represent the evaluation level of the target evaluation talent in terms of emotional attitude;

[0023] According to the position distribution relationship between the ontology point set and the object point set, the sentiment analysis confidence of the target evaluation talent is obtained. The sentiment analysis confidence is used to indicate the authenticity of the sentiment analysis value.

[0024] In a further solution: the sentiment analysis value of the target evaluation talent is obtained according to the position distribution relationship between the ontology point set and the sentiment evaluation anchor point, including:

[0025] Count and calculate the variance of the ontology point set in each dimension in the vector space, calculate the variance of the variance of the ontology point set in multiple dimensions in the vector space, and obtain the distribution discreteness value of the ontology point set in the vector space;

[0026] According to the distribution dispersion value, the neighborhood radius value is obtained, and the neighborhood radius value is proportional to the distribution dispersion value;

[0027] Count the number of points in the ontology point set whose Euclidean distance from each sentiment evaluation anchor point is less than the neighborhood radius value, and obtain the neighborhood statistics corresponding to each sentiment evaluation anchor point;

[0028] The preset sentiment value corresponding to each sentiment evaluation anchor point is obtained, and the sentiment analysis value of the target evaluation talent is obtained according to the preset sentiment value and neighborhood statistical value corresponding to each sentiment evaluation anchor point.

[0029] In a further scheme: the preset sentiment evaluation words include a plurality of positive words and a plurality of negative words corresponding to each other, and the preset sentiment value of the sentiment evaluation anchor point corresponding to each positive word is the inverse of the preset sentiment value of the sentiment evaluation anchor point corresponding to the negative word corresponding to the positive word; the method of obtaining the preset sentiment value corresponding to each sentiment evaluation anchor point and obtaining the sentiment analysis value of the target evaluation talent according to the preset sentiment value corresponding to each sentiment evaluation anchor point and the neighborhood statistical value includes:

[0030] The sentiment analysis value of the target evaluation talent is calculated by the following formula:

[0031] Vemo=∑Vp×Nn;

[0032] Among them, Vemo is the sentiment analysis value of the target evaluation talent, Vp is the preset sentiment value corresponding to the sentiment evaluation anchor point, and Nn is the neighborhood statistical value corresponding to the sentiment evaluation anchor point.

[0033] In a further solution, the sentiment analysis confidence of the target evaluation talent is obtained according to the position distribution relationship between the ontology point set and the object point set, including:

[0034] Obtaining a preset neighborhood range and a preset threshold;

[0035] Counting the sentiment evaluation anchor points whose number of points in the ontology point set within the preset neighborhood exceeds a preset threshold, to obtain multiple first-category anchor points;

[0036] Counting the sentiment evaluation anchor points whose number of points in the object point set within the preset neighborhood exceeds a preset threshold, to obtain a plurality of second-category anchor points;

[0037] According to the similarity between multiple first-category anchor points and second-category anchor points, the sentiment analysis confidence of the target evaluation talent is obtained.

[0038] In a further solution, obtaining the sentiment analysis confidence of the target evaluation talent according to the similarity between the plurality of first-category anchor points and the second-category anchor points includes:

[0039] Counting the number of the same type of sentiment evaluation anchor points among the sentiment evaluation anchor points corresponding to the first type of anchor points and the sentiment evaluation anchor points corresponding to the second type of anchor points, to obtain a first value;

[0040] The sentiment analysis confidence of the target evaluation talent is obtained according to the ratio of the first value to the total number of all types of sentiment evaluation anchor points.

[0041] In a further solution: the target evaluation talent is managed according to the sentiment analysis result of the target evaluation talent, including:

[0042] Obtain the initial evaluation scores of the target evaluation talents;

[0043] Based on the preset formula, the initial evaluation score is corrected according to the sentiment analysis results of the target evaluation talent to obtain the target score of the target evaluation talent;

[0044] Manage target evaluation talents according to target scores;

[0045] The preset formulas include:

[0046] S = ω1 × S0 + ω2 × Vemo;

[0047] ω2=arctan(r×C);

[0048] ω1+ω2=1;

[0049] Among them, S is the target score of the target evaluation talent, S0 is the initial evaluation score of the target evaluation talent, ω1 and ω2 are the first weight and the second weight respectively, C is the sentiment analysis confidence of the target evaluation talent, r is the preset scaling factor, and arctan() is the inverse tangent function1.

