Artificial intelligence-based recruitment interview evaluation method and system, and storage medium

Abstract

The application discloses a recruitment interview evaluation method and system based on artificial intelligence and a storage medium, relates to the technical field of recruitment interview evaluation, and solves the technical problem that the prior art only completes morphological analysis and pragmatic reasoning through voice recognition and corpus comparison, does not deeply mine dynamic characteristics of voice signals, and does not perform correlation analysis on personality descriptions in resumes and real-time emotional performances in interviews, so that the evaluation dimension is single, and it is difficult to comprehensively and objectively reflect the language expression ability, psychological state and personality matching degree of a job seeker; the application obtains voice data and emotional data of a job seeker; the voice data is preprocessed to obtain a formant sequence corresponding to the voice data; a voice evaluation coefficient is calculated based on the formant sequence; an emotional similarity is calculated based on the emotional data; an interview evaluation coefficient is calculated based on the voice evaluation coefficient and the emotional similarity; and whether the job seeker is qualified is judged based on the interview evaluation coefficient, so that the above technical problem is solved.

Description

TECHNICAL FIELD

[0001] The present application belongs to the field of recruitment interview evaluation, and specifically relates to a recruitment interview evaluation method and system based on artificial intelligence and a storage medium. BACKGROUND

[0002] As a key link of enterprise talent selection, the accuracy and fairness of recruitment interview evaluation are crucial. Traditional recruitment interview evaluation methods mainly rely on the subjective judgment of interviewers, and have many shortcomings. For example, interviewers may be influenced by personal biases, emotional states, fatigue levels, and other factors, resulting in less objective evaluation results. At the same time, traditional methods are difficult to comprehensively and timely record and analyze various performances of candidates during the interview process, such as language expression, body language, emotional changes, etc., making the evaluation lack depth and comprehensiveness. In addition, with the expansion of recruitment scale, the low efficiency of traditional methods is increasingly prominent.

[0003] The prior art (invention patent with application number 2021108786489) discloses an intelligent interview method based on expression analysis, including: interview start; real-time voice data collection; the server performs voice recognition on the answer content of the job seeker to obtain text information; the server compares the text information with the vocabulary information in the corpus to complete morphological analysis; the server completes pragmatic reasoning according to the morphological analysis result; the server generates an evaluation result according to the morphological analysis and pragmatic reasoning results; the server pushes the evaluation result to the employer in real time. The invention patent realizes a practical and targeted talent evaluation interview full-process product, so that the employer can screen talents matched with the recruitment position, and each job seeker can get a more fair and objective personality evaluation. In addition, not only does the interview complete the examination of the job seeker in the traditional interview, but also accesses the data processing process, making the personality evaluation result more fair, objective and accurate, and improving the efficiency of enterprise talent selection. However, the prior art only completes morphological analysis and pragmatic reasoning by comparing voice recognition with the corpus, does not deeply mine the dynamic characteristics of voice signals (such as the fractal dimension of the formant sequence, Lyapunov exponent, etc. nonlinear dynamic characteristics), and does not perform correlation analysis on the personality description in the resume and the real-time emotional performance in the interview, resulting in a single evaluation dimension and difficulty in comprehensively and objectively reflecting the language expression ability, psychological state and personality matching degree of the job applicant.

[0004] Therefore, the present application proposes a recruitment interview evaluation method based on artificial intelligence to solve the above problems. SUMMARY

[0005] The present application aims to solve at least one of the technical problems existing in the prior art; for this purpose, the present application proposes an artificial intelligence-based recruitment interview evaluation method and system and a storage medium, to solve the technical problem that the prior art only completes morphological analysis and pragmatic reasoning through voice recognition and corpus comparison, does not deeply mine the dynamic characteristics of voice signals, and does not perform correlation analysis on personality descriptions in resumes and real-time emotional performance in interviews, resulting in a single evaluation dimension and difficulty in comprehensively and objectively reflecting the language expression ability, psychological state and personality matching degree of job applicants.

[0006] To achieve the above-mentioned purpose, the first aspect of the present application provides an artificial intelligence-based recruitment interview evaluation method, comprising:

[0007] obtaining voice data and emotional data of the job applicant; wherein the emotional data comprises resume personality data and interview emotional data;

[0008] obtaining voice data and emotional data of the job applicant; wherein the emotional data comprises resume personality data and interview emotional data;

[0009] calculating a voice evaluation coefficient based on the formant sequence;

[0010] calculating an emotional similarity based on the emotional data;

[0011] calculating an interview evaluation coefficient based on the voice evaluation coefficient and the emotional similarity;

[0012] determining whether the job applicant is qualified based on the interview evaluation coefficient.

