Interview method and device

Abstract

The invention discloses an interview method and device, relates to the technical field of online interview, and mainly aims to adjust interview problems in real time according to the state of an interviewer. According to the main technical scheme, the method comprises the steps of collecting audio and video data of an interviewer in an interview process; when an interview question is issued to the interviewer terminal every time, the collected audio and video data are analyzed to obtain emotion feature data, language behavior feature data and other behavior feature data except the language behavior feature data of the interviewer responding to the current interview question; selecting a next interview question based on the emotion feature data, the language behavior feature data and other behavior feature data, and issuing the next interview question to the interviewer terminal; and if it is determined that the interview process of the interviewer is finished, generating an interview report of the interviewer based on the corresponding emotional feature data, language behavior feature data, other behavior feature data and answers of the interviewer responding to each interview question in the interview process.

Description

Technical Field

[0001] This application relates to the technical field of online interviews, and particularly to an interview method and device. Background Art

[0002] With the development of AI (Artificial Intelligence) technology, AI technology has been widely applied to interview scenarios. In an interview scenario, an interview system based on AI technology acts as an interviewer to ask questions to an interviewee, and automatically evaluates the interviewee based on the interviewee's responses to the questions.

[0003] In the existing interview method, the interview system usually selects some questions with a fixed order from a preset question bank and asks the interviewee in the fixed order. The inventor found that there are problems in the related technology: this question-and-answer mode in a fixed order makes the interview process seem mechanical and unable to flexibly adjust interview questions in real time according to the interviewee's state. Once the interviewee has excessive psychological pressure during the interview process, the interview may be interrupted and unable to continue.

[0004] Therefore, how to adjust interview questions in real time according to the interviewee's state has become an urgent problem to be solved currently. Summary of the Invention

[0005] This application proposes an interview method and device, and the main purpose is to adjust interview questions in real time according to the interviewee's state, so as to enable the interview to proceed smoothly and improve the interview effect.

[0006] To achieve the above purpose, this application mainly provides the following technical solutions:

[0007] In a first aspect, this application provides an interview method, and the interview method includes: during the interview process, collecting audio and video data of the interviewee; every time an interview question is sent to the interviewee's terminal, analyzing the collected audio and video data to obtain emotional characteristic data, language behavior characteristic data, and other behavior characteristic data other than the language behavior characteristic data of the interviewee's response to the current interview question, and selecting the next interview question based on the emotional characteristic data, the language behavior characteristic data, and the other behavior characteristic data and sending it to the interviewee's terminal; if it is determined that the interview process of the interviewee ends, generating an interview report of the interviewee based on the emotional characteristic data, language behavior characteristic data, other behavior characteristic data, and answers corresponding to each interview question answered by the interviewee during the interview process.

[0008] In a second aspect, this application provides an interview device, and the interview device may include:

[0009] A collection module, configured to collect audio and video data of the interviewee during the interview process;

[0010] A selection module, configured to analyze the collected audio and video data to obtain emotional characteristic data, language behavior characteristic data, and other behavior characteristic data other than the language behavior characteristic data of the interviewee's response to the current interview question every time an interview question is sent to the interviewee's terminal, and select the next interview question based on the emotional characteristic data, the language behavior characteristic data, and the other behavior characteristic data and send it to the interviewee's terminal;

[0011] A generation module, configured to generate an interview report of the interviewee based on the emotional characteristic data, language behavior characteristic data, other behavior characteristic data, and answers corresponding to each interview question answered by the interviewee during the interview process if it is determined that the interview process of the interviewee ends.

[0012] In a third aspect, the present application provides a computer-readable storage medium, where the storage medium includes a stored program, and when the program runs, it controls the device where the storage medium is located to execute the interview method described in the first aspect.

[0013] In a fourth aspect, the present application provides an electronic device, which includes: a memory for storing a program; a processor coupled to the memory for running the program to execute the interview method described in the first aspect.

[0014] In a fifth aspect, the present application provides a computer program product, including computer programs / instructions, and when the computer programs / instructions are executed by a processor, the above-mentioned interview method is implemented.

[0015] The interview method and device provided by this application collect the audio and video data of the interviewee during the interview process. Each time an interview question is sent to the interviewee's terminal, the collected audio and video data is analyzed to obtain the emotional characteristic data, language behavior characteristic data, and other behavior characteristic data (excluding language behavior characteristic data) of the interviewee's response to the current interview question. Then, based on the emotional characteristic data, language behavior characteristic data, and other behavior characteristic data, the next interview question is selected and sent to the interviewee's terminal. When it is determined that the interview process of the interviewee ends, an interview report of the interviewee is generated based on the emotional characteristic data, language behavior characteristic data, other behavior characteristic data, and answers corresponding to each interview question during the interview process. It can be seen that the solution provided in this embodiment can achieve the following effects: First, each time an interview question is sent to the interviewee's terminal during the interview process, the emotional characteristic data, language behavior characteristic data, and other behavior characteristic data of the interviewee's response to the current interview question are obtained by analyzing the audio and video data of the interviewee. Based on these characteristic data (which reflect the interviewee's mental state from the emotional state and real behavior of the interviewee), the next interview question is adaptively adjusted. The interview question adjusted by this adaptive adjustment mechanism can timely adjust the psychological pressure of the interviewee, thereby reducing the possibility of interview interruption caused by the increase in the interviewee's psychological pressure, realizing the humanization of the interview process, and ensuring the smooth completion of the interview. Second, after the interview ends, the characteristic data (which reflect the interviewee's mental state from the emotional state and real behavior of the interviewee) and the answers of the interviewee to the interview questions are used together as the basis for evaluating the interviewee in the interview report. In this way, the interviewee can be evaluated more comprehensively to assist in more accurately screening out the interviewees who are competent for the corresponding positions, thereby improving the interview accuracy rate.

[0016] The above description is only an overview of the technical solution of this application. In order to be able to understand the technical means of this application more clearly, it can be implemented according to the content of the specification. And in order to make the above and other purposes, features, and advantages of this application more obvious and understandable, the following specifically illustrates the specific implementation manners of this application. Brief Description of the Drawings

[0017] In order to more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the drawings in the following description are some embodiments of this application. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0018] Figure 1 The flowchart of an interview method provided by an embodiment of this application is shown;

[0019] Figure 2 It shows a schematic diagram of an interview process provided by an embodiment of the present application;

[0020] Figure 3 It shows a schematic structural diagram of an interview device provided by an embodiment of the present application;

[0021] Figure 4 It shows a schematic structural diagram of an interview device provided by another embodiment of the present application. Detailed implementation manners

[0022] Exemplary embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although the exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure can be implemented in various forms and should not be limited by the embodiments set forth herein. On the contrary, these embodiments are provided so that the present disclosure can be more thoroughly understood and the scope of the present disclosure can be fully conveyed to those skilled in the art.

[0023] In multiple embodiments of the detailed implementation manners, the audio-visual data, sensitive questions and other data involved in each embodiment are all authorized by the user or fully authorized by all parties, and the collection, use and processing of the relevant data all comply with the relevant laws, regulations and standards of the relevant countries and regions.

[0024] In the existing interview method, the interview system selects some questions with a fixed order from a preset question bank and asks the interviewees in a fixed order. This existing interview method is relatively rigid and cannot flexibly adjust the interview questions in real time according to the interviewees' states. Once the interviewees experience an increase in psychological pressure due to reasons such as being unable to answer questions, the interview may be interrupted and unable to continue. In addition, since it is impossible to obtain the emotional states and real behavioral responses of the interviewees during the interview process and only rely on the interviewees' responses to the questions to evaluate the interviewees, it is difficult to comprehensively evaluate the interviewees, resulting in poor interview effects.

[0025] After research, it is found that if during the interview process, each interview question is sent to the interviewee, and then by analyzing the interviewee's audio-visual data to determine the characteristic data that reflects the interviewee's psychological state from the interviewee's emotional state and real behavior when answering the current interview question, and adaptively adjusting the next interview question based on this characteristic data, the following effects can be achieved: First, based on the characteristic data that reflects the interviewee's psychological state from the interviewee's emotional state and real behavior when answering the current interview question, the next interview question is adaptively adjusted. The interview question adjusted by this adaptive adjustment mechanism can timely adjust the interviewee's psychological pressure, thereby reducing the possibility of interview interruption caused by the increase in the interviewee's psychological pressure, realizing the humanization of the interview process, and then ensuring the smooth completion of the interview. Second, after the interview, the characteristic data that reflects the interviewee's psychological state from the interviewee's emotional state and real behavior, together with the answers of the interviewee to the interview questions, are used as the basis for evaluating the interviewee. This can more comprehensively evaluate the interviewee to assist in more accurately screening out the interviewees who can be competent for the corresponding positions, thereby improving the interview accuracy rate.

[0026] Based on the above findings, the embodiment of the present application proposes an interview technical solution, which specifically includes: During the interview process, collect the interviewee's audio-visual data. Each time an interview question is sent to the interviewee, analyze the collected audio-visual data to obtain the emotional characteristic data, language behavior characteristic data, and other behavior characteristic data (excluding language behavior characteristic data) of the interviewee when answering the current interview question (these characteristic data at least include the characteristic data that reflects the interviewee's psychological state from the interviewee's emotional state and real behavior); select the next interview question based on the emotional characteristic data, language behavior characteristic data, and other behavior characteristic data and send it to the interviewee. If it is determined that the interview process of the interviewee ends, then based on the emotional characteristic data, language behavior characteristic data, other behavior characteristic data, and answers corresponding to each interview question answered by the interviewee during the interview process, generate an interview report of the interviewee.

[0027] The interview technical solution provided in the above embodiment can be applied to any interview scenario, and this embodiment does not limit the interview scenario. Exemplarily, the interview scenario can include but is not limited to: the interview scenario for enterprise and institution recruitment, the interview scenario for school enrollment. In addition, the interview system applying the interview technical solution provided in this embodiment can include but is not limited to: AI interview robots, AI interview cloud platforms, etc.