[0050] The present invention also provides a talent evaluation management system based on artificial intelligence, comprising:

[0051] A record acquisition module is used to acquire the historical chat records of the target evaluation talent within a preset time period, wherein the historical chat records include the target entity chat records and the target object chat records;

[0052] A word segmentation and embedding module is used to perform word segmentation and word embedding processing on historical chat records to obtain a point set in a vector space corresponding to the word embedding, wherein the point set includes an ontology point set corresponding to the target ontology chat record and an object point set corresponding to the target object chat record;

[0053] An anchor point setting module, used to obtain the corresponding sentiment evaluation anchor points of multiple preset sentiment evaluation words in the vector space;

[0054] The sentiment analysis module is used to obtain the sentiment analysis results of the target evaluation talent according to the position distribution relationship between the ontology point set and the sentiment evaluation anchor point, and the position distribution relationship between the ontology point set and the object point set;

[0055] The talent management module is used to manage target evaluation talents based on the sentiment analysis results of target evaluation talents.

[0056] The beneficial effects of adopting the above embodiment are:

[0057] The present invention provides a talent evaluation management method based on artificial intelligence, which mainly realizes sentiment analysis in the talent evaluation management process by analyzing the historical chat records of the target evaluation talent within a preset time period. Specifically, the present invention maps the vocabulary in the historical chat records to the vector space by word embedding, and obtains the sentiment analysis result of the target evaluation talent according to the position distribution relationship between the ontology point set and the sentiment evaluation anchor point, and the position distribution relationship between the ontology point set and the object point set. The preset sentiment evaluation words are artificially set words that can evaluate emotions. By analyzing the position distribution relationship between the ontology point set and the sentiment evaluation anchor point, the emotional tendency of the target evaluation talent can be identified. In addition, the present invention also analyzes the position distribution relationship between the ontology point set and the object point set to reflect the commonality of the target ontology chat record and the target object chat record, thereby obtaining a more accurate analysis and evaluation result, solving the problem of how to combine sentiment analysis technology with talent evaluation. BRIEF DESCRIPTION OF THE DRAWINGS

[0058] Figure 1 A method flow chart of an embodiment of a talent evaluation management method based on artificial intelligence provided by the present invention;

[0059] Figure 2 for Figure 1 A method flow chart of an embodiment of step S104;

[0060] Figure 3 for Figure 2 A method flow chart of an embodiment of step S201;

[0061] Figure 4 for Figure 2 A method flow chart of an embodiment of step S202;

[0062] Figure 5 This is a system structure diagram of an embodiment of the talent evaluation management system based on artificial intelligence provided by the present invention. DETAILED DESCRIPTION

[0063] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0064] Combination Figure 1 As shown, a specific embodiment of the present invention discloses a talent evaluation management method based on artificial intelligence, including:

[0065] S101, obtaining the historical chat records of the target evaluation talent within a preset time period, wherein the historical chat records include the target entity chat records and the target object chat records;

[0066] S102, performing word segmentation and word embedding processing on the historical chat records to obtain a point set in the vector space corresponding to the word embedding, wherein the point set includes an entity point set corresponding to the target entity chat record and an object point set corresponding to the target object chat record;

[0067] S103, obtaining sentiment evaluation anchor points corresponding to a plurality of preset sentiment evaluation words in the vector space;

[0068] S104, obtaining a sentiment analysis result of the target evaluation talent according to the position distribution relationship between the ontology point set and the sentiment evaluation anchor point, and the position distribution relationship between the ontology point set and the object point set;

[0069] S105. Manage the target evaluation talents according to the sentiment analysis results of the target evaluation talents.

[0070] In the above process, the target evaluation talent refers to the person who is currently to be analyzed and managed. The preset time period refers to a period of time in the past that is artificially set. According to actual needs, it can be the past quarter, a project cycle, etc. The historical chat records refer to the chat records of the target evaluation talent when working within the preset time period, which can be obtained through existing office social software.