[0013] Preferably, the voice data and emotional data of the job applicant are obtained by:

[0014] collecting voice data and image data of the job applicant and resume personality data through a data collection device; wherein the voice data refers to the voice of the job applicant during the interview; the image data refers to a plurality of images of the job applicant during the interview;

[0015] obtaining interview emotional data by analyzing the image data.

[0016] It should be noted that the resume personality data refers to the personality type of the job applicant in the resume or the personality type of the job applicant when the interviewer asks about the personality of the job applicant during the interview; the interview emotional data refers to the emotional type exhibited by the job applicant during the interview.

[0017] Preferably, the analysis of the image data comprises:

[0018] using the FER2013 dataset as training data;

[0019] extracting face images and corresponding emotions in the FER2013 dataset;

[0020] The face image is integrated as standard input data, and the face image and the corresponding emotion type are integrated as standard output data;

[0021] The artificial intelligence model is trained based on the standard input data and the standard output data to obtain an emotion recognition model; wherein the artificial intelligence model comprises a convolutional neural network or a deep belief network;

[0022] The image data is recognized based on the emotion recognition model to obtain interview emotion data; wherein the interview emotion data refers to the emotion type corresponding to the image data; the emotion type includes confidence, nervousness, happiness, fear, etc.

[0023] It should be noted that the FER2013 dataset is a dataset for facial expression recognition, which is provided by researchers at the University of California, Los Angeles and the International Face Recognition Competition (FERET), and contains 35887 images, each image being 48x48 pixels in size.

[0024] Preferably, the resonance peak sequence corresponding to the voice data is obtained by preprocessing the voice data, comprising:

[0025] The voice data is converted into a digital signal;

[0026] The digital signal corresponding to each frame of voice data is subjected to a fast Fourier transform, and a resonance peak estimation method is used to obtain a resonance peak sequence corresponding to the voice data; wherein the resonance peak estimation method includes linear predictive cepstral coefficient method, autocorrelation method or cepstral filtering method.

[0027] Preferably, the voice evaluation coefficient is calculated based on the resonance peak sequence, comprising:

[0028] The resonance peak sequence is analyzed by a box dimension algorithm to obtain a D value corresponding to the voice data;

[0029] The resonance peak sequence is analyzed by a Wolf algorithm to obtain a maximum Lyapunov exponent corresponding to the voice data;

[0030] The D value is marked as D, and the maximum Lyapunov exponent is marked as L;

[0031] The voice evaluation coefficient is calculated by the formula P=A1*ln(D+1)*tanh(D)+A2 / e^L; wherein P is the voice evaluation coefficient, and A1 and A2 are weight coefficients.

[0032] It should be noted that the weight coefficients A1 and A2 are set by those skilled in the art based on actual experience;

[0033] In the voice analysis of the candidate, the larger the D value (fractal dimension), the stronger the dynamic complexity and structural richness of the voice signal; a high D value means that the time domain waveform or frequency domain features (such as the formant trajectory) of the voice present more complex fluctuation patterns - this corresponds to the candidate's varied intonation and natural rhythm changes when expressing, even with the diversity of spectral structure brought by emotional involvement; for example, a confident and composed candidate's voice is often more rhythmic when expressing an opinion due to the fullness of emotion, and this complexity is quantified by a larger D value; on the contrary, nervous or monotonous speech has a smaller D value due to rigid rhythm and flat intonation; therefore, a larger D value indicates that the candidate has strong language expression ability and positive emotional transmission, which is a positive indicator in interview evaluation;

[0034] In the interview situation, if the candidate's voice has a smaller maximum Lyapunov exponent, it means that the dynamic trajectory of the voice system converges strongly in the phase space, i.e. the intonation, speed and rhythm of the speech are more stable, and there are fewer irregular changes caused by nervousness and anxiety; for example, a calm and clear-minded candidate has a low divergence rate of adjacent trajectories of the voice signal, corresponding to a negative value of the exponent, which reflects the coherence of expression and the stability of the state of mind; while a candidate with large emotional fluctuations or nervousness, the voice dynamics system is more susceptible to initial condition sensitivity dependence, resulting in an exponent tending to positive, which is manifested as sudden changes in intonation or rhythm disorder; therefore, a smaller maximum Lyapunov exponent is usually considered as a sign of a stable psychological state and a calm expression of the candidate, which is more advantageous in interview evaluation.