[0028] Based on the above interview technical solution, this embodiment specifically provides an interview method and device. The following specifically describes the interview method and device provided in this embodiment.

[0029] The embodiment of the present application provides an interview method, as Figure 1As shown, the interview method provided in this embodiment may at least include the following steps 101 to 103.

[0030] 101. During the interview process, collect the audio-visual data of the interviewee.

[0031] The interview method provided in this embodiment needs to collect the audio-visual data of the interviewee during the interview process, so as to obtain the characteristic data for reflecting the emotional state and real behavior of the interviewee when answering interview questions based on the audio-visual data, so as to adaptively adjust the subsequent interview questions based on these characteristic data. The audio-visual data refers to a data set containing the audio data and video data of the interviewee during the interview process.

[0032] The interview method provided in this embodiment is applied to an interview system. The information, data, and signals involved in the application of the interview method by the interview system are all authorized by the user or fully authorized by all parties, and the collection, use, and processing of relevant data comply with the relevant laws, regulations, and standards of relevant countries and regions. The interview system is communicatively connected to the audio-visual acquisition device. Based on this, the specific process of collecting the audio-visual data of the interviewee during the interview process may include the following steps: Detect whether the interviewee enters and stays in the acquisition area of the audio-visual acquisition device. If so, after determining that the interview start condition is met, call the audio-visual acquisition device to continuously collect the audio-visual data of the interviewee during the interview process. The method for determining whether the interview start condition is met may include at least one of the following:

[0033] One method is to monitor whether there is an interview start instruction. If an interview start instruction is monitored, it is determined that the interview start condition is met. The interview start instruction is an instruction generated after the interview start button is triggered. The interview start button can be triggered by the interviewee who enters the acquisition area or by the interview control personnel who observe the interviewee entering the acquisition area. In this way, the interviewee or the interview control personnel can flexibly control the start time of the interview based on the interview needs. In other embodiments, the interview can also be started after monitoring the interview start voice instruction or gesture instruction issued by the interviewee. For example: The interview system issues an inquiry voice of "Do you want to start the interview", and when it monitors the "Start the interview" voice answered by the interviewee or the OK gesture indicating no problem, the interview system starts the interview. The voice instruction and gesture instruction can be flexibly set according to needs and will not be elaborated here.

[0034] Another method is to monitor whether the cumulative duration of the interviewee staying continuously in the acquisition area is not less than the preset duration. If it is not less than, it is determined that the interview start condition is met. In this way, the interview can automatically start after the interviewee enters the acquisition area, thereby reducing the manual intervention in the interview.

[0035] 102. Each time an interview question is sent to the interviewee's terminal, the collected audio and video data is analyzed to obtain the emotional characteristic data, language behavior characteristic data, and other behavior characteristic data (excluding language behavior characteristic data) of the interviewee's response to the current interview question. Based on the emotional characteristic data, language behavior characteristic data, and other behavior characteristic data, the next interview question is selected and sent to the interviewee's terminal.

[0036] In some embodiments, during the interview process, the interviewee obtains interview questions from the interviewee's terminal to promote the interview by answering the interview questions. The interviewee's terminal is used to provide interview questions to the interviewee. The ways for the interviewee's terminal to provide interview questions to the interviewee can include the following three: First, the interviewee's terminal is an electronic terminal with at least a voice playback function, and it provides interview questions to the interviewee in the way of voice-playing the interview questions; Second, the interviewee's terminal is an electronic terminal with at least a display function, and it provides interview questions to the interviewee by displaying the interview questions on the screen; Third, the interviewee's terminal is an electronic terminal with at least a voice playback function and a display function, and it provides interview questions to the interviewee in the way of voice-playing the interview questions and by displaying the interview questions on the screen, so that the interviewee can more accurately obtain the interview questions from both hearing and vision. It should be noted that during the interview, the interviewee can flexibly select any of the above ways for the interviewee's terminal to provide interview questions based on their actual situation, and this embodiment does not make any restrictions on this.

[0037] In some embodiments, each time a current interview question is sent to the interviewee's terminal during the interview process, the collected audio and video data is analyzed to obtain the emotional characteristic data, language behavior characteristic data, and other behavior characteristic data of the interviewee's response to the current interview question, so as to adaptively adjust the next interview question based on these characteristic data (which reflect the interviewee's psychological state from the interviewee's emotional state and real behavior), and timely adjust the interviewee's psychological pressure through the interview questions adjusted by this adaptive adjustment mechanism, so as to reduce the possibility of interview interruption caused by the increase in the interviewee's psychological pressure, thus ensuring the smooth completion of the interview.

[0038] In some embodiments, the emotional characteristic data of the interviewee's response to the current interview question is used to reflect the emotion of the interviewee's response to the current interview question, and it reflects the mental state of the interviewee when answering the current interview question from the dimension of emotion. When evaluating the interviewee, based on the emotional characteristic data, the behavior and mental health status of the interviewee can be more accurately evaluated from the dimension of emotion, providing a more powerful screening basis for screening out interviewees who can be competent for the corresponding positions in the future, thereby improving the interview accuracy rate. Based on this, it is necessary to perform the step of analyzing the collected audio-visual data to obtain the emotional characteristic data of the interviewee's response to the current interview question. The specific process of this step may include the following steps 102A to 102C.

[0039] 102A. Invoke the first model to analyze and identify the audio data in the audio-visual data to obtain the voice emotional characteristic data of the interviewee when answering the current interview question.

[0040] The audio characteristics of the interviewee's audio data are related to the interviewee's emotion. Therefore, the voice emotional characteristic data of the interviewee when answering the current interview question can be obtained by analyzing the audio data in the audio-visual data. The voice emotional characteristic data may include, but is not limited to, at least one of the following: pitch, tone, volume, speech rate, timbre, intonation, rhythm, pause, formant, clarity, tremor, voice duration, voice intensity change, voice frequency range, and harmonic structure. The voice emotional characteristic data reflects the emotions of the interviewee such as happiness / joy, sadness, anger, fear, disgust, surprise, contempt, etc. through the content it includes.

[0041] The first model is used to identify the voice emotional characteristic data from the audio data. When determining the voice emotional characteristic data of the interviewee when answering the current interview question, the audio data in the audio-visual data is used as the input of the first model and input into the first model. The first model identifies the corresponding voice emotional characteristic data from the input audio data and outputs the identified voice emotional characteristic data. It should be noted that in this embodiment, the audio data used as the input of the first model can be any of the following: one is that the audio data used as the input of the first model is only the audio data of the interviewee when answering the current interview question; the other is that considering that a person's emotion has temporality, in order to more accurately determine the voice emotional characteristic data of the interviewee when answering the current interview question, the audio data used as the input of the first model includes the audio data of the interviewee when answering the current interview question and the audio data of the interviewee when answering the previous interview question.

[0042] The first model can be trained with multiple sets of data. Each set of data includes the audio data of a sample interviewee answering a sample interview question and the corresponding voice emotion feature data of the audio data. The specific type of the first model can be flexibly selected based on business needs, and this embodiment does not make any limitations in this regard. Exemplarily, the first model can select a voice analysis model that adopts a multi-level feature extraction strategy. This voice analysis model is used to obtain the voice emotion feature data by analyzing basic features such as the 13-dimensional MFCC coefficients, zero-crossing rate, and sound intensity of the audio data, prosodic features such as pitch, duration, and stress position, and emotional features such as the energy spectrum and harmonic-to-noise ratio of the audio data. In terms of algorithm implementation, the first model can perform noise reduction processing through wavelet transform, extract frequency domain features using Fourier transform, and use the LSTM-CTC model to implement audio data recognition, and use the CNN-Attention structure for emotion classification. The processing delay of such a first model is controlled within 100 milliseconds, effectively improving the determination efficiency of the voice emotion feature data.

[0043] 102B. Invoke the second model to analyze and recognize the first video frame corresponding to the face of the interviewee when answering the current interview question in the audio-visual data, and obtain the micro-expression emotion feature data of the interviewee when answering the current interview question.

[0044] Considering that the emotions of the interviewee are not only reflected in the voice but also in the micro-expressions. Based on this, this embodiment also obtains the first video frame corresponding to the face of the interviewee when answering the current interview question from the audio-visual data, and analyzes the first video frame by invoking the second model to obtain the micro-expression emotion feature data of the interviewee when answering the current interview question. The micro-expression emotion feature data can include, but is not limited to, at least one of the following contents: the shape and position change data of facial key points (the key points include at least one of the following: eyes, nose, mouth, eyebrows), facial texture feature data, and facial color feature data. The micro-expression emotion feature data reflects the emotions of the interviewee such as happiness / joy, sadness, anger, fear, disgust, surprise, contempt, etc. through the contents it includes.

[0045] The second model is used to identify micro-expression emotion feature data from video frames of a face. When determining the micro-expression emotion feature data of an interviewee in response to a current interview question, the first video frame corresponding to the interviewee's face when answering the current interview question is used as the input to the second model and input into the second model. The second model identifies the corresponding micro-expression emotion feature data from the input first video frame and outputs the identified micro-expression emotion feature data. It should be noted that in this embodiment, the first video frame used as the input to the second model can be any of the following: one is that the first video frame used as the input to the second model is only the video frame corresponding to the interviewee's face when answering the current interview question; the other is that considering that human emotions are time-sequential, in order to more accurately determine the micro-expression emotion feature data of the interviewee when answering the current interview question, the first video frame used as the input to the second model includes the video frame corresponding to the interviewee's face when answering the current interview question and the video frame corresponding to the interviewee's face when answering the previous interview question before answering the current interview question.