[0071] Obviously, a conversation chat is usually conducted by two or more people. Therefore, in this embodiment, the historical chat records are divided into target entity chat records and target object chat records according to different topics in the conversation. The target entity chat records are the words spoken by the target evaluation talent itself, and the target object chat records are the words spoken by the object chatting with the target evaluation talent.

[0072] In addition, in the above process, tokenization is the process of dividing text into meaningful units, usually words or subwords. In sentiment recognition, the input text needs to be tokenized first, and divided into independent words or subwords for subsequent processing and analysis. Word embedding is the process of mapping words to real domain vectors. It can represent words as dense vectors, which contain the semantic information of words. Through word embedding, words can be converted from discrete symbolic representations to continuous real vector representations, so as to better capture the semantic relationship between words. The word embedding vector space is the set space of word embedding representations, in which each word embedding vector corresponds to a word in the vocabulary. In this vector space, the positions of words and the distances between them reflect their semantic relationship, that is, words with similar semantics are closer in the vector space, while words with different semantics are farther away.

[0073] In sentiment recognition, commonly used word embedding models include Word2Vec, GloVe, and FastText, all of which can embed words into a high-dimensional vector space, allowing the model to better understand the semantic information of the text. Obviously, the specific process of word embedding is a prior art in this field, so this article will not go into detail.

[0074] The preset sentiment evaluation words are artificially set benchmark words for evaluating sentiment based on chat records, and the sentiment evaluation anchor points are the positions where these preset sentiment evaluation words are mapped in the word embedding vector space.

[0075] The present invention provides a talent evaluation management method based on artificial intelligence, which mainly realizes sentiment analysis in the talent evaluation management process by analyzing the historical chat records of the target evaluation talent within a preset time period. Specifically, the present invention maps the vocabulary in the historical chat records to the vector space by word embedding, and obtains the sentiment analysis result of the target evaluation talent according to the position distribution relationship between the ontology point set and the sentiment evaluation anchor point, and the position distribution relationship between the ontology point set and the object point set. The preset sentiment evaluation words are artificially set words that can evaluate emotions. By analyzing the position distribution relationship between the ontology point set and the sentiment evaluation anchor point, the emotional tendency of the target evaluation talent can be identified. In addition, the present invention also analyzes the position distribution relationship between the ontology point set and the object point set to reflect the commonality of the target ontology chat record and the target object chat record, thereby obtaining a more accurate analysis and evaluation result, solving the problem of how to combine sentiment analysis technology with talent evaluation.

[0076] Furthermore, in a preferred embodiment, the above step S102, performing word segmentation and word embedding processing on the historical chat records to obtain a point set in the vector space corresponding to the word embedding, specifically includes:

[0077] Segment the historical chat records to obtain multiple record words, each of which has two to four characters.

[0078] Based on the preset word embedding matrix, each record word is embedded, and each record word is mapped to a point in the vector space to obtain a point set.

[0079] When working, people often choose voice or other communication methods for overly complex communication dialogues. In text-based chat records, sentences are usually short, and two to four words in the sentence can well reflect the speaker's emotional tendency (such as two-word modal particles to four-word idioms). In view of this, in this embodiment, two to four words are selected for word segmentation to simplify the complexity of dictionary preparation, and at the same time simplify the word segmentation steps and improve analysis efficiency.

[0080] Specifically, in a preferred embodiment, the historical chat records are segmented to obtain multiple record words, specifically including:

[0081] Get the target chat record, which is a historical chat record to be segmented;

[0082] From the beginning to the end of the sentence, four consecutive characters in the target chat record can be overlapped and taken as the first word to be matched;

[0083] From the beginning to the end of the sentence, three consecutive characters in the target chat record can be taken in an overlapping manner as the second words to be matched;

[0084] From the beginning to the end of the sentence, two consecutive words in the target chat record can be taken in an overlapping manner as the third word to be matched;

[0085] The first to-be-matched word, the second to-be-matched word and the third to-be-matched word are matched with words in a preset dictionary, and the words that are successfully matched are used as recorded words.