[0035] Preferably, the emotional similarity is calculated based on the emotional data, comprising:

[0036] The key words in the resume personality data and the interview emotional data are represented as fixed-dimensional vectors through a word vector model, obtaining resume personality word vector data and interview emotional word vector data; wherein the word vector model includes Word2Vec or GloVe;

[0037] The similarity between each word vector in the resume personality word vector data and each word vector in the interview emotional word vector data is calculated by a similarity algorithm, and the average value is taken to obtain the emotional similarity; wherein the similarity algorithm includes: cosine similarity, Pearson correlation coefficient or Jaccard similarity.

[0038] Preferably, the interview evaluation coefficient is calculated based on the voice evaluation coefficient and the emotional similarity, comprising:

[0039] The voice evaluation coefficient P is extracted, and the emotional similarity is marked as Y;

[0040] The interview evaluation coefficient is calculated by the formula MS=alpha*P / (P+1)+beta*e^tanh(Y), wherein MS is the interview evaluation coefficient, alpha and beta are weight coefficients.

[0041] It should be noted that the weight coefficients alpha and beta are set by the person skilled in the art according to actual experience.

[0042] Preferably, the method further comprises:

[0043] determining whether the interview evaluation coefficient is greater than a preset interview evaluation threshold value; if yes, determining that the candidate is qualified; and if no, determining that the candidate is unqualified.

[0044] It should be noted that the preset interview evaluation threshold value is set by the person skilled in the art according to actual experience.

[0045] The second aspect of the present application provides an artificial intelligence-based recruitment interview evaluation system, comprising a data acquisition module, a data analysis module and an interview evaluation module.

[0046] The data acquisition module is used to acquire the voice data and the emotional data of the candidate.

[0047] The data analysis module is used to obtain a formant sequence corresponding to the voice data by pre-processing the voice data, calculate a voice evaluation coefficient based on the formant sequence, calculate an emotional similarity based on the emotional data, and

[0048] The interview evaluation module is used to determine whether the candidate is qualified based on the interview evaluation coefficient.

[0049] The interview evaluation module is used to determine whether the candidate is qualified based on the interview evaluation coefficient.

[0050] Preferably, the data analysis module is in communication and / or electrical connection with the data acquisition module and the interview evaluation module, respectively.

[0051] The third aspect of the present application provides a storage medium for artificial intelligence-based recruitment interview evaluation, wherein the storage medium stores computer program instructions, and the program instructions are executed by a processor to realize the above method steps.

[0052] Compared with the prior art, the present application has the following advantages:

[0053] This invention acquires applicant speech data and multimodal sentiment data including resume personality data and interview sentiment data. The speech data is preprocessed to obtain formant sequences, and the fractal dimension (D-value) and maximum Lyapunov exponent are calculated using the box-counting algorithm and the Wolf algorithm. Speech evaluation coefficients are calculated to quantify the dynamic complexity of the speech signal and the stability of the vocal system, overcoming the deficiency of existing technologies in exploring the nonlinear dynamic features of speech. Simultaneously, a word vector model is used to transform resume personality and interview sentiment keywords into vectors, and a similarity algorithm is used to calculate sentiment similarity, enabling correlation analysis between resume personality descriptions and real-time interview sentiment performance, thus solving the problems of existing technologies. The problem of limited technical evaluation dimensions was addressed by constructing an interview evaluation coefficient based on speech evaluation coefficients and emotional similarity. This coefficient, combined with preset thresholds, determines the applicant's suitability, thereby comprehensively and objectively reflecting the applicant's language expression ability, psychological state, and personality fit. This improves the accuracy and fairness of recruitment interview evaluations and solves the technical problem that existing technologies only perform lexical analysis and pragmatic reasoning by comparing speech recognition with a corpus, without deeply exploring the dynamic features of speech signals or correlating personality descriptions in resumes with real-time emotional performance during interviews. This results in a limited evaluation dimension, making it difficult to comprehensively and objectively reflect the applicant's language expression ability, psychological state, and personality fit. Attached Figure Description

[0054] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0055] Figure 1 This is a schematic diagram of the method steps in an embodiment of the present invention;

[0056] Figure 2 This is a schematic diagram illustrating the applicant qualification assessment method according to an embodiment of the present invention.