[0046] The second model is trained with multiple sets of data. Each set of data includes the video frame corresponding to the face of a sample interviewee when answering a sample interview question and the micro-expression emotion feature data corresponding to the video frame. The specific type of the second model can be flexibly selected based on business needs, and this embodiment does not make any limitations in this regard. Exemplarily, the second model can select a model using a deep learning architecture with ResNet-50 as the backbone network. This model combines an attention mechanism to achieve precise positioning of 68 key points (68 key points refer to the facial muscle action units defined in the Facial Action Coding System, FACS), and analyzes the sequence of expression changes through a temporal LSTM network to obtain micro-expression emotion feature data. This model can ensure that the recognition accuracy of seven basic expressions (the seven basic expressions include happiness / joy, sadness, anger, fear, disgust, surprise, contempt) reaches more than 95%, the average error of key point positioning is controlled within 3 pixels, the real-time processing delay is less than 50 milliseconds, and the recognition remains stable within a 30-degree deflection angle of the face. In addition, to ensure the recognition quality, this model has implemented preprocessing mechanisms such as adaptive light compensation for video frames, multi-scale face detection, and facial alignment and normalization.

[0047] 102C. Integrate the voice emotion feature data and the micro-expression emotion feature data to obtain the emotion feature data of the interviewee when answering the current interview question.

[0048] The emotional characteristics of the interviewee are reflected in the voice and micro-expressions. Based on this, in this embodiment, the voice emotional characteristic data and the micro-expression emotional characteristic data are fused to obtain the emotional characteristic data of the interviewee when answering the current interview question. In this way, through the fusion of the voice emotional characteristic data and the micro-expression emotional characteristic data, the obtained emotional characteristic data can be made closer to the real emotional state of the interviewee when answering the current interview question.

[0049] The fusion of the voice emotional characteristic data and the micro-expression emotional characteristic data essentially means using one modality (such as the emotional analysis result of the voice modality, "voice emotional characteristic data") to fuse another modality (such as the emotional analysis result of the micro-expression modality, "micro-expression emotional characteristic data") to achieve the complementarity of the two. This fusion can at least achieve the following two effects: (1) Different aspects of emotional characteristics may be captured from the two modalities of audio and micro-expressions. For example, audio may better capture the intonation changes, while micro-expressions may better capture the subtle emotional changes on the face. In this way, by fusing these two types of data, namely the voice emotional characteristic data and the micro-expression emotional characteristic data, more comprehensive emotional characteristic data of the interviewee can be obtained. (2) There are limitations in using either the audio modality or the micro-expression modality alone for emotional analysis. For example, audio analysis may be inaccurate in a noisy environment, or facial expression analysis may be limited when part of the face is blocked. Through fusion, the limitations of a single modality can be reduced, thereby improving the accuracy of the interviewee's emotional characteristic data.

[0050] The specific process of fusing the voice emotional characteristic data and the micro-expression emotional characteristic data to obtain the emotional characteristic data of the interviewee when answering the current interview question may include: projecting the voice emotional characteristic data into a first vector in the first preset vector space, projecting the micro-expression emotional characteristic data into a second vector in the first preset vector space, and based on the first vector and the second vector, obtaining a third vector through a preset first algorithm; based on the correspondence relationship between the preset vector and the emotional characteristic data, determining the target emotional characteristic data corresponding to the third vector, and determining the target emotional characteristic data as the emotional characteristic data of the interviewee when answering the current interview question. It should be noted that the first preset vector space is a vector space used to determine the emotional characteristic data of the interviewee. The preset first algorithm can be flexibly determined based on business needs, and this embodiment does not make any limitations in this regard. Exemplarily, the preset first algorithm may include, but is not limited to, at least one of the following operations: addition operation, multiplication operation.

[0051] Further, if there is a large difference between the voice emotion feature data and the micro-expression emotion feature data, it indicates that there is an abnormality in the emotion analysis of the audio modality or the micro-expression modality. In this case, the emotion feature data of the interviewee obtained by fusing the voice emotion feature data and the micro-expression emotion feature data is probably inaccurate. Based on this, in order to improve the accuracy of the emotion feature data, before fusing the voice emotion feature data and the micro-expression emotion feature data to obtain the emotion feature data of the interviewee when answering the current interview question, the following steps may also be included: If it is determined that the difference between the voice emotion feature data and the micro-expression emotion feature data is large, then stop executing the step of fusing the voice emotion feature data and the micro-expression emotion feature data to obtain the emotion feature data of the interviewee when answering the current interview question, and analyze the reason for the large difference, and conduct corresponding anomaly detection based on the reason until the difference between the re-obtained voice emotion feature data and the micro-expression emotion feature data after the anomaly detection meets the requirements, then continue to execute the step of fusing the voice emotion feature data and the micro-expression emotion feature data to obtain the emotion feature data of the interviewee when answering the current interview question.

[0052] When evaluating the difference between the voice emotion feature data and the micro-expression emotion feature data, determine the first emotion information of the interviewee based on the voice emotion feature data, determine the second emotion information of the interviewee based on the micro-expression emotion feature data, and determine the emotion difference information between the first emotion indicated by the first emotion information and the second emotion indicated by the second emotion information. The emotion difference information is used to indicate the difference degree between the first emotion and the second emotion. The greater the difference degree, the greater the difference between the first emotion and the second emotion; if it is determined that the difference degree indicated by the emotion difference information is not less than the difference threshold, then it is determined that the difference between the voice emotion feature data and the micro-expression emotion feature data is large. If it is determined that the difference degree indicated by the emotion difference information is less than the difference threshold, then it is determined that the difference between the voice emotion feature data and the micro-expression emotion feature data is not large. It should be noted that the process of determining the emotion difference information between the first emotion indicated by the first emotion information and the second emotion indicated by the second emotion information may include: taking the first emotion indicated by the first emotion information and the second emotion indicated by the second emotion information as model inputs and inputting them into the emotion difference information evaluation model. The emotion difference information evaluation model evaluates the corresponding emotion difference information based on the model inputs and outputs it. The emotion difference information evaluation model is trained based on multiple groups of data. Each group of data includes the first sample emotion indicated by the first sample emotion information and the second sample emotion indicated by the second sample emotion information, as well as the emotion difference information corresponding to the first sample emotion indicated by the first sample emotion information and the second sample emotion indicated by the second sample emotion information. The specific type of the emotion difference information evaluation model can be flexibly selected based on business needs, and this embodiment does not make any limitations in this regard.

[0053] In some embodiments, the language behavior feature data is used to reflect the language behavior of the interviewee in answering the current interview question, and it reflects the mental state of the interviewee in answering the current interview question from the dimension of language behavior. When evaluating the interviewee, based on the language behavior feature data, the behavior and mental health status of the interviewee can be more accurately evaluated from the dimension of language behavior, so as to provide a more powerful screening basis for screening out interviewees who can be competent for the corresponding positions in the follow-up, thereby improving the interview accuracy rate. Based on this, it is necessary to perform the step of analyzing the collected audio-visual data to obtain the language behavior feature data of the interviewee in answering the current interview question, and the specific process of this step can include the following step 102D.

[0054] 102D. Invoke the third model to analyze and identify the audio data in the audio-visual data to obtain the language behavior feature data of the interviewee in answering the current interview question.

[0055] The audio data reflects the language behavior features related to the mental state of the interviewee. Therefore, the audio data in the audio-visual data can be analyzed and identified to obtain the language behavior feature data of the interviewee in answering the current interview question. The language behavior feature data may include but is not limited to at least one of the following: volume, intonation, timbre, speech rate, pause frequency, word choice, repeated words, changes in speaking rhythm, etc.

[0056] The third model is used to identify the language behavior feature data from the audio data. When determining the language behavior feature data of the interviewee in answering the current interview question, the audio data in the audio-visual data is used as the input of the third model and input into the third model. The third model identifies the corresponding language behavior feature data from the input audio data and outputs the identified language behavior feature data. It should be noted that in this embodiment, the audio data used as the input of the third model can be any one of the following: First, the audio data used as the input of the third model is only the audio data of the interviewee in answering the current interview question; Second, considering that human emotions have temporality, in order to more accurately determine the voice emotion feature data of the interviewee in answering the current interview question, the audio data used as the input of the third model includes the audio data of the interviewee in answering the current interview question and the audio data of the interviewee in answering the previous interview questions.

[0057] The third model is trained with multiple groups of data. Each group of data includes the audio data of the sample interviewee in answering the sample interview question and the corresponding language behavior feature data. The specific type of the third model can be flexibly selected based on business needs, and this embodiment does not make any limitations in this regard. Exemplarily, the second model can be selected but is not limited to the Recurrent Neural Networks (RNN).

[0058] In some embodiments, other behavioral feature data in addition to the language behavioral feature data are used to reflect other behavioral features of the interviewee in answering the current interview question (for example, micro-expression behavior, body behavior), which reflect the mental state of the interviewee in answering the current interview question from the dimension of non-verbal behavior. When evaluating the interviewee, based on the other behavioral feature data, the behavior and mental health status of the interviewee can be more accurately evaluated from the dimension of other behaviors, providing a more powerful screening basis for screening out interviewees who are competent for the corresponding positions subsequently, thereby improving the interview accuracy rate. Based on this, it is necessary to perform the step of analyzing the collected audio-visual data to obtain other behavioral feature data of the interviewee in answering the current interview question. The specific process of this step may include the following steps 102E to 102G.

[0059] 102E. Invoke the fourth model to analyze and identify the second video frame corresponding to the face of the interviewee when answering the current interview question in the audio-visual data, and obtain the micro-expression behavior feature data of the interviewee when answering the current interview question.