[0086] The above process is an improvement on the existing forward maximum matching algorithm. The forward maximum matching algorithm is an existing word segmentation algorithm, which "cuts" the sentence during the word segmentation process. For example, after extracting the first words to be matched of four characters, the existing forward maximum matching algorithm will extract these first words to be matched from the original sentence, and perform the next three-sub-word segmentation on the remaining characters. In this embodiment, only the words are extracted without sentence segmentation, that is, the first words to be matched, the second words to be matched, and the third words to be matched with different numbers of characters are extracted at the same time. On the one hand, based on the foregoing, because the historical chat records in the work are generally short, the method of this embodiment will not increase the complexity of the algorithm too much. On the other hand, for shorter sentences, the method of this embodiment extracts as many words as possible from them, which will improve the robustness of word segmentation to a certain extent and optimize the accuracy of the final emotion recognition.

[0087] Further, combined with Figure 2 As shown, in a preferred embodiment, the above step S104, based on the position distribution relationship between the ontology point set and the sentiment evaluation anchor point, and the position distribution relationship between the ontology point set and the object point set, obtains the sentiment analysis result of the target evaluation talent, specifically including:

[0088] S201, obtaining a sentiment analysis value of the target evaluation talent according to the position distribution relationship between the ontology point set and the sentiment evaluation anchor point, where the sentiment analysis value is used to represent the evaluation level of the target evaluation talent in terms of emotional attitude;

[0089] S202. According to the position distribution relationship between the ontology point set and the object point set, the sentiment analysis confidence of the target evaluation talent is obtained. The sentiment analysis confidence is used to indicate the authenticity of the sentiment analysis value.

[0090] In the above process, sentiment evaluation is performed using the position distribution relationship between the ontology point set and the sentiment evaluation anchor point. However, in practice, considering the different speaking habits of each person, the pre-prepared preset sentiment evaluation words may not be suitable for everyone. Therefore, in this embodiment, the position distribution relationship between the ontology point set and the object point set is further used to evaluate the degree of match between the word embedding and sentiment analysis rules and the current target evaluation talent, and the sentiment analysis confidence index is obtained to evaluate the authenticity of the sentiment analysis value obtained this time. Combining the two indicators of sentiment analysis value and confidence for talent evaluation can obtain more reasonable results.

[0091] Specifically, combined Figure 3 As shown, in a preferred embodiment, the above step S201, obtaining the sentiment analysis value of the target evaluation talent according to the position distribution relationship between the ontology point set and the sentiment evaluation anchor point, specifically includes:

[0092] S301, counting and calculating the variance of the ontology point set in each dimension in the vector space, calculating the variance of the variance of the ontology point set in multiple dimensions in the vector space, and obtaining the distribution discreteness value of the ontology point set in the vector space;

[0093] S302, according to the distribution dispersion value, obtain the neighborhood radius value, the neighborhood radius value is proportional to the distribution dispersion value;

[0094] S303, counting the number of points in the ontology point set whose Euclidean distance from each sentiment evaluation anchor point is less than the neighborhood radius value, and obtaining the neighborhood statistics corresponding to each sentiment evaluation anchor point;

[0095] S304: Obtain the preset emotion value corresponding to each emotion evaluation anchor point, and obtain the emotion analysis value of the target evaluation talent according to the preset emotion value and neighborhood statistics corresponding to each emotion evaluation anchor point.

[0096] The above process mainly describes the position distribution relationship between the ontology point set and the emotion evaluation anchor point based on the number of points near each emotion evaluation anchor point. The more points there are near a certain emotion evaluation anchor point, the more the chat record has the emotion tendency represented by the preset emotion evaluation word corresponding to the emotion evaluation anchor point. In addition, in this embodiment, the degree of discreteness of the ontology point set in the vector space is mainly represented by the distribution discreteness value, and the domain radius value is further adjusted according to this discreteness to control the tightness of the comparison scale of the position distribution relationship between the ontology point set and the emotion evaluation anchor point. The higher the degree of discreteness of the ontology point set, the more reasonable the effect of word segmentation, i.e., word embedding, can be appropriately increased at this time to refer to more data for emotion evaluation. The lower the degree of discreteness of the ontology point set, the poorer the effect of word segmentation, i.e., word embedding, or the insufficient number of samples of the historical chat technology, can be appropriately reduced at this time to ensure the validity of the data.