[0057] Figure 3 This is a schematic diagram of the system modules in an embodiment of the present invention. Detailed Implementation

[0058] The technical solution of the present invention will be clearly and completely described below with reference to the embodiments. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0059] Please see Figure 1The embodiment of the first aspect of the application provides a recruitment interview evaluation method based on artificial intelligence, which comprises the following steps:

[0060] Obtaining voice data and emotion data of a job applicant; wherein the emotion data comprises resume character data and interview emotion data;

[0061] Obtaining voice data and emotion data of a job applicant; wherein the emotion data comprises resume character data and interview emotion data;

[0062] The voice data is preprocessed to obtain a formant sequence corresponding to the voice data;

[0063] The voice data is preprocessed to obtain a formant sequence corresponding to the voice data;

[0064] The voice data is preprocessed to obtain a formant sequence corresponding to the voice data;

[0065] The voice data is preprocessed to obtain a formant sequence corresponding to the voice data;

[0066] Obtaining voice data and emotion data of a job applicant; wherein the emotion data comprises resume character data and interview emotion data;

[0067] The voice data and image data of the job applicant and the resume character data are collected through a data collection device; wherein the voice data refers to the voice of the job applicant during the interview; and the image data refers to a plurality of images of the job applicant during the interview;

[0068] The voice data and image data of the job applicant and the resume character data are collected through a data collection device; wherein the voice data refers to the voice of the job applicant during the interview; and the image data refers to a plurality of images of the job applicant during the interview;

[0069] The voice data and image data of the job applicant and the resume character data are collected through a data collection device; wherein the voice data refers to the voice of the job applicant during the interview; and the image data refers to a plurality of images of the job applicant during the interview;

[0070] The FER2013 data set is used as training data;

[0071] The face images in the FER2013 data set and the corresponding emotions are extracted;

[0072] The face images are integrated into standard input data, and the face images and the corresponding emotion types are integrated into standard output data;

[0073] An artificial intelligence model is trained based on the standard input data and the standard output data to obtain an emotion recognition model; wherein the artificial intelligence model comprises a convolutional neural network or a deep belief network;

[0074] The image data is recognized based on the emotion recognition model to obtain interview emotion data; wherein the interview emotion data refers to the emotion type corresponding to the image data; and the emotion type comprises confidence, nervousness, happiness, fear, etc.

[0075] The voice data is preprocessed to obtain a formant sequence corresponding to the voice data, comprising:

[0076] The voice data is converted into a digital signal;

[0077] Performing fast Fourier transform on the digital signal corresponding to each frame of voice data, and obtaining a formant sequence corresponding to the voice data through a formant estimation method; wherein the formant estimation method includes: linear prediction cepstral coefficient method, autocorrelation method or cepstral filtering method.

[0078] Calculating a voice evaluation coefficient based on the formant sequence, including:

[0079] Analyzing the formant sequence through a box dimension algorithm to obtain a D value corresponding to the voice data;

[0080] Analyzing the formant sequence through a Wolf algorithm to obtain a maximum Lyapunov exponent corresponding to the voice data;

[0081] Marking the D value as D and the maximum Lyapunov exponent as L;

[0082] Calculating the voice evaluation coefficient through a formula P=A1*ln(D+1)*tanh(D)+A2 / e^L; wherein P is the voice evaluation coefficient, and A1 and A2 are weight coefficients.

[0083] Calculating a sentiment similarity based on sentiment data, including:

[0084] Expressing keywords in resume personality data and interview sentiment data as fixed-dimension vectors through a word vector model to obtain resume personality word vector data and interview sentiment word vector data; wherein the word vector model includes Word2Vec or GloVe;

[0085] Calculating the similarity between each word vector in the resume personality word vector data and each word vector in the interview sentiment word vector data through a similarity algorithm, and taking the average to obtain the sentiment similarity; wherein the similarity algorithm includes: cosine similarity, Pearson correlation coefficient or Jaccard similarity.

[0086] Calculating an interview evaluation coefficient based on the voice evaluation coefficient and the sentiment similarity, including:

[0087] Extracting the voice evaluation coefficient P and marking the sentiment similarity as Y;

[0088] Calculating the interview evaluation coefficient through a formula MS=α*P / (P+1)+β*e^tanh(Y); wherein MS is the interview evaluation coefficient, and α and β are weight coefficients.

