Multi-dimensional talent assessment method and system based on artificial intelligence technology

Abstract

The invention relates to the technical field of talent assessment, and discloses a multi-dimensional talent assessment method and system based on an artificial intelligence technology, and the method comprises the following steps: S101, collecting resume text data and interview video data of a to-be-tested person; step S102, preprocessing the resume text data and the interview video data to respectively obtain a first feature sequence and a second feature sequence; s103, a scoring model is constructed according to the first feature sequence and the second feature sequence, the input of the scoring model is the first feature sequence and the second feature sequence, and the output of the scoring model is the score of the person to be tested; and step S104, training and optimizing the scoring model. According to the method, the multi-modal data is fused, the resume text data and the interview video data are combined, the multi-dimensional features of the to-be-tested person are extracted, and the scoring model is constructed to output the score of the to-be-tested person, so that the subjectivity in the evaluation process is reduced, and objective and standardized scores are provided.

Description

Technical Field

[0001] The present invention belongs to the technical field of talent assessment, and specifically relates to a multi-dimensional talent assessment method and system based on artificial intelligence technology. Background Art

[0002] Talent assessment is an important part of human resource management, which directly affects the effectiveness of recruitment and the overall competitiveness of the organization. Through scientific talent assessment, enterprises can identify the skills, qualities and potential of candidates and arrange suitable talents for different positions. Therefore, accurate and effective talent assessment is of great significance to ensure that enterprises obtain high-quality talents, improve employee performance and reduce recruitment errors.

[0003] Currently, companies generally use resume screening, interviews, personality tests and other methods to conduct talent assessment. Although these methods help companies understand the abilities and characteristics of candidates to a certain extent, they generally rely on the subjective judgment of interviewers and lack unified objective standards. The personal experience, preferences and emotions of interviewers may affect the evaluation results, resulting in different evaluations of the same candidate by different evaluators. At the same time, the ambiguity and inconsistency of evaluation criteria may also lead to the neglect of some candidates with potential but non-traditional backgrounds.

[0004] Therefore, a multidimensional talent assessment method is urgently needed. Summary of the invention

[0005] The present invention provides a multi-dimensional talent assessment method and system based on artificial intelligence technology to solve the technical problems of over-reliance on subjective judgment and lack of unified objective standards in related technologies.

[0006] The present invention provides a multi-dimensional talent evaluation method and system based on artificial intelligence technology, comprising the following steps:

[0007] Step S101, collecting the resume text data and interview video data of the test subject;

[0008] Step S102, preprocessing the resume text data and the interview video data to obtain a first feature sequence and a second feature sequence respectively;

[0009] Step S103, constructing a scoring model based on the first feature sequence and the second feature sequence, the input of the scoring model is the first feature sequence and the second feature sequence, and the output is the score of the subject;

[0010] Step S104: training and optimizing the scoring model.

[0011] Furthermore, the step of preprocessing the resume text data and the interview video data includes:

[0012] Step S201, cleaning the resume text data, specifically including: removing special symbols, removing irrelevant numbers, removing spaces, removing line breaks, and unifying upper and lower case;

[0013] Step S202, segmenting and encoding the resume text data, specifically, using a BERT model-based segmenter to segment the text into words or subwords, using the BERT model to encode the resume text, converting the resume text into a sequence based on word vectors, and representing it through a first feature sequence;

[0014] Step S203, extracting video frames from the interview video data at a preset time interval;

[0015] Step S204: normalize the video frame using the maximum-minimum normalization method and represent it through a second feature sequence.

[0016] Further, the scoring model includes: a first hidden layer, a second hidden layer, an expansion layer, a third hidden layer, a fusion layer and a first classifier;

[0017] The first hidden layer inputs the first feature sequence and outputs the first hidden feature;

[0018] The second hidden layer inputs the sequence unit of the second feature sequence and outputs the second hidden feature;

[0019] The expansion layer is used to expand the first hidden feature into a first feature vector and expand the second hidden feature into a second feature vector;

[0020] The third hidden layer inputs a time series based on the second feature vector and outputs a third feature vector;

[0021] The fusion layer is used to fuse the first eigenvector and the third eigenvector to obtain the fourth eigenvector;

[0022] The fourth eigenvector is input into the first classifier, and the classification space of the first classifier represents the score of the subject.