[0097] Further, in a preferred embodiment, the preset sentiment evaluation words include a plurality of positive words and a plurality of negative words corresponding to each other, and the preset sentiment value of the sentiment evaluation anchor point corresponding to each positive word is the inverse of the preset sentiment value of the sentiment evaluation anchor point corresponding to the negative word corresponding to the positive word. The above step S304, obtaining the preset sentiment value corresponding to each sentiment evaluation anchor point, and obtaining the sentiment analysis value of the target evaluation talent according to the preset sentiment value corresponding to each sentiment evaluation anchor point and the neighborhood statistics, specifically includes:

[0098] The sentiment analysis value of the target evaluation talent is calculated by the following formula:

[0099] Vemo=∑Vp×Nn;

[0100] Among them, Vemo is the sentiment analysis value of the target evaluation talent, Vp is the preset sentiment value corresponding to the sentiment evaluation anchor point, and Nn is the neighborhood statistical value corresponding to the sentiment evaluation anchor point.

[0101] The above process provides a specific process for calculating sentiment analysis values, which divides sentiment into two categories: positive and negative. The sentiment analysis values ​​obtained in the end also represent positive and negative tendencies. For example, positive words and negative words can be:

[0102]

[0103] At this time, the sentiment analysis value calculated by the above formula can indicate whether the target evaluation talent has a positive tendency or a negative tendency. The larger the sentiment analysis value, the more positive the target evaluation talent may be at work, and vice versa.

[0104] Further, combined with Figure 4 As shown, in a preferred embodiment, the above step S202, based on the position distribution relationship between the ontology point set and the object point set, obtains the sentiment analysis confidence of the target evaluation talent, specifically including:

[0105] S401, obtaining a preset neighborhood range and a preset threshold;

[0106] S402, counting the sentiment evaluation anchor points whose number of points in the ontology point set within the preset neighborhood exceeds a preset threshold, to obtain a plurality of first-category anchor points;

[0107] S403, counting the number of sentiment evaluation anchor points whose number in the object point set within the preset neighborhood exceeds a preset threshold, to obtain a plurality of second-category anchor points;

[0108] S404. Obtain the sentiment analysis confidence of the target evaluation talent according to the similarity between the plurality of first-category anchor points and the second-category anchor points.

[0109] The above process is to evaluate the authenticity of this sentiment analysis by comparing the word embedding effect of the target object chat record and the word embedding effect of the target entity chat record. The main reason is that the emotional tone of the two parties in communication is relatively close in most cases, that is, it is rare for one party to be too positive and the other party to be too negative. Therefore, in this embodiment, the first type of anchor point and the second type of anchor point are used to respectively reflect the different emotional characteristics represented by the entity point set and the object point set, and the confidence is obtained according to the difference between the two.

[0110] Further, in a preferred embodiment, the above step S404, based on the similarity between the plurality of first-type anchor points and the second-type anchor points, obtains the sentiment analysis confidence of the target evaluation talent, specifically including:

[0111] Counting the number of the same type of sentiment evaluation anchor points among the sentiment evaluation anchor points corresponding to the first type of anchor points and the sentiment evaluation anchor points corresponding to the second type of anchor points, to obtain a first value;

[0112] The sentiment analysis confidence of the target evaluation talent is obtained according to the ratio of the first value to the total number of all types of sentiment evaluation anchor points.