[0089] Referring to Figure 2 , judging whether the job applicant is qualified based on the interview evaluation coefficient, including:

[0090] If the interview evaluation coefficient is greater than the preset interview evaluation threshold, the candidate is determined to be qualified; otherwise, the candidate is determined to be unqualified.

[0091] For example: in the interview of a software engineer position in an Internet company, the 2-minute voice and image data of the candidate when answering the question "Please describe the distributed system optimization project you led", and the character description in the resume is "confident, logical and rigorous, and team cooperation".

[0092] 1. Voice data collection:

[0093] The sampling rate is 16 kHz, 16-bit quantization, noise reduction microphone collection, and the frame length is 25 ms and the frame shift is 10 ms, a total of 480 frames of voice data.

[0094] 2. Image data collection:

[0095] 720p resolution, 30fps camera recording, and 100 clear face images are extracted.

[0096] 3. Resume character data:

[0097] The keywords are "confidence", "logical rigor", and "team cooperation".

[0098] 4. Interview emotion data extraction:

[0099] The CNN model (ResNet-18) trained based on the FER2013 dataset recognizes facial expressions, and the results are "confidence (probability 0.75)" and "focus (0.68)".

[0100] II. Voice evaluation coefficient calculation;

[0101] 1. Formant sequence extraction;

[0102] Preprocessing: convert the voice to a digital signal and perform pre-emphasis operation, and perform fast Fourier transform on each frame of voice.

[0103] Formant estimation (linear predictive cepstrum coefficient method, order p=12):

[0104] Extract the F2 formant sequence (unit: Hz) of each frame, and the first 5 frame data are: 1300, 1350, 1400, 1380, and 1420.

[0105] 2. Fractal dimension (D value) calculation;

[0106] Phase space reconstruction: embedding dimension m=7, time delay τ=12, and 7-dimensional trajectory points are generated.

[0107] Box dimension algorithm: Take δ=0.05σ (σ=80Hz), calculate the number of grids of the coverage trajectory at different scales, and obtain D=1.82 by linear fitting (which meets the excellent candidate range of 1.7-1.9).

[0108] It should be noted that σ here represents the standard deviation of formant frequencies, a statistical measure of the frequency fluctuation range of a speech signal. In the example, σ = 80Hz means that the standard deviation of the formant frequencies (such as F1, F2, etc.) of the applicant's speech in the time domain is 80Hz. δ is the grid scale (or box side length) used in the box-counting algorithm to cover the phase space trajectory. In the example, δ = 0.05σ = 0.05 × 80Hz = 4Hz, which means that the grid is divided into frequency axes in units of 4Hz, used to count the number of boxes required to cover the formant trajectory.

[0109] 3. Calculation of the maximum Lyapunov exponent (L);

[0110] Wolf algorithm: Tracks the evolution of the distance between adjacent trajectory points, with a time window of T=100ms, and calculates L=0.23 (L<0.3, indicating that the speech dynamics system is stable).

[0111] 4. Calculation of speech evaluation coefficient P;

[0112] Assume the weighting coefficients A1 = 0.6 and A2 = 0.4;

[0113] The speech evaluation coefficients are calculated using the formula P=A1×ln(D+1)×tanh(D)+A2 / e^L;

[0114] Substituting the data, we get P≈0.921.

[0115] III. Calculation of Emotional Similarity;

[0116] 1. Word vector conversion;

[0117] Personality word vectors in resumes: using Word2Vec (skip-gram model, 100 dimensions), the vector for "confidence" is V1, "logical rigor" is V2, and "teamwork" is V3.

[0118] Interview emotional word vectors: "confidence" is W1, "focus" is W2.

[0119] 2. Cosine similarity calculation;

[0120] Calculate the cosine similarity of each pair of word vectors:

[0121] sim(V1,W1)=0.85, sim(V1,W2)=0.62;

[0122] sim(V2, W1) = 0.58, sim(V2, W2) = 0.71;

[0123] sim(V3, W1) = 0.49, sim(V3, W2) = 0.36;

[0124] Take the average value:

[0125] Y = (0.85 + 0.62 + 0.58 + 0.71 + 0.49 + 0.36) / 6 ≈ 0.602.