[0023] Furthermore, the calculation formula of the first hidden layer is:

[0024]

[0025] Where Attention represents the first hidden feature output by the first hidden layer, Query, Key and Value represent the first intermediate feature, the second intermediate feature and the third intermediate feature of the first hidden layer respectively, T represents the transposition operation, and softmax represents the softmax activation function.

[0026] Furthermore, the second hidden layer is constructed based on a convolutional neural network.

[0027] Furthermore, the calculation formula of the third hidden layer includes:

[0028] The calculation formula of the reset gate of the third hidden layer is:

[0029]

[0030] Among them, x t represents the tth sequence unit of the time series, h t-1 represents the t-1th third eigenvector, and Respectively represent the first weight parameter, the second weight parameter and the bias parameter corresponding to the reset gate of the t-th sequence unit of the time series;

[0031] The calculation formula of the update gate of the third hidden layer is:

[0032]

[0033] in, and Respectively represent the first weight parameter, the second weight parameter and the bias parameter corresponding to the update gate of the t-th sequence unit of the time series;

[0034] The calculation formula for the candidate hidden features of the third hidden layer is:

[0035]

[0036] in, r t and Respectively represent the first weight parameter, the second weight parameter and the bias parameter corresponding to the candidate hidden feature of the t-th sequence unit of the time series;

[0037] The calculation formula for the hidden state update of the third hidden layer is:

[0038]

[0039] Among them, z t and represents the update gate and candidate hidden features of the t-th sequence unit of the time series, and ⊙ represents point-by-point multiplication.

[0040] Furthermore, the fusion layer is used to fuse the first eigenvector and the third eigenvector by weighted summation, and the calculation formula is:

[0041] V3=α*V1+β*V2;

[0042] Among them, V3 represents the fourth eigenvector after fusion, V1 and V2 represent the first eigenvector and the third eigenvector respectively, and α and β represent the weight parameter of the first eigenvector and the weight parameter of the third eigenvector respectively.

[0043] Furthermore, manual annotation is used as sample labels for scoring model training.

[0044] Furthermore, the cross entropy loss function is used as the loss function of the scoring model, and the adaptive moment estimation optimizer is used to adjust the weight of the scoring model. The formula of the adaptive moment estimation optimizer is:

[0045]

[0046] Among them, θ represents the weight parameter in the scoring model, ρ represents the learning rate, and m t and v t represents the first-order and second-order momentum, and ε represents a random number ranging from 0 to 1.

[0047] The present invention also provides a multi-dimensional talent evaluation system based on artificial intelligence technology, comprising the following modules:

[0048] A data collection module is used to collect the resume text data and interview video data of the test subject;

[0049] A preprocessing module, which is used to preprocess the resume text data and the interview video data to obtain a first feature sequence and a second feature sequence respectively;

[0050] A scoring module, which is used to construct a scoring model based on the first feature sequence and the second feature sequence to output a score of the subject;

[0051] The training and optimization module is used to train and optimize the scoring model.

[0052] The beneficial effects of the present invention are as follows: the present invention combines resume text data with interview video data by fusing multimodal data, extracts multidimensional features of the test subject, and constructs a scoring model to output the test subject's score. Compared with traditional methods, the present invention can reduce the subjectivity in the evaluation process, provide objective and standardized scores, and effectively evaluate the soft and hard skills of candidates. In addition, the use of automated data processing and optimization mechanisms, such as dynamic adjustment of learning rate, K-fold cross-validation, etc., improves the generalization ability and evaluation efficiency of the model, and significantly improves the accuracy and fairness of talent evaluation. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] Figure 1 It is a flow chart of a multi-dimensional talent assessment method based on artificial intelligence technology of the present invention;

[0054] Figure 2 It is a flow chart of preprocessing resume text data and interview video data of the present invention. DETAILED DESCRIPTION

[0055] The subject matter described herein will now be discussed with reference to example embodiments. It should be understood that the discussion of these embodiments is only to enable those skilled in the art to better understand and implement the subject matter described herein, and the functions and arrangements of the elements discussed may be changed without departing from the scope of protection of the contents of this specification. Each example may omit, replace or add various processes or components as needed. In addition, the features described relative to some examples may also be combined in other examples.