[0113] The above process is a preferred way to calculate confidence. Furthermore, the present invention also provides a more detailed embodiment to clearly illustrate the application of the above sentiment analysis in talent evaluation:

[0114] Assume that at the end of a project, it is necessary to evaluate the personnel involved in the project in order to determine the staffing for the next project of the same type. At this point, the initial evaluation score for each person has been obtained based on other indicators (such as performance, overtime hours, etc.). It is now necessary to combine sentiment analysis technology to take into account factors such as employees' willingness and enthusiasm for engaging in this type of project. At this point, the sentiment analysis value and sentiment analysis confidence of the target evaluation talent can be obtained in the above manner. At this point, the above step S105 can be performed. In this embodiment, step S105 manages the target evaluation talent according to the sentiment analysis results of the target evaluation talent, specifically including:

[0115] Obtain the initial evaluation scores of the target evaluation talents;

[0116] Based on the preset formula, the initial evaluation score is corrected according to the sentiment analysis results of the target evaluation talent to obtain the target score of the target evaluation talent;

[0117] Manage target evaluation talents according to target scores;

[0118] The preset formulas include:

[0119] S = ω1 × S0 + ω2 × Vemo;

[0120] ω2=arctan(r×C);

[0121] ω1+ω2=1;

[0122] Among them, S is the target score of the target evaluation talent, S0 is the initial evaluation score of the target evaluation talent, ω1 and ω2 are the first weight and the second weight respectively, C is the sentiment analysis confidence of the target evaluation talent, r is the preset scaling factor, and arctan() is the inverse tangent function1.

[0123] The higher the target score, the more suitable the target evaluation talent is for the project. In the above process, the first weight and the second weight are adjusted by the sentiment analysis confidence to adjust the influence of the sentiment analysis value on the initial evaluation score. The lower the sentiment analysis confidence, the lower the reference value of the sentiment analysis value. Then the smaller the value of the second weight is, the smaller the influence of the sentiment analysis value on the initial evaluation score is, and vice versa.

[0124] Combination Figure 5 As shown, the present invention also provides a talent evaluation management system based on artificial intelligence, comprising:

[0125] The record acquisition module 510 is used to acquire the historical chat records of the target evaluation talent within a preset time period, wherein the historical chat records include the target entity chat records and the target object chat records;

[0126] A word segmentation and embedding module 520 is used to perform word segmentation and word embedding processing on the historical chat records to obtain a point set in the vector space corresponding to the word embedding, wherein the point set includes an ontology point set corresponding to the target ontology chat record and an object point set corresponding to the target object chat record;

[0127] Anchor point setting module 530, used to obtain sentiment evaluation anchor points corresponding to a plurality of preset sentiment evaluation words in the vector space;

[0128] The sentiment analysis module 540 is used to obtain the sentiment analysis result of the target evaluation talent according to the position distribution relationship between the ontology point set and the sentiment evaluation anchor point, and the position distribution relationship between the ontology point set and the object point set;

[0129] The talent management module 550 is used to manage the target evaluation talents according to the sentiment analysis results of the target evaluation talents.

[0130] It should be noted here that the artificial intelligence-based talent evaluation and management system provided in the above embodiments can implement the technical solutions described in the above method embodiments. The specific implementation principles of the above modules or units can be found in the corresponding contents in the above method embodiments, which will not be repeated here.

[0131] The present invention provides a talent evaluation management method based on artificial intelligence, which mainly realizes sentiment analysis in the talent evaluation management process by analyzing the historical chat records of the target evaluation talent within a preset time period. Specifically, the present invention maps the vocabulary in the historical chat records to the vector space by word embedding, and obtains the sentiment analysis result of the target evaluation talent according to the position distribution relationship between the ontology point set and the sentiment evaluation anchor point, and the position distribution relationship between the ontology point set and the object point set. The preset sentiment evaluation words are artificially set words that can evaluate emotions. By analyzing the position distribution relationship between the ontology point set and the sentiment evaluation anchor point, the emotional tendency of the target evaluation talent can be identified. In addition, the present invention also analyzes the position distribution relationship between the ontology point set and the object point set to reflect the commonality of the target ontology chat record and the target object chat record, thereby obtaining a more accurate analysis and evaluation result, solving the problem of how to combine sentiment analysis technology with talent evaluation.

[0132] It should be noted that the various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the various embodiments can be referenced to each other.