[0126] Four, interview evaluation coefficient calculation and judgment;

[0127] 1. Interview evaluation coefficient MS calculation:

[0128] Assume the weight coefficient α = 0.7, β = 0.3;

[0129] Calculate the interview evaluation coefficient by the formula MS = α × ln (e^(P / (P+1))) + β × e^tanh(Y);

[0130] Substitute the data to get: MS ≈ 0.855.

[0131] 2. Eligibility judgment:

[0132] Assume the preset threshold is 0.8, because 0.855 > 0.8, it is determined that the candidate is qualified.

[0133] Five, result analysis;

[0134] The speech evaluation coefficient P = 0.921: the D value 1.82 indicates that the candidate's language expression rhythm is rich (such as natural fluctuation of intonation when explaining technical details), and L = 0.23 reflects the candidate's stable state of mind (no obvious intonation mutation caused by tension).

[0135] Emotional similarity Y = 0.602: the key word matching degree of the resume character and the interview emotion is high, especially the similarity of "self-confidence" reaches 0.85, which verifies the consistency of the candidate's self-description and actual performance.

[0136] Interview evaluation coefficient MS = 0.855: comprehensive reflects the candidate's language expression ability and character matching degree, which exceeds the threshold and meets the requirements of the software engineer position for clear logic and stable state of mind.

[0137] This example shows how the present application breaks through the limitations of the prior art from two aspects of speech dynamic characteristics and emotional matching degree through full-process quantitative analysis, and realizes more objective interview evaluation.

[0138] Reference Figure 3The second aspect embodiment of the present application provides an artificial intelligence-based recruitment interview evaluation system, comprising a data acquisition module, a data analysis module and an interview evaluation module.

[0139] The data acquisition module is used to acquire the voice data and the emotional data of the job applicant.

[0140] The data analysis module is used to obtain the formant sequence corresponding to the voice data by preprocessing the voice data, calculate the voice evaluation coefficient based on the formant sequence, calculate the emotional similarity based on the emotional data, and the like.

[0141] The interview evaluation module is used to judge whether the job applicant is qualified based on the interview evaluation coefficient.

[0142] The interview evaluation module is used to judge whether the job applicant is qualified based on the interview evaluation coefficient.

[0143] Some data in the above formula are calculated by removing the dimension, and the formula is obtained by software simulation of a large amount of collected data to be closest to the real situation; the preset parameters and the preset threshold in the formula are set by the person skilled in the art according to the actual situation or obtained by a large amount of data simulation.

[0144] Working principle of the present application:

[0145] The present application acquires the voice data and the emotional data of the job applicant, obtains the formant sequence corresponding to the voice data by preprocessing the voice data, calculates the voice evaluation coefficient based on the formant sequence, calculates the emotional similarity based on the emotional data, calculates the interview evaluation coefficient based on the voice evaluation coefficient and the emotional similarity, and judges whether the job applicant is qualified based on the interview evaluation coefficient, thereby solving the technical problem that the prior art only completes morphological analysis and pragmatic reasoning by comparing the voice recognition with the corpus library, does not deeply mine the dynamic characteristics of the voice signal, and does not perform correlation analysis on the personality description in the resume and the real-time emotional performance in the interview, resulting in single evaluation dimension and difficulty in comprehensively and objectively reflecting the language expression ability, psychological state and personality matching degree of the job applicant.

[0146] The above embodiments are only used to illustrate the technical method of the present application but not limit the present application, and although the present application has been described in detail with reference to the preferred embodiments, it should be understood by those skilled in the art that the technical method of the present application can be modified or replaced equivalently without departing from the spirit and scope of the technical method of the present application.