[0060] The other behavioral feature data may include but are not limited to at least one of the following: micro-expression behavior feature data, body behavior feature data. Among them, the micro-expression behavior feature data is used to reflect the micro-expression behavior from the subtle facial muscle changes, and it may include but is not limited to at least one of the following contents: eyebrow behavior data (such as raising, pressing down), eye behavior data (such as widening, narrowing, frequent blinking), nose behavior data (such as wrinkling), mouth behavior data (such as upward curling of the corners of the mouth, downward curling of the corners of the mouth, tight lips, open lips). The body behavior feature data is used to reflect the body behavior from the body movement changes, and it may include but is not limited to at least one of the following: sitting posture data (such as hunchback, uncrossed arms and legs, leaning forward, leaning backward, crossing arms, crossing legs or curling up the body), gesture data, head movement data (such as nodding, shaking the head, tilting the head, fixing the head), leg movement data (such as jittering, swaying, crossing legs, direction of the feet), arm movement data (such as crossing arms, spreading arms, stiff arms), eye contact data (such as frequent eye contact, avoiding eye contact, wandering eyes), etc.

[0061] The other behaviors of the interviewee are reflected in the micro-expressions on the face. Based on this, in this embodiment, the second video frame corresponding to the face of the interviewee when answering the current interview question is obtained from the audio-visual data, and the fourth model is invoked to analyze and identify the second video frame, and the micro-expression behavior feature data of the interviewee when answering the current interview question is obtained.

[0062] The fourth model is used to identify micro-expression behavior feature data from the video frames of the face. When determining the micro-expression behavior feature data of the interviewee in response to the current interview question, the second video frame corresponding to the face of the interviewee in response to the current interview question is used as the input of the fourth model and input into the fourth model. The fourth model identifies the corresponding micro-expression behavior feature data from the input second video frame and outputs the identified micro-expression behavior feature data. It should be noted that in this embodiment, the second video frame used as the input of the fourth model can include any one of the following: First, the second video frame used as the input of the fourth model is only the video frame corresponding to the face of the interviewee in response to the current interview question; Second, considering that human emotions are time-sequential, in order to more accurately determine the micro-expression behavior feature data of the interviewee in response to the current interview question, the second video frame used as the input of the fourth model includes the video frame corresponding to the face of the interviewee in response to the current interview question and the video frame corresponding to the face of the interviewee in response to the interview questions before the current interview question.

[0063] The fourth model is trained with multiple groups of data. Each group of data includes the video frame corresponding to the face of the sample interviewee in response to the sample interview question and the micro-expression behavior feature data corresponding to the video frame. The specific type of the fourth model can be flexibly selected based on business needs, and this embodiment does not make any limitations in this regard. Exemplarily, the fourth model can be selected but not limited to a recurrent neural network model, a long short-term memory network model

[0064] 102F. Invoke the fifth model to analyze and identify the third video frame corresponding to the body of the interviewee in response to the current interview question included in the audio-visual data, and obtain the limb behavior feature data of the interviewee in response to the current interview question.

[0065] Considering that the non-verbal behavior of the interviewee is not only reflected in micro-expressions but also in limb behavior. Based on this, the third video frame corresponding to the body of the interviewee in response to the current interview question is obtained from the audio-visual data, and the fifth model is invoked to analyze and identify the third video frame corresponding to the body of the interviewee in response to the current interview question included in the audio-visual data, and obtain the limb behavior feature data of the interviewee in response to the current interview question.

[0066] The fifth model is used to identify limb behavior feature data from the video frames of the body. When determining the limb behavior feature data of the interviewee when answering the current interview question, the third video frame corresponding to the face of the interviewee when answering the current interview question is used as the input of the fifth model and input into the fifth model. The fifth model identifies the corresponding limb behavior feature data from the input third video frame and outputs the identified limb behavior feature data. It should be noted that in this embodiment, the third video frame used as the input of the fifth model can be any of the following: one is that the third video frame used as the input of the fifth model is only the video frame corresponding to the limb of the interviewee when answering the current interview question; the other is that considering that human emotions are time-sequential, in order to more accurately determine the limb behavior feature data of the interviewee when answering the current interview question, the third video frame used as the input of the fifth model includes the video frame corresponding to the limb of the interviewee when answering the current interview question and the video frame corresponding to the face of the interviewee when answering the previous interview question before answering the current interview question.

[0067] The fifth model is trained with multiple sets of data. Each set of data includes the video frame corresponding to the body of the sample interviewee when answering the sample interview question and the limb behavior feature data corresponding to the video frame. The specific type of the fifth model can be flexibly selected based on business needs, and this embodiment does not make any limitations in this regard. Exemplarily, the fifth model can select a model based on a behavior tracking technical solution. This model is used to implement human pose estimation based on OpenPose, a custom gesture recognizer based on MediaPipe, and eye movement tracking technology based on PyGaze to obtain limb behavior feature data. To ensure the tracking accuracy of limb behavior, the fifth model can adopt a multi-model fusion strategy, specifically including key point detection, action classification, and trajectory prediction. In addition, this model can also use mechanisms such as a Kalman filter to eliminate jitter, double exponential smoothing to process outliers, adaptive threshold adjustment, and dynamic light compensation to ensure the stable recognition of limb behavior feature data in various environments. The accuracy of this model in pose estimation and gesture recognition both exceeds 92%, and the eye movement tracking accuracy is better than 1 degree of view.

[0068] 102G. Integrate the micro-expression behavior feature data and the limb behavior feature data to obtain other behavior feature data of the interviewee when answering the current interview question.

[0069] The other behavior features (i.e., non-verbal behavior features) of the interviewee are reflected in micro-expression behavior and limb behavior. Based on this, this embodiment integrates the micro-expression behavior feature data and the limb behavior feature data to obtain other behavior feature data of the interviewee when answering the current interview question. In this way, through the integration of the micro-expression behavior feature data and the limb behavior feature data, it can be realized that the obtained other behavior feature data is more in line with the real behavior state of the interviewee when answering the current interview question.

[0070] The fusion of micro-expression behavior feature data and body behavior feature data essentially means using one modality (such as the behavior analysis result of the micro-expression modality, "micro-expression behavior feature data") to fuse another modality (such as the behavior analysis result of the body modality, "body behavior feature data") to achieve their complementarity. This fusion can at least achieve the following two effects: (1) Different aspects of behavior features may be captured from the two modalities of micro-expression and body language. In this way, by fusing the micro-expression behavior feature data and the body behavior feature data, more real and comprehensive other behavior feature data of the interviewee can be obtained. (2) There are limitations in using either the micro-expression modality or the body modality alone for behavior analysis. For example, when the face part is blocked, the facial micro-expression behavior analysis may be restricted. Through fusion, the limitations of a single modality can be reduced, thereby improving the accuracy of other behavior feature data of the interviewee.

[0071] The specific process of fusing the micro-expression behavior feature data and the body behavior feature data to obtain other behavior feature data of the interviewee when answering the current interview question may include: projecting the micro-expression behavior feature data into a fourth vector in a second preset vector space, taking the body behavior feature data as a fifth vector in the second preset vector space, and based on the fourth vector and the fifth vector, obtaining a sixth vector through a preset second algorithm; based on the corresponding relationship between the preset vector and other behavior feature data, determining the target other behavior feature data corresponding to the sixth vector, and determining the target other behavior feature data as the other behavior feature data of the interviewee when answering the current interview question. It should be noted that the second preset vector space is a vector space used to determine other behavior feature data of the interviewee. The preset second algorithm can be flexibly determined based on business needs, and this embodiment does not make any limitations in this regard. Exemplarily, the second algorithm may include at least one of the following operations: addition operation, multiplication operation.

[0072] In some instances, considering that different interviewees may have different emotions and behaviors expressing their psychological stress states, based on this, in order to more accurately identify the true emotion feature data, language behavior feature data, and other behavior feature data of the interviewee, the interview method provided in this embodiment may further include the following steps: collecting target audio-visual data of the interviewee in a natural state; based on the target audio-visual data, fine-tuning the target model so that the fine-tuned target model is applicable to the interviewee. The target model can be selected from at least one of the first model, the second model, the third model, the fourth model, and the fifth model in the above steps 102A to 102G according to business requirements, and this embodiment does not make any limitations in this regard.

[0073] Within a preset duration before the interviewee starts the interview (for example, 5 seconds before the interview), collect the target audio-visual data of the interviewee in a natural state. Using this target audio-visual data as a benchmark, fine-tune the target model so that it is more suitable for the interviewee, and based on the more accurate target model, identify the true emotional characteristic data, language behavior characteristic data, and other behavior characteristic data of the interviewee.

[0074] In some embodiments, after determining the emotional characteristic data, language behavior characteristic data, and other behavior characteristic data of the interviewee's response to the current interview question, select the next interview question based on the emotional characteristic data, language behavior characteristic data, and other behavior characteristic data and send it to the interviewee's terminal, so as to timely adjust the interviewee's psychological pressure through the selected interview question, thereby ensuring the smooth progress of the interview.

[0075] The specific process of selecting the next interview question based on the emotional characteristic data, language behavior characteristic data, and other behavior characteristic data and sending it to the interviewee's terminal may include the following steps 102H to 102K:

[0076] 102H. Based on the emotional characteristic data, language behavior characteristic data, and other behavior characteristic data, determine the psychological pressure characteristic data of the interviewee.

[0077] Whether the interview can proceed smoothly depends on the interviewee's psychological pressure. Based on this, based on the emotional characteristic data, language behavior characteristic data, and other behavior characteristic data, which are the characteristic data used to reflect the interviewee's psychological pressure when answering the current interview question, determine the psychological pressure characteristic data of the interviewee, so as to select the next interview question suitable for the interviewee's current psychological pressure through the psychological pressure characteristic data, and timely adjust the interviewee's psychological pressure through the selected next interview question, reducing the possibility of the interviewee interrupting the interview, thereby ensuring the successful completion of the interview. The psychological pressure characteristic data is the data used to reflect the magnitude of the interviewee's psychological pressure when answering the current interview question.