[0056] It should be noted that, unless otherwise defined, the technical terms or scientific terms used in one or more embodiments of the present invention should be understood by people with ordinary skills in the field to which the present invention belongs. The words "first", "second" and similar words used in one or more embodiments of the present invention do not indicate any order, quantity or importance, but are only used to distinguish different components. "Include" or "comprise" and similar words mean that the elements or objects appearing before the word include the elements or objects listed after the word and their equivalents, without excluding other elements or objects. "Connect" or "connected" and similar words are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. "Up", "down", "left", "right" and the like are only used to indicate relative positional relationships. When the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0057] like Figure 1-Figure 2 As shown, a multi-dimensional talent evaluation method and system based on artificial intelligence technology includes the following steps:

[0058] Step S101, collecting the resume text data and interview video data of the test subject;

[0059] Step S102, preprocessing the resume text data and the interview video data to obtain a first feature sequence and a second feature sequence respectively;

[0060] Step S103, constructing a scoring model based on the first feature sequence and the second feature sequence, the input of the scoring model is the first feature sequence and the second feature sequence, and the output is the score of the subject;

[0061] Step S104: training and optimizing the scoring model.

[0062] In one embodiment of the present invention, the present invention supports automated data collection and real-time processing functions, simplifies the data collection process and ensures the timeliness of the data. The present invention can be integrated with recruitment websites and HR systems to automatically obtain resume text data from major platforms, avoiding the tedious process of manual uploading. For interview video data, the present invention can record the interview video in real time through a webcam and microphone with the consent of the testee, automatically save and pre-process it to improve efficiency.

[0063] In one embodiment of the present invention, the step of preprocessing the resume text data and the interview video data includes:

[0064] Step S201, cleaning the resume text data, specifically including: removing special symbols, removing irrelevant numbers (except for years and experience related), removing spaces, removing line breaks, and unifying upper and lower case;

[0065] Step S202, segmenting and encoding the resume text data, specifically, using a BERT model-based segmenter to segment the text into words or subwords, using the BERT model to encode the resume text, converting the resume text into a sequence based on word vectors, and representing it through a first feature sequence;

[0066] Step S203, extracting video frames from the interview video data at a preset time interval;

[0067] Step S204: normalize the video frame using the maximum-minimum normalization method and represent it through a second feature sequence.

[0068] In one embodiment of the present invention, the scoring model includes: a first hidden layer, a second hidden layer, an expansion layer, a third hidden layer, a fusion layer and a first classifier;

[0069] The first hidden layer inputs the first feature sequence and outputs the first hidden feature;

[0070] The second hidden layer inputs the sequence unit of the second feature sequence and outputs the second hidden feature;

[0071] The expansion layer is used to expand the first hidden feature into a first feature vector and expand the second hidden feature into a second feature vector;

[0072] The third hidden layer inputs a time series based on the second feature vector and outputs a third feature vector;

[0073] The fusion layer is used to fuse the first eigenvector and the third eigenvector to obtain the fourth eigenvector;

[0074] The fourth eigenvector is input into the first classifier, and the classification space of the first classifier represents the score of the subject.

[0075] In one embodiment of the present invention, the calculation formula of the first hidden layer is:

[0076]

[0077] Where Attention represents the first hidden feature output by the first hidden layer, Query, Key and Value represent the first intermediate feature, the second intermediate feature and the third intermediate feature of the first hidden layer respectively, T represents the transposition operation, and softmax represents the softmax activation function.

[0078] In one embodiment of the present invention, the second hidden layer is constructed based on a convolutional neural network.

[0079] In one embodiment of the present invention, the calculation formula of the third hidden layer includes:

[0080] The calculation formula of the reset gate of the third hidden layer is:

[0081]

[0082] Among them, x t represents the tth sequence unit of the time series, h t-1 represents the t-1th third eigenvector, and Respectively represent the first weight parameter, the second weight parameter and the bias parameter corresponding to the reset gate of the t-th sequence unit of the time series;

[0083] The calculation formula of the update gate of the third hidden layer is:

[0084]

[0085] in, and Respectively represent the first weight parameter, the second weight parameter and the bias parameter corresponding to the update gate of the t-th sequence unit of the time series;

[0086] The calculation formula for the candidate hidden features of the third hidden layer is:

[0087]

[0088] in, r t and Respectively represent the first weight parameter, the second weight parameter and the bias parameter corresponding to the candidate hidden feature of the t-th sequence unit of the time series;

[0089] The calculation formula for the hidden state update of the third hidden layer is:

[0090]

[0091] Among them, z t and represents the update gate and candidate hidden features of the t-th sequence unit of the time series, and ⊙ represents point-by-point multiplication.