Claims

1. An artificial intelligence-based recruitment interview evaluation method, characterized in that, include: Acquire applicants' voice data and emotional data; the emotional data includes: resume personality data and interview emotional data; By preprocessing the speech data, the formant sequence corresponding to the speech data is obtained; Speech evaluation coefficients are calculated based on formant sequences; Emotional similarity is calculated based on sentiment data; The interview evaluation coefficient is calculated based on the voice evaluation coefficient and the emotional similarity coefficient; The suitability of an applicant is determined based on an interview evaluation coefficient. The speech evaluation coefficients calculated based on formant sequences include: The formant sequence is analyzed using the box-dimensional algorithm to obtain the D value corresponding to the speech data; The Wolf algorithm was used to analyze the formant sequences to obtain the maximum Lyapunov exponent corresponding to the speech data. Label the D value as D, and the maximum Lyapunov exponent as L; The speech evaluation coefficients are calculated using the formula P=A1×ln(D+1)×tanh(D)+A2 / e^L, where P is the speech evaluation coefficient and A1 and A2 are weighting coefficients. The emotional similarity calculated based on emotional data includes: By using a word vector model, keywords in resume personality data and interview sentiment data are represented as fixed-dimensional vectors, resulting in resume personality word vector data and interview sentiment word vector data. The similarity between each word vector in the resume personality word vector data and each word vector in the interview sentiment word vector data is calculated using a similarity algorithm, and the average value is taken to obtain the sentiment similarity. The interview evaluation coefficient, calculated based on voice evaluation coefficient and emotional similarity, includes: Extract the speech evaluation coefficient P and label the emotional similarity as Y; The interview evaluation coefficient is calculated using the formula MS=α×P / (P+1)+β×e^tanh(Y); where MS is the interview evaluation coefficient, and α and β are weighting coefficients.

2. The artificial intelligence-based recruitment interview evaluation method according to claim 1, characterized in that, The acquisition of applicants' voice and emotional data includes: The data collection equipment collects the applicant's voice data, image data, and personality data from their resume; the voice data refers to the applicant's voice responses during the interview; the image data refers to several images of the applicant during the interview. Interview emotion data was obtained by analyzing the image data.

3. The artificial intelligence-based recruitment interview evaluation method according to claim 2, characterized in that, The analysis of image data includes: Use the FER2013 dataset as training data; Extract face images and corresponding emotions from the FER2013 dataset; Facial images are integrated into standard input data, and facial images and corresponding emotion types are integrated into standard output data. An artificial intelligence model is trained based on standard input and standard output data to obtain an emotion recognition model; the artificial intelligence model includes: convolutional neural network or deep belief network; Based on the emotion recognition model, image data is identified to obtain interview emotion data; where interview emotion data refers to the emotion type corresponding to the image data.

4. The artificial intelligence-based recruitment interview evaluation method according to claim 1, characterized in that, The process of preprocessing the speech data to obtain the formant sequence corresponding to the speech data includes: Convert voice data into digital signals; A fast Fourier transform is performed on the digital signal corresponding to each frame of speech data, and the formant sequence corresponding to the speech data is obtained through formant estimation methods. Among them, the formant estimation methods include: linear predictive cepstral coefficient method, autocorrelation method or cepstral filtering method.

5. The artificial intelligence-based recruitment interview evaluation method according to claim 1, characterized in that, The emotional similarity calculated based on emotional data includes: Keywords in resume personality data and interview sentiment data are represented as fixed-dimensional vectors using word vector models, resulting in resume personality word vector data and interview sentiment word vector data; the word vector models include Word2Vec or GloVe. The similarity between each word vector in the resume personality word vector data and each word vector in the interview sentiment word vector data is calculated using a similarity algorithm, and the average value is taken to obtain the sentiment similarity. The similarity algorithms include: cosine similarity, Pearson correlation coefficient or Jaccard similarity.

6. The artificial intelligence-based recruitment interview evaluation method according to claim 1, characterized in that, The method of determining whether an applicant is qualified based on an interview evaluation coefficient includes: Determine if the interview evaluation coefficient is greater than the preset interview evaluation threshold; if yes, the applicant is deemed qualified; otherwise, the applicant is deemed unqualified.

7. An AI-based recruitment interview evaluation system, executing the AI-based recruitment interview evaluation method according to any one of claims 1-6, characterized in that, include: Data acquisition module, data analysis module, and interview evaluation module; Data acquisition module: used to acquire applicants' voice and emotional data; Data analysis module: By preprocessing the speech data, the formant sequence corresponding to the speech data is obtained; Speech evaluation coefficients are calculated based on formant sequences; Emotional similarity is calculated based on sentiment data; as well as, The interview evaluation coefficient is calculated based on the voice evaluation coefficient and the emotional similarity coefficient; Interview evaluation module: Determines whether an applicant is qualified based on an interview evaluation coefficient.

8. A recruitment interview evaluation storage medium based on artificial intelligence, characterized in that, The storage medium stores computer program instructions that, when executed by a processor, implement the steps of the method described in any one of claims 1-6.

Images