[0133] The above description of the disclosed embodiments enables one skilled in the art to implement or use the present invention. Various modifications to these embodiments will be apparent to one skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but rather to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A talent evaluation management method based on artificial intelligence, characterized in that: include: Obtain the historical chat records of the target evaluation talent within a preset time period, wherein the historical chat records include the target entity chat records and the target object chat records; Perform word segmentation and word embedding processing on the historical chat records to obtain a point set in the vector space corresponding to the word embedding, wherein the point set includes an ontology point set corresponding to the target ontology chat records and an object point set corresponding to the target object chat records; Obtaining sentiment evaluation anchor points corresponding to multiple preset sentiment evaluation words in the vector space; According to the position distribution relationship between the ontology point set and the emotional evaluation anchor point, as well as the position distribution relationship between the ontology point set and the object point set, the emotional analysis results of the target evaluation talent are obtained; Manage target evaluation talents based on their sentiment analysis results; The sentiment analysis result of the target evaluation talent is obtained according to the position distribution relationship between the ontology point set and the sentiment evaluation anchor point, and the position distribution relationship between the ontology point set and the object point set, including: According to the position distribution relationship between the ontology point set and the emotional evaluation anchor point, the emotional analysis value of the target evaluation talent is obtained. The emotional analysis value is used to represent the evaluation level of the target evaluation talent in terms of emotional attitude; According to the position distribution relationship between the ontology point set and the object point set, the sentiment analysis confidence of the target evaluation talent is obtained. The sentiment analysis confidence is used to indicate the authenticity of the sentiment analysis value.

2. The talent evaluation management method based on artificial intelligence according to claim 1 is characterized in that: The word segmentation and word embedding processing of the historical chat records is performed to obtain a point set in the vector space corresponding to the word embedding, including: Segment the historical chat records to obtain multiple record words, each of which has two to four characters. Based on the preset word embedding matrix, each record word is embedded, and each record word is mapped to a point in the vector space to obtain a point set.

3. The talent evaluation management method based on artificial intelligence according to claim 2 is characterized in that: The historical chat records are segmented to obtain multiple record words, including: Get the target chat record, which is a historical chat record to be segmented; From the beginning to the end of the sentence, four consecutive characters in the target chat record can be taken in an overlapping manner as the first word to be matched; From the beginning to the end of the sentence, three consecutive characters in the target chat record can be taken in an overlapping manner as the second words to be matched; From the beginning to the end of the sentence, two consecutive words in the target chat record can be taken in an overlapping manner as the third word to be matched; The first to-be-matched word, the second to-be-matched word and the third to-be-matched word are matched with words in a preset dictionary, and the words that are successfully matched are used as recorded words.

4. The talent evaluation management method based on artificial intelligence according to claim 1 is characterized in that: The method of obtaining the sentiment analysis value of the target evaluation talent according to the position distribution relationship between the ontology point set and the sentiment evaluation anchor point includes: Count and calculate the variance of the ontology point set in each dimension in the vector space, calculate the variance of the variance of the ontology point set in multiple dimensions in the vector space, and obtain the distribution discreteness value of the ontology point set in the vector space; According to the distribution dispersion value, the neighborhood radius value is obtained, and the neighborhood radius value is proportional to the distribution dispersion value; In the ontology point set, the number of points whose Euclidean distance from each sentiment evaluation anchor point is less than the neighborhood radius value is counted to obtain the neighborhood statistics value corresponding to each sentiment evaluation anchor point; The preset sentiment value corresponding to each sentiment evaluation anchor point is obtained, and the sentiment analysis value of the target evaluation talent is obtained according to the preset sentiment value and neighborhood statistical value corresponding to each sentiment evaluation anchor point.