[0092] In one embodiment of the present invention, the fusion layer is used to fuse the first eigenvector and the third eigenvector by weighted summation, and the calculation formula is:

[0093] V3=α*V1+β*V2;

[0094] Among them, V3 represents the fourth eigenvector after fusion, V1 and V2 represent the first eigenvector and the third eigenvector respectively, and α and β represent the weight parameter of the first eigenvector and the weight parameter of the third eigenvector respectively.

[0095] In one embodiment of the present invention, manual annotation is used as sample labels for scoring model training.

[0096] In one embodiment of the present invention, data enhancement technology can be used for resume text data to generate more training data to improve the generalization ability of the model. Data enhancement techniques include synonym replacement, random deletion, random insertion, etc. Synonym replacement refers to generating new data by replacing certain words in the resume text with their synonyms; random deletion refers to randomly deleting some words to simulate formatting problems or simplify expressions that may occur in the resume text. For interview video data, video enhancement technology can be used, including rotation, cropping, changing brightness, adding noise, etc., to increase the diversity of training data. Video rotation refers to rotating video frames within a certain range to simulate performance under different shooting angles; random cropping refers to randomly cropping video frames to ensure the adaptability of the scoring model to incomplete frames; adding noise refers to adding Gaussian noise to the video frame to enhance the robustness of the scoring model in a noisy environment.

[0097] In one embodiment of the present invention, a cross entropy loss function is used as the loss function of the scoring model, and an adaptive moment estimation optimizer is used to adjust the weight of the scoring model. The formula of the adaptive moment estimation optimizer is:

[0098]

[0099] Among them, θ represents the weight parameter in the scoring model, ρ represents the learning rate, and m t and v t represents the first-order and second-order momentum, and ε represents a random number ranging from 0 to 1.

[0100] In one embodiment of the present invention, in the initial stage of training, the learning rate is set between 0.0001 and 0.001, and during the training process, the value of the learning rate is dynamically adjusted according to the change of the value of the loss function.

[0101] In one embodiment of the present invention, K-fold cross validation is used to evaluate the generalization ability of the scoring model to avoid overfitting of the scoring model; during the training process, when the value of the loss function no longer changes, the training is stopped.

[0102] The present invention also provides a multi-dimensional talent evaluation system based on artificial intelligence technology, comprising the following modules:

[0103] A data collection module is used to collect the resume text data and interview video data of the test subject;

[0104] A preprocessing module, which is used to preprocess the resume text data and the interview video data to obtain a first feature sequence and a second feature sequence respectively;

[0105] A scoring module, which is used to construct a scoring model based on the first feature sequence and the second feature sequence to output a score of the subject;

[0106] The training and optimization module is used to train and optimize the scoring model.

[0107] The above describes an embodiment of the present embodiment, but the present embodiment is not limited to the above-mentioned specific implementation methods. The above-mentioned specific implementation methods are merely illustrative and not restrictive. Under the guidance of the present embodiment, ordinary technicians in this field can also make many forms, all of which are within the protection of the present embodiment.

Claims

1. A multi-dimensional talent assessment method based on artificial intelligence technology, characterized in that: The following steps are involved: Step S101, collecting the resume text data and interview video data of the test subject; Step S102, preprocessing the resume text data and the interview video data to obtain a first feature sequence and a second feature sequence respectively; Step S103, constructing a scoring model based on the first feature sequence and the second feature sequence, the input of the scoring model is the first feature sequence and the second feature sequence, and the output is the score of the subject; Step S104: training and optimizing the scoring model.

2. According to claim 1, a multi-dimensional talent assessment method based on artificial intelligence technology is characterized in that: The steps of preprocessing the resume text data and the interview video data include: Step S201, cleaning the resume text data, specifically including: removing special symbols, removing irrelevant numbers, removing spaces, removing line breaks, and unifying upper and lower case; Step S202, segmenting and encoding the resume text data, specifically, using a BERT model-based segmenter to segment the text into words or phrases, using the BERT model to encode the resume text, converting the resume text into a sequence based on word vectors, and representing it through a first feature sequence; Step S203, extracting video frames from the interview video data at a preset time interval; Step S204: normalize the video frame using the maximum-minimum normalization method and represent it through a second feature sequence.