[0078] Invoke the sixth model to determine the psychological stress characteristic data of the interviewee based on the emotion characteristic data, language behavior characteristic data, and other behavior characteristic data. The sixth model is used to determine the psychological stress characteristic data of the interviewee. When determining the psychological stress characteristic data of the interviewee in response to the current interview question, the emotion characteristic data, language behavior characteristic data, and other behavior characteristic data are used as the input of the sixth model and input into the sixth model. The sixth model identifies the corresponding psychological stress characteristic data through the input emotion characteristic data, language behavior characteristic data, and other behavior characteristic data, and outputs the identified psychological stress characteristic data. The sixth model can be trained with multiple sets of data. Each set of data includes the emotion characteristic data, language behavior characteristic data, and other behavior characteristic data of the sample interviewee in response to the sample interview question, and the psychological stress characteristic data corresponding to the emotion characteristic data, language behavior characteristic data, and other behavior characteristic data. The specific type of the sixth model can be flexibly selected based on business needs, and this embodiment does not make any limitations in this regard.

[0079] 102I. Determine the mental state information of the interviewee based on the psychological stress characteristic data.

[0080] The mental state information is used to indicate the degree of psychological stress of the interviewee. Specifically, the mental state information can be a psychological stress score, and the psychological stress score is positively correlated with the degree of psychological stress. The higher the psychological stress score, the greater the degree of psychological stress of the interviewee.

[0081] Invoke the seventh model to determine the mental state information of the interviewee based on the psychological stress characteristic data. The seventh model is used to determine the mental state information of the interviewee. When determining the mental state information of the interviewee in response to the current interview question, the psychological stress characteristic data is used as the input of the seventh model and input into the seventh model. The seventh model determines the corresponding mental state information through the input psychological stress characteristic data and outputs the determined mental state information. The seventh model can be trained with multiple sets of data. Each set of data includes the psychological stress characteristic data of the sample interviewee in response to the sample interview question and the mental state information corresponding to the psychological stress characteristic data. The specific type of the seventh model can be flexibly selected based on business needs, and this embodiment does not make any limitations in this regard.

[0082] 102J. If the degree of psychological stress indicated by the mental state information is not less than the preset psychological stress threshold, select an interview question with a lower difficulty than the current interview question as the next interview question and send it to the interviewee terminal.

[0083] If the degree of psychological stress indicated by the psychological state information is not less than the preset psychological stress threshold, it indicates that the interviewee is currently in a high-pressure state, and the interviewee may not be able to complete the subsequent interview or may not be able to perform at their true level during the subsequent interview due to the stress. Based on this, an interview question with a lower difficulty than the current interview question is selected as the next interview question and sent to the interviewee's terminal to relieve the interviewee's psychological stress.

[0084] In some embodiments, the interview system usually sets a corresponding set of interview questions for each interview position. The interview questions in the set of interview questions corresponding to each interview position are divided into at least two difficulty levels based on difficulty. Based on this, the next interview question selected can be any one of the following: (1) The next interview question selected is an interview question that is N difficulty levels lower than the current interview question, where N is a positive integer other than zero, and N can be flexibly set based on business needs. For example, N is set to 1; (2) The next interview question selected is an interview question with a lower difficulty than the current interview question and is adapted to the degree of psychological stress indicated by the psychological state information. Each interview question in the set of interview questions not only has a corresponding difficulty level but also has a corresponding degree of psychological stress. Therefore, the next interview question selected not only has a lower difficulty than the current interview question but also needs to be adapted to the degree of psychological stress indicated by the psychological state information; (3) The next interview question selected is a basic interpretation question, which is used to instruct the interviewee to interpret and answer the basic professional knowledge corresponding to their interview position. It serves as a buffer question to adjust the interviewee's psychological stress. The set of interview questions corresponding to each interview position may also include basic interpretation questions, and the basic interpretation questions can be the interview questions with the lowest difficulty level in the set of interview questions.

[0085] Furthermore, the interview method provided in this embodiment may further include the following steps: If the degree of psychological stress indicated by the psychological state information is not less than the preset psychological stress threshold, and it is determined that the similarity between the answer given by the interviewee to the current interview question and the preset answer to the current interview question is not less than the similarity threshold, it indicates that although the interviewee is currently under relatively high psychological stress, the high psychological stress may not be caused by the high difficulty of the interview question. Therefore, in order to understand the interviewee's true level, an interview question with a difficulty not lower than the current interview question (for example, an interview question at the same difficulty level as the current interview question) is selected as the next interview question and sent to the interviewee's terminal. If the degree of psychological stress indicated by the information is not less than the preset psychological stress threshold, and it is determined that the similarity between the answer given by the interviewee to the current interview question and the preset answer to the current interview question is less than the similarity threshold, it indicates that the interviewee is currently under relatively high psychological stress, which may be caused by the high difficulty of the interview question. Therefore, in order to relieve the interviewee's stress, the step of selecting an interview question with a lower difficulty than the current interview question as the next interview question and sending it to the interviewee's terminal is executed.

[0086] 102K. If the psychological stress level indicated by the psychological state information is less than the preset psychological stress threshold, select an interview question with a difficulty level not lower than the current interview question as the next interview question and send it to the interviewee's terminal.

[0087] If the psychological stress level indicated by the psychological state information is less than the preset psychological stress threshold, it indicates that the interviewee is not currently in a high-pressure state and can still complete the subsequent interview. Based on this, select an interview question with a difficulty level not lower than the current interview question as the next interview question and send it to the interviewee's terminal.

[0088] Specifically, if the psychological stress level indicated by the psychological state information is less than the preset psychological stress threshold, determine whether the psychological stress levels indicated by the psychological state information of the interviewee for a consecutive target number of interview questions are all less than the preset psychological stress threshold; if they are all less than, select an interview question with a difficulty level lower than the current interview question as the next interview question and send it to the interviewee's terminal; if they are not all less than, continue to determine whether the similarity between the preset answer of each interview question in the target number and the answer given by the interviewee is greater than the first similarity; if they are all greater, select an interview question with a difficulty level higher than the current interview question or select an interview with the same difficulty level as the current interview question as the next interview question and send it to the interviewee's terminal to better understand the true level of the interviewee; if they are not all greater, select an interview question with the same difficulty level as the current interview question as the next interview question and send it to the interviewee's terminal.

[0089] Furthermore, the interview method provided in this embodiment may further include the following steps: If the psychological stress level indicated by the psychological state information is less than the preset psychological stress threshold and it is determined that the current interview question is less difficult than the previous interview question, it indicates that the interviewee's psychological stress has decreased. To understand the true level of the interviewee, select an interview question with a difficulty level not lower than the previous interview question as the next interview question and send it to the interviewee's terminal. If the psychological stress level indicated by the psychological state information is less than the preset psychological stress threshold and it is determined that the current interview question is not less difficult than the previous interview question, it indicates that the interviewee's psychological stress is not high and the possibility of the interview being interrupted due to psychological stress is small when the interviewee continues to answer interview questions not less difficult than the current interview question. Therefore, execute the step of selecting an interview question with a difficulty level not lower than the current interview question as the next interview question and send it to the interviewee's terminal to understand the true level of the interviewee.

[0090] 103. If it is determined that the interview process of the interviewee has ended, generate an interview report for the interviewee based on the corresponding emotional characteristic data, language behavior characteristic data, other behavior characteristic data, and answers of the interviewee to each interview question during the interview process.

[0091] The interview report is generated when the interview process of the interviewee is determined to be over. Therefore, the interview method provided in this embodiment may further include the step of determining whether the interview process of the interviewee is over. The implementation method of this step may include at least one of the following four methods:

[0092] First, the interviewee has a corresponding interview position, and each interview position has at least one corresponding type of position professional knowledge. Then, determining whether the interview process of the interviewee is over can be achieved through the following steps: If it is determined that interview questions covering each type of position professional knowledge corresponding to the interviewee's interview position have been issued to the interviewee, and the number of interview questions for each type of position professional knowledge is not less than the corresponding quantity threshold, then it is determined that the interview process of the interviewee is over.

[0093] The interview position is provided by enterprises and institutions to the interviewee and is the position that the interviewee expects to obtain through the interview. Each interview position has at least one corresponding type of position professional knowledge, and the type of position professional knowledge is the type of professional knowledge covered by the professional knowledge that a person competent for the corresponding interview position needs to possess.

[0094] During the interview process, if it is determined that interview questions covering each type of position professional knowledge corresponding to the interviewee's interview position have been issued to the interviewee, and the number of interview questions for each type of position professional knowledge is not less than the corresponding quantity threshold, it means that enough interview questions have been issued to the interviewee to verify whether he or she is competent for the corresponding interview position, and there is enough data to evaluate the interviewee. Therefore, it is determined that the interview process of the interviewee is over.

[0095] During the interview process, if it is determined that the interview questions issued to the interviewee do not cover each type of position professional knowledge corresponding to the interviewee's interview position, or, if it is determined that interview questions covering each type of position professional knowledge corresponding to the interviewee's interview position have been issued to the interviewee, and there is a type of position professional knowledge for which the number of interview questions is not greater than the corresponding quantity threshold, it means that the interview questions issued to the interviewee are not sufficient to verify whether the interviewee is competent for the corresponding interview position. Therefore, it is determined that the interview process of the interviewee is not over, and the interviewee needs to be interviewed continuously.

[0096] Second, determining whether the interview process of the interviewee is over can be achieved through the following steps: Determine the similarity between the answer of the interviewee to the current interview question and the preset answer of the current interview question; Determine the fitness between the emotional characteristic data, language behavior characteristic data, and other behavior characteristic data and the similarity, where the fitness is used to indicate the possibility that the interviewee does not answer the current interview question truthfully based on his or her own professional knowledge and uses cheating means to answer the interview question; If the fitness is less than the fitness threshold, then it is determined that the interview process of the interviewee is over.

[0097] The similarity between the answer given by the interviewee to the current interview question and the preset answer to the current interview question reflects the correctness of the interviewee's answer. The higher the similarity, the closer the answer is to the preset answer, and the higher the correctness of the answer. In principle, the response of the interviewee to the current interview question will be reflected in the interviewee's mood, speech rhythm and other language behaviors, sitting posture and other non-verbal behaviors, that is, other behaviors. Therefore, the similarity between the answer given by the interviewee to the current interview question and the preset answer to the current interview question, and the fitness between the mood characteristic data, language behavior characteristic data and other behavior characteristic data of the interviewee's response to the current interview question can be used to evaluate whether the answer of the respondent is a true response based on his own professional knowledge.