5. The talent evaluation management method based on artificial intelligence according to claim 4 is characterized in that: The preset emotion evaluation words include a plurality of positive words and a plurality of negative words corresponding to each other, and the preset emotion value of the emotion evaluation anchor point corresponding to each positive word is the inverse of the preset emotion value of the emotion evaluation anchor point corresponding to the negative word corresponding to the positive word; the step of obtaining the preset emotion value corresponding to each emotion evaluation anchor point and obtaining the emotion analysis value of the target evaluation talent according to the preset emotion value corresponding to each emotion evaluation anchor point and the neighborhood statistics value includes: The sentiment analysis value of the target evaluation talent is calculated by the following formula: ； in, Evaluate the sentiment analysis value of the talent for the target, is the preset sentiment value corresponding to the sentiment evaluation anchor point, It is the neighborhood statistics corresponding to the sentiment evaluation anchor point.

6. The talent evaluation management method based on artificial intelligence according to claim 5 is characterized in that: The sentiment analysis confidence of the target evaluation talent is obtained according to the position distribution relationship between the ontology point set and the object point set, including: Obtaining a preset neighborhood range and a preset threshold; Counting the sentiment evaluation anchor points whose number of points in the ontology point set within the preset neighborhood exceeds a preset threshold, to obtain multiple first-category anchor points; Counting the sentiment evaluation anchor points whose number of points in the object point set within the preset neighborhood exceeds a preset threshold, to obtain a plurality of second-category anchor points; According to the similarity between multiple first-category anchor points and second-category anchor points, the sentiment analysis confidence of the target evaluation talent is obtained.

7. The talent evaluation management method based on artificial intelligence according to claim 6 is characterized in that: The step of obtaining the sentiment analysis confidence of the target evaluation talent according to the similarity between the plurality of first-category anchor points and the second-category anchor points includes: Counting the number of the same type of sentiment evaluation anchor points among the sentiment evaluation anchor points corresponding to the first type of anchor points and the sentiment evaluation anchor points corresponding to the second type of anchor points, to obtain a first value; The sentiment analysis confidence of the target evaluation talent is obtained according to the ratio of the first value to the total number of all types of sentiment evaluation anchor points.

8. The talent evaluation management method based on artificial intelligence according to claim 7 is characterized in that: The target evaluation talent is managed according to the sentiment analysis result of the target evaluation talent, including: Obtain the initial evaluation scores of the target evaluation talents; Based on the preset formula, the initial evaluation score is corrected according to the sentiment analysis results of the target evaluation talent to obtain the target score of the target evaluation talent; Manage target evaluation talents according to target scores; The preset formulas include: ； ; ； in, Target score for target evaluation talent, The initial evaluation score for the target evaluation talent. and are the first weight and the second weight respectively, To evaluate the confidence of sentiment analysis of talents for the target, is the preset zoom factor, is the inverse tangent function1.

9. A talent evaluation management system based on artificial intelligence, characterized in that: include: A record acquisition module is used to acquire the historical chat records of the target evaluation talent within a preset time period, wherein the historical chat records include the target entity chat records and the target object chat records; A word segmentation and embedding module is used to perform word segmentation and word embedding processing on historical chat records to obtain a point set in a vector space corresponding to the word embedding, wherein the point set includes an ontology point set corresponding to the target ontology chat record and an object point set corresponding to the target object chat record; An anchor point setting module, used to obtain the corresponding sentiment evaluation anchor points of multiple preset sentiment evaluation words in the vector space; The sentiment analysis module is used to obtain the sentiment analysis results of the target evaluation talent according to the position distribution relationship between the ontology point set and the sentiment evaluation anchor point, and the position distribution relationship between the ontology point set and the object point set; The talent management module is used to manage the target evaluation talents according to the sentiment analysis results of the target evaluation talents; The sentiment analysis result of the target evaluation talent is obtained according to the position distribution relationship between the ontology point set and the sentiment evaluation anchor point, and the position distribution relationship between the ontology point set and the object point set, including: According to the position distribution relationship between the ontology point set and the emotional evaluation anchor point, the emotional analysis value of the target evaluation talent is obtained. The emotional analysis value is used to represent the evaluation level of the target evaluation talent in terms of emotional attitude; According to the position distribution relationship between the ontology point set and the object point set, the sentiment analysis confidence of the target evaluation talent is obtained. The sentiment analysis confidence is used to indicate the authenticity of the sentiment analysis value.

Images