3. According to claim 1, a multi-dimensional talent assessment method based on artificial intelligence technology is characterized in that: The scoring model includes: a first hidden layer, a second hidden layer, an expansion layer, a third hidden layer, a fusion layer and a first classifier; The first hidden layer inputs the first feature sequence and outputs the first hidden feature; The second hidden layer inputs the sequence unit of the second feature sequence and outputs the second hidden feature; The expansion layer is used to expand the first hidden feature into a first feature vector and expand the second hidden feature into a second feature vector; The third hidden layer inputs a time series based on the second feature vector and outputs a third feature vector; The fusion layer is used to fuse the first eigenvector and the third eigenvector to obtain the fourth eigenvector; The fourth eigenvector is input into the first classifier, and the classification space of the first classifier represents the score of the subject.

4. The multi-dimensional talent assessment method based on artificial intelligence technology according to claim 3 is characterized in that: The calculation formula of the first hidden layer is: Where Attention represents the first hidden feature output by the first hidden layer, Query, Key and Value represent the first intermediate feature, the second intermediate feature and the third intermediate feature of the first hidden layer respectively, T represents the transposition operation, and softmax represents the softmax activation function.

5. According to claim 3, a multi-dimensional talent assessment method based on artificial intelligence technology is characterized in that: The second hidden layer is constructed based on a convolutional neural network.

6. The multi-dimensional talent assessment method based on artificial intelligence technology according to claim 3 is characterized in that: The calculation formula of the third hidden layer includes: The calculation formula of the reset gate of the third hidden layer is: Among them, x t represents the tth sequence unit of the time series, h t-1 represents the t-1th third eigenvector, and Respectively represent the first weight parameter, the second weight parameter and the bias parameter corresponding to the reset gate of the t-th sequence unit of the time series; The calculation formula of the update gate of the third hidden layer is: in, and Respectively represent the first weight parameter, the second weight parameter and the bias parameter corresponding to the update gate of the t-th sequence unit of the time series; The calculation formula for the candidate hidden features of the third hidden layer is: in, r t and Respectively represent the first weight parameter, the second weight parameter and the bias parameter corresponding to the candidate hidden feature of the t-th sequence unit of the time series; The calculation formula for the hidden state update of the third hidden layer is: Among them, z t and represents the update gate and candidate hidden features of the t-th sequence unit of the time series, and ⊙ represents point-by-point multiplication.

7. The multi-dimensional talent assessment method based on artificial intelligence technology according to claim 3 is characterized in that: The fusion layer is used to fuse the first eigenvector and the third eigenvector by weighted summation, and the calculation formula is: V3=α*V1+β*V2; Among them, V3 represents the fourth eigenvector after fusion, V1 and V2 represent the first eigenvector and the third eigenvector respectively, and α and β represent the weight parameter of the first eigenvector and the weight parameter of the third eigenvector respectively.

8. The multi-dimensional talent assessment method based on artificial intelligence technology according to claim 1 is characterized in that: Manual annotation is used as sample labels for scoring model training.

9. The multi-dimensional talent assessment method based on artificial intelligence technology according to claim 1, characterized in that: The cross entropy loss function is used as the loss function of the scoring model, and the adaptive moment estimation optimizer is used to adjust the weight of the scoring model. The formula of the adaptive moment estimation optimizer is: Among them, θ represents the weight parameter in the scoring model, ρ represents the learning rate, and m t and v t represents the first-order and second-order momentum, and ε represents a random number ranging from 0 to 1.

10. The present invention also provides a multi-dimensional talent evaluation system based on artificial intelligence technology, comprising the following modules: A data collection module is used to collect the resume text data and interview video data of the test subject; A preprocessing module, which is used to preprocess the resume text data and the interview video data to obtain a first feature sequence and a second feature sequence respectively; A scoring module, which is used to construct a scoring model based on the first feature sequence and the second feature sequence to output a score of the subject; The training and optimization module is used to train and optimize the scoring model.

Images