[0098] In some embodiments, the specific process of determining the similarity between the answer given by the interviewee to the current interview question and the preset answer to the current interview question may include the following steps: obtaining the audio data of the interviewee's response to the current interview question from the collected audio-visual data, and extracting the answer of the interviewee's response to the current interview question from the audio data; calling a preset similarity model to determine the similarity between the answer given by the interviewee to the current interview question and the preset answer to the current interview question.

[0099] In some embodiments, after determining the similarity value, a fitness determination model is called to determine the fitness between the mood characteristic data, language behavior characteristic data and other behavior characteristic data and the similarity. Specifically, the mood characteristic data, language behavior characteristic data and other behavior characteristic data, and similarity of the interviewee's response to the current interview question are used as model inputs and input into the fitness determination model to obtain the fitness between the mood characteristic data, language behavior characteristic data and other behavior characteristic data and the similarity. The fitness determination model is obtained based on multiple groups of data, and each group of data includes the fitness and the mood characteristic data, language behavior characteristic data and other behavior characteristic data, and similarity corresponding to the fitness.

[0100] If the fitness is less than the fitness threshold, it means that the answer given by the interviewee to the current interview question is probably not a true response to the current interview question based on the interviewee's own professional knowledge. It may be that the interviewee uses cheating means to answer the interview question, or uses a deliberately accurate answer to answer the interview question. The interviewee has a deception behavior and there is no need to continue the subsequent interview. Therefore, it is determined that the interview process of the interviewee ends.

[0101] If the fitness is not less than the fitness threshold, it means that the answer given by the interviewee to the current interview question is probably a true response to the current interview question based on the interviewee's own professional knowledge. Therefore, the subsequent interview can be continued.

[0102] Exemplarily, the similarity between the answer given by the interviewee to the current interview question and the preset answer to the current interview question reaches 90%, which indicates that the interviewee's answer is correct. In principle, the corresponding emotional characteristic data, language behavior characteristic data, and other behavior characteristic data are emotional stability, large fluctuations in speaking rhythm, looking straight ahead, etc. However, the emotional characteristic data, language behavior characteristic data, and other behavior characteristic data of the interviewee answering the current interview question are emotional tension, small fluctuations in speaking rhythm, and looking down. The degree of fit between the determined emotional characteristic data, language behavior characteristic data, and other behavior characteristic data and the similarity is relatively low and less than the fit threshold. At this time, it indicates that the interviewee may have used cheating means to answer the interview question, and the interviewee may have deceptive behavior. There is no need to continue the subsequent interview, so it is determined that the interview process of the interviewee ends.

[0103] Thirdly, determining whether the interview process of the interviewee ends can be achieved through the following steps: Based on the language behavior characteristic data and other behavior characteristic data, identify abnormal interview behaviors of the interviewee; if abnormal interview behaviors are identified, determine that the interview process of the interviewee ends.

[0104] During the interview process, abnormal interview behaviors related to the language behavior characteristic data and other behavior characteristic data may occur. Abnormal interview behaviors may include, but are not limited to: impolite behaviors, physical discomfort behaviors. Based on this, the method for identifying abnormal interview behaviors of the interviewee based on the language behavior characteristic data and other behavior characteristic data can include the following two types:

[0105] One is that considering that the impolite language and impolite body movements of the interviewee are reflected in the language behavior characteristic data and other behavior characteristic data, based on this, call the first abnormal interview behavior recognition model to identify abnormal interview behaviors of the interviewee based on the language behavior characteristic data and other behavior characteristic data. If abnormal interview behaviors such as impolite language or impolite body movements are identified, it means that the interviewee is not competent for the corresponding interview position, there is no need to continue the subsequent interview, and it is determined that the interview process of the interviewee ends.

[0106] The first abnormal interview behavior recognition model takes the language behavior characteristic data and other behavior characteristic data as inputs, and based on the input language behavior characteristic data and other behavior characteristic data, identifies abnormal interview behaviors such as impolite language or impolite body movements. The first abnormal interview behavior recognition model is trained based on multiple groups of data. Each group of data includes language behavior characteristic data and other behavior characteristic data, and the corresponding abnormal interview behaviors such as impolite language or impolite body movements to the language behavior characteristic data and other behavior characteristic data. The specific type of the first abnormal interview behavior recognition model can be selected based on business needs, and this embodiment does not make a limitation on this.

[0107] Another is that considering that the interviewee may experience physical discomfort during the interview, and physical discomfort will be reflected in the language behavior feature data and other behavior feature data. Based on this, the second abnormal interview behavior recognition model is called, and based on the language behavior feature data and other behavior feature data, the abnormal interview behavior of the interviewee is recognized. If the interviewee shows physical discomfort behavior, it means that the interviewee cannot continue the interview.

[0108] The second abnormal interview behavior recognition model takes the language behavior feature data and other behavior feature data as inputs, and recognizes abnormal interview behaviors such as physical discomfort behavior based on the input language behavior feature data and other behavior feature data. The second abnormal interview behavior recognition model is trained based on multiple groups of data. Each group of data includes language behavior feature data, other behavior feature data, and physical discomfort behavior corresponding to the language behavior feature data and other behavior feature data. The specific type of the second abnormal interview behavior recognition model can be selected based on business needs, and this embodiment does not make any limitations in this regard.

[0109] For the above two methods of recognizing the abnormal interview behavior of the interviewee based on the language behavior feature data and other behavior feature data, at least one can be selected based on business needs, and this embodiment does not make any limitations in this regard.

[0110] In some embodiments, after determining that the interview process of the interviewee ends, it is necessary to perform the step of generating an interview report of the interviewee based on the emotional feature data, language behavior feature data, other behavior feature data, and answers corresponding to each interview question answered by the interviewee during the interview process, so as to evaluate the ability of the interviewee based on the interview report and thereby evaluate whether the interviewee is competent for the corresponding position. The step of generating the interview report of the interviewee may include the following steps 103A to 103C.

[0111] 103A. Based on the emotional feature data, language behavior feature data, and other behavior feature data corresponding to each interview question during the interview process, determine the psychological stress feature data corresponding to the interviewee's answer to each interview question.

[0112] The psychological stress feature data is used to indicate the stress resistance ability of the interviewee. For the specific process of determining the psychological stress feature data corresponding to the interviewee's answer to each interview question, reference can be made to the above step 102H, which will not be elaborated here.

[0113] 103B. Based on the similarity between the preset answer of each interview question during the interview process and the answer of the interviewee, determine the response evaluation data of each interview question.

[0114] The similarity between the preset answer to an interview question and the answer given by the interviewee is used to indicate the correctness of the interviewee's answer to the interview question. The higher the similarity, the more accurate the interviewee's answer. Based on this, the similarity reflects the interviewee's true professional level. Therefore, based on the similarity between the preset answer to each interview question and the answer given by the interviewee during the interview process, response evaluation data for each interview question is determined. The response evaluation data is used to indicate the professional and technical capabilities of the interviewee for the interviewed position. The response evaluation data includes the similarity, the knowledge mastered by the interviewee determined based on the similarity, and the degree of mastery of the knowledge.

[0115] 103C. Generate an interview report for the interviewee based on the psychological stress characteristic data and the response evaluation data.

[0116] Summarize the psychological stress characteristic data and the response evaluation data corresponding to each interview question together to generate an interview report for the interviewee. In this way, the interviewee can be comprehensively evaluated from the two dimensions of stress resistance ability and professional ability based on the content in the interview report, so as to better and more comprehensively understand the true level of the interviewee and decide whether to hire the interviewee.

[0117] In the interview method provided by the embodiments of the present application, during the interview process, audio and video data of the interviewee is collected. Each time an interview question is sent to the interviewee's terminal, the collected audio and video data is analyzed to obtain the emotional characteristic data, language behavior characteristic data, and other behavior characteristic data other than the language behavior characteristic data of the interviewee's response to the current interview question. And based on the emotional characteristic data, language behavior characteristic data, and other behavior characteristic data, the next interview question is selected and sent to the interviewee's terminal. When it is determined that the interview process of the interviewee ends, an interview report for the interviewee is generated based on the emotional characteristic data, language behavior characteristic data, other behavior characteristic data, and answers corresponding to each interview question during the interview process. It can be seen that the solution provided by this embodiment can achieve the following effects:

[0118] First, each time an interview question is sent to the interviewee's terminal during the interview process, the emotional characteristic data, language behavior characteristic data, and other behavior characteristic data of the interviewee's response to the current interview question are obtained by analyzing the audio and video data of the interviewee. And based on these characteristic data, namely the emotional characteristic data, language behavior characteristic data, and other behavior characteristic data, which are used to reflect the interviewee's psychological state from the interviewee's emotional state and real behavior, the next interview question is adaptively adjusted. The interview question adjusted by this adaptive adjustment mechanism can timely adjust the psychological stress of the interviewee, thereby reducing the possibility of interview interruption caused by the increase in the interviewee's psychological stress, realizing the humanization of the interview process, and ensuring the smooth completion of the interview.

[0119] Second, after the interview, the characteristic data reflecting the interviewee's psychological state from their emotional state and real behaviors during the interview, together with the answers to the interview questions, will be used as the basis for evaluating the interviewee in the interview report. This can more comprehensively evaluate the interviewee to assist in more accurately screening out interviewees who are competent for the corresponding position, thereby improving the interview accuracy rate.

[0120] In some embodiments of the present application, if the interviewee has insufficient stress resistance, it will be difficult for them to hold the corresponding position. Therefore, the interview method provided in this embodiment also adds the interviewee's psychological stress characteristic data as auxiliary data to the interview report for the interviewer to evaluate the interviewee. Based on this, the interview method provided in this embodiment may further include: determining the psychological stress characteristic data corresponding to each interview question answered by the interviewee based on the emotional characteristic data, language behavior characteristic data, and other behavior characteristic data corresponding to each interview question during the interview; generating the psychological stress change trend data based on the psychological stress characteristic data corresponding to each interview question, and adding the psychological stress change trend data to the interview report to enable the interviewer to understand the interviewee's psychological stress change process through the psychological stress change trend data, so that the interviewer can more accurately understand the interviewee's stress resistance from the dimension of the psychological stress change process.

[0121] Furthermore, if abnormal psychological stress characteristic data is found in the psychological stress characteristic data corresponding to each interview question answered by the interviewee, the abnormal psychological stress characteristic data will be marked in the interview report so that the interviewer can intuitively understand the abnormal psychological stress characteristic data and more quickly understand the interviewee's stress resistance. The abnormal psychological stress characteristic data may include, but is not limited to, psychological stress characteristic data with a psychological stress level indicated by the psychological state information not less than the abnormal psychological stress threshold.

[0122] In some embodiments of the present application, if the interviewee has insufficient emotional control ability, it will also be difficult for them to hold the corresponding position. Therefore, the interview method provided in this embodiment also adds the interviewee's emotional characteristic data as auxiliary data to the interview report for the interviewer to evaluate the interviewee. Based on this, the interview method provided in this embodiment may further include: generating the emotional change trend data based on the emotional characteristic data corresponding to each interview question during the interview, and adding the emotional trend data to the interview report to enable the interviewer to understand the interviewee's emotional change process through the emotional change trend data, so that the interviewer can more accurately understand the interviewee's competence for the position from the dimension of the emotional change process.

[0123] In some embodiments of the present application, if the interviewee has a bad personality or insufficient ability, it is difficult for the interviewee to hold the corresponding position even if the interviewee has excellent stress resistance. Based on this, the interview method provided in this embodiment may further include the following steps: determining the interviewee's own characteristic data based on the emotional characteristic data, language behavior characteristic data, and other behavior characteristic data corresponding to each interview question during the interview process, where the own characteristic data includes at least one of the following: personality characteristic data and ability characteristic data; adding the own characteristic data to the interview report. The personality characteristic data is data obtained by comprehensively considering the emotional characteristic data, language behavior characteristic data, and other behavior characteristic data corresponding to each question answered by the interviewee, and it is used to reflect the interviewee's personality traits. The ability characteristic data is data obtained by comprehensively considering the emotional characteristic data, language behavior characteristic data, and other behavior characteristic data corresponding to each question answered by the interviewee, and it is used to reflect the interviewee's ability level.

[0124] In some embodiments, a personality characteristic recognition model is called to determine the interviewee's personality characteristic data based on the emotional characteristic data, language behavior characteristic data, and other behavior characteristic data corresponding to each interview question. The personality characteristic data is data used to indicate the interviewee's personality. The personality characteristic recognition model is used to determine the interviewee's personality characteristic data. When determining the interviewee's personality characteristic data, the emotional characteristic data, language behavior characteristic data, and other behavior characteristic data corresponding to each interview question are used as the input of the personality characteristic recognition model and input into the personality characteristic recognition model. The personality characteristic recognition model determines the corresponding personality characteristic data of the interviewee through the input emotional characteristic data, language behavior characteristic data, and other behavior characteristic data, and outputs the determined personality characteristic data. The personality characteristic recognition model can be trained through multiple sets of data. Among them, each set of data includes the emotional characteristic data, language behavior characteristic data, and other behavior characteristic data when a sample interviewee answers each sample interview question in at least one sample interview question, and the personality characteristic data corresponding to the emotional characteristic data, language behavior characteristic data, and other behavior characteristic data of the at least one sample interview question. The specific type of the personality characteristic recognition model can be flexibly selected based on business needs, and this embodiment does not make any limitations in this regard.

[0125] In some embodiments, the ability feature recognition model is called to determine the ability feature data of the interviewee based on the emotional feature data, language behavior feature data and other behavior feature data corresponding to each interview question, and the ability feature data is used to indicate the data of the interviewee's ability. The ability feature data is used to indicate but not limited to the following ability of the interviewee: emotional stability ability, communication and expression ability, self-control ability, adaptability, on-the-spot reaction ability, and teamwork ability. The ability feature recognition model is used to determine the ability feature data of the interviewee. When determining the ability feature data of the interviewee, the emotional feature data, language behavior feature data and other behavior feature data corresponding to each interview question are used as inputs of the ability feature recognition model and input into the ability feature recognition model. The ability feature recognition model determines the corresponding ability feature data through the input emotional feature data, language behavior feature data and other behavior feature data, and outputs the determined ability feature data. The ability feature recognition model can be obtained by training multiple groups of data, wherein each group of data includes the emotional feature data, language behavior feature data and other behavior feature data of the sample interviewee when answering each sample interview question in at least one sample interview question and the ability feature data corresponding to the emotional feature data, language behavior feature data and other behavior feature data of the at least one sample interview question. The specific type of the capability feature identification model can be flexibly selected based on business needs, and this embodiment does not limit this.

[0126] Add personality trait data and ability trait data to the interview report as a reference for person-job matching, so that the interviewer can use the personality trait data and ability trait data to comprehensively evaluate the interviewee's work adaptability and teamwork ability from the personality and ability dimensions, thereby improving the accuracy of the interview.

[0127] In some embodiments of the present application, the interviewee may be required to provide sensitive data such as project participation and past salary and benefits during the interview process to provide a basis for job matching and salary grading. Considering that such sensitive information is usually provided by the interviewee, once the interviewee dishonestly conceals the real data and provides false data, it may cause the interviewee to be hired because he is not competent for the job. Based on this, the interview method provided in this embodiment may also include the following steps: if the current interview question is a sensitive question that requires the interviewee to provide his true situation, then based on the emotional feature data, the language behavior feature data and other behavior feature data, it is determined whether the interviewee has concealed the true situation when answering the sensitive question. If so, after the interview process is over, the interviewee's concealment of the true situation on the sensitive question is summarized in the interview report.

[0128] Call the concealment behavior recognition model to identify the concealment behavior recognition result of the interviewee based on the corresponding emotional feature data, language behavior feature data, and other behavior feature data of the interview question. The concealment behavior recognition result is used to indicate whether the interviewee has concealed the truth about sensitive questions, and when it exists, it specifies the specific behavior of concealing the truth. The concealment behavior recognition model is used to determine the concealment behavior recognition result of the interviewee. When determining the concealment behavior recognition result of the interviewee, the corresponding emotional feature data, language behavior feature data, and other behavior feature data of the interview question are used as the input of the concealment behavior recognition model and input into the concealment behavior recognition model. The concealment behavior recognition model determines the corresponding concealment behavior recognition result through the input emotional feature data, language behavior feature data, and other behavior feature data, and outputs the determined concealment behavior recognition result. The concealment behavior recognition model can be trained with multiple sets of data, where each set of data includes the emotional feature data, language behavior feature data, and other behavior feature data when the sample interviewee answers the sample interview question and the corresponding concealment behavior recognition result of the emotional feature data, language behavior feature data, and other behavior feature data.

[0129] Summarize the interviewee's behavior of concealing the truth about sensitive questions in the interview report to generate a risk reminder for the interviewer, so as to help the interviewer evaluate whether to hire the interviewee based on the interviewee's behavior of concealing the truth about sensitive questions.

[0130] In some embodiments of the present application, although the probability of causing an interview interruption is low when the interviewee is in a low-confidence state, it may affect the interviewee's true level of performance. Based on this, the interview method provided in this embodiment may further include the following steps: identify the low-confidence state of the interviewee based on the emotional feature data, language behavior feature data, and other behavior feature data; if it is identified that the interviewee is in a low-confidence state, send an encouraging prompt to the interviewee's terminal to enable the interviewee to regain confidence.

[0131] Call the low-confidence state recognition model to recognize the low-confidence state of the interviewee based on the corresponding emotional feature data, language behavior feature data, and other behavior feature data of the interview question. The low-confidence state recognition model is used to determine the low-confidence state of the interviewee. When recognizing the low-confidence state of the interviewee, the corresponding emotional feature data, language behavior feature data, and other behavior feature data of the interview question are used as the input of the low-confidence state recognition model and input into the low-confidence state recognition model. The low-confidence state recognition model determines the corresponding low-confidence state recognition result through the input emotional feature data, language behavior feature data, and other behavior feature data, and outputs the determined low-confidence state result. The low-confidence state recognition model can be trained through multiple groups of data, where each group of data includes the emotional feature data, language behavior feature data, and other behavior feature data when the sample interviewee answers the sample interview question, and the low-confidence state data corresponding to the emotional feature data, language behavior feature data, and other behavior feature data.

[0132] In some embodiments of the present application, the following uses Figure 2 The schematic diagram of the interview process shown to illustrate the interview method provided in this embodiment. As can be seen from Figure 2 When applying the interview method provided in this embodiment, first determine the interviewee and start the interview for the interviewee. During the interview process of the interviewee, collect the audio-visual data of the interviewee through the collection device. Then, every time an interview question is sent to the interviewee's terminal, the following steps are executed:

[0133] (1) Analyze the collected audio-visual data to obtain the emotional feature data, language behavior feature data, and other behavior feature data except the language behavior feature data when the interviewee answers the current interview question;

[0134] (2) Dispose of the emotional feature data, language behavior feature data, and other behavior feature data through the AI engine to identify the psychological stress feature data and target content when the interviewee answers the current interview question, where the target content includes fitness, abnormal interview behavior, interview question sending situation, behavior of concealing the true situation, and low-confidence state;

[0135] (3) According to the recognition result of the target content when the interviewee answers the current interview question, judge whether it is determined that the interview process of the interviewee ends; if so, determine that the interview process of the interviewee ends and execute step (6); if not, continue to execute step (4);

[0136] The solutions for judging whether it is determined that the interview process of the interviewee ends according to the recognition result of the target content when the interviewee answers the current interview question include the following situations:

[0137] Case ①, regarding the fitness degree. The fitness degree is used to indicate the possibility that the interviewee does not answer the current interview question truthfully based on the professional knowledge he / she has, but uses cheating means to answer the interview question. The determination process of the fitness degree can be referred to the foregoing embodiments. If it is determined that the fitness degree is less than the fitness degree threshold, it means that the answer of the interviewee to the current interview question is probably not a truthful answer based on the professional knowledge he / she has. It may be that the interviewee uses cheating means to answer the interview question, or uses deliberately accurate answers to answer the interview question. The interviewee has deceptive behavior and there is no need to continue the subsequent interview. Therefore, it is determined that the interview process of the interviewee ends.

[0138] Case ②, regarding abnormal interview behaviors. The identification process of abnormal interview behaviors can be referred to the foregoing embodiments. If abnormal interview behaviors are identified, it means that the interviewee is not competent for the corresponding interview position or the interviewee is physically unwell, and there is no need to continue the subsequent interview. Then it is determined that the interview process of the interviewee ends.

[0139] Case ③, regarding the distribution of interview questions. If it is determined that after the current interview question is distributed, interview questions covering each type of professional knowledge of the interviewee's interview position have been distributed to the interviewee, and the number of interview questions for each type of professional knowledge is not less than the corresponding quantity threshold, then it is determined that the interview process of the interviewee ends.

[0140] Case ④, regarding the behavior of concealing the truth. If the current interview question is a sensitive question and it is determined that the interviewee has the behavior of concealing the truth when answering the sensitive question, then record the behavior of concealing the truth of the current interview question, so that after the interview process ends, the behavior of the interviewee concealing the truth of the current interview question can be summarized in the interview report.

[0141] Case ⑤, regarding the low-confidence state. If it is identified that the interviewee is in a low-confidence state when answering the current interview question, then send an encouraging prompt to the interviewee's terminal to enable the interviewee to regain confidence in the interview.

[0142] If it is not determined that the interview process of the interviewee ends in any of Cases ① to ③, then step (4) is executed. If it is determined that the interview process of the interviewee ends in any of Cases ① to ③, then step (6) is executed.

[0143] (4) Select the next interview question in combination with the psychological stress characteristic data, the self-characteristic data, and the behavior of concealing the truth;

[0144] The method of selecting the next interview question can be referred to step 102 in the foregoing embodiments and will not be elaborated here.

[0145] (5) Provide the selected interview questions to the interviewer for the interviewer to decide whether to provide the next selected interview question to the interviewee's terminal. If so, send the next selected interview question to the interviewee's terminal. If not, send the default interview question after the current interview question to the interviewee's terminal. The default interview question is the interview question preset between the current interview questions;

[0146] (6) If it is determined that the interview process of the interviewee ends, generate an interview report for the interviewee based on the emotional characteristic data, language behavior characteristic data, other behavior characteristic data, and answers corresponding to each interview question during the interview process.

[0147] In addition to including the emotional characteristic data, language behavior characteristic data, other behavior characteristic data, and answers corresponding to each interview question during the interview process, the interview report may also include the following Contents One to Five to assist in evaluating whether the interviewee is competent for the corresponding interview position through these contents and improve the accuracy of the interview.

[0148] Content One, emotional trend data. The emotional trend data is generated based on the emotional characteristic data corresponding to each interview question during the interview process.

[0149] Content Two, psychological stress change trend data. The psychological stress change trend data is obtained based on the emotional characteristic data, language behavior characteristic data, and other behavior characteristic data corresponding to each interview question during the interview process.

[0150] Content Three, abnormal psychological stress characteristic data. The abnormal psychological stress characteristic data is the data evaluated from the psychological stress characteristic data corresponding to each interview question answered by the interviewee.

[0151] Content Four, the interviewee's own characteristic data. The own characteristic data includes at least one of the following: personality characteristic data and ability characteristic data. The own characteristic data is obtained based on the emotional characteristic data, language behavior characteristic data, and other behavior characteristic data corresponding to each interview question during the interview process.

[0152] Content Five, fitness. The determination process of the fitness is as follows: Determine the similarity between the answer of the interviewee to the current interview question and the preset answer of the current interview question, and determine the fitness of the emotional characteristic data, language behavior characteristic data, and other behavior characteristic data with the similarity. To evaluate the possibility that the interviewee does not answer the current interview question truthfully based on the professional knowledge he / she has and uses cheating means to answer the interview question through the fitness.

[0153] After obtaining the interview report of the interviewee, the interview report is stored. In addition, in order to better assist the interviewer in evaluating the interviewee, the original interview data can also be stored. The original data includes interview questions, corresponding audio and video data for each interview question, and emotional characteristic data, language behavior characteristic data, and other behavior characteristic data obtained from the analysis of the audio and video data. After obtaining the interview report and the original data, the interviewer analyzes the employment ability of the interviewee based on the interview report and the original data, forms an evaluation report for indicating whether to hire the interviewee, and stores the evaluation report for the personnel system to call, so that the personnel system can initiate the corresponding hiring process for the corresponding interviewee.

[0154] An embodiment of the present application further provides an interview device, as Figure 3 shown. The interview device provided in this embodiment includes:

[0155] An acquisition module 21, configured to acquire audio and video data of the interviewee during the interview process;

[0156] A selection module 22, configured to, for each interview question sent to the interviewee terminal, analyze the acquired audio and video data to obtain emotional characteristic data, language behavior characteristic data, and other behavior characteristic data except the language behavior characteristic data of the interviewee's response to the current interview question; select the next interview question based on the emotional characteristic data, the language behavior characteristic data, and the other behavior characteristic data and send it to the interviewee terminal;

[0157] A generation module 23, configured to, if it is determined that the interview process of the interviewee ends, generate an interview report of the interviewee based on the emotional characteristic data, language behavior characteristic data, other behavior characteristic data, and answers corresponding to each interview question answered by the interviewee during the interview process.

[0158] The interview device provided by the embodiment of the present application collects the audio-visual data of the interviewee during the interview process. Every time an interview question is sent to the interviewee's terminal: analyze the collected audio-visual data to obtain the emotional characteristic data, language behavior characteristic data, and other behavior characteristic data of the interviewee's response to the current interview question, and select the next interview question based on the emotional characteristic data, language behavior characteristic data, and other behavior characteristic data and send it to the interviewee's terminal. When it is determined that the interview process of the interviewee ends, an interview report of the interviewee is generated based on the emotional characteristic data, language behavior characteristic data, other behavior characteristic data, and answers corresponding to each interview question during the interview process. It can be seen that the solution provided by this embodiment can achieve the following effects: First, every time an interview question is sent to the interviewee's terminal during the interview process, the emotional characteristic data, language behavior characteristic data, and other behavior characteristic data other than the language behavior characteristic data of the interviewee's response to the current interview question are obtained by analyzing the audio-visual data of the interviewee, and based on these characteristic data that reflect the interviewee's mental state from the interviewee's emotional state and real behavior, the next interview question is adaptively adjusted. The interview question adjusted by this adaptive adjustment mechanism can timely adjust the mental pressure of the interviewee, thereby reducing the possibility of interview interruption caused by the increase in the interviewee's mental pressure, realizing the humanization of the interview process, and ensuring the smooth completion of the interview. Second, after the interview, the characteristic data that reflect the interviewee's mental state from the interviewee's emotional state and real behavior, together with the answers of the interviewee to the interview questions, are used as the basis for evaluating the interviewee in the interview report. In this way, the interviewee can be evaluated more comprehensively to assist in more accurately screening out the interviewees who are competent for the corresponding positions, thereby improving the interview accuracy.

[0159] In some embodiments of the present application, as Figure 4 shown, the selection module 22 includes:

[0160] The first determination unit 221 is configured to determine the mental pressure characteristic data of the interviewee based on the emotional characteristic data, the language behavior characteristic data, and the other behavior characteristic data;

[0161] The evaluation unit 222 is configured to determine the mental state information of the interviewee based on the mental pressure characteristic data, and the mental state information is used to indicate the mental pressure degree of the interviewee;

[0162] The first selection unit 223 is configured to, if the mental pressure degree indicated by the mental state information is not less than a preset mental pressure threshold, select an interview question with a lower difficulty than the current interview question as the next interview question and send it to the interviewee's terminal;

[0163] A second selection unit 224, configured to, if the degree of psychological stress indicated by the psychological state information is less than the preset psychological stress threshold, select an interview question whose difficulty is not lower than that of the current interview question as the next interview question and send it to the interviewee terminal.

[0164] In some embodiments of the present application, as Figure 4 shown, the first selection unit 223 is specifically configured to, if the degree of psychological stress indicated by the psychological state information is not less than the preset psychological stress threshold, and it is determined that the similarity between the answer of the interviewee to the current interview question and the preset answer of the current interview question is not less than the similarity threshold, select an interview question whose difficulty is not lower than that of the current interview question as the next interview question and send it to the interviewee terminal; if the degree of psychological stress indicated by the information is not less than the preset psychological stress threshold, and it is determined that the similarity between the answer of the interviewee to the current interview question and the preset answer of the current interview question is less than the similarity threshold, then perform the step of selecting an interview question with a difficulty lower than that of the current interview question as the next interview question and sending it to the interviewee terminal.

[0165] In some embodiments of the present application, as Figure 4 shown, the second selection unit 224 is specifically configured to, if the degree of psychological stress indicated by the psychological state information is less than the preset psychological stress threshold, and it is determined that the current interview question is lower in difficulty than the previous interview question, select an interview question whose difficulty is not lower than that of the previous interview question as the next interview question and send it to the interviewee terminal; if the degree of psychological stress indicated by the psychological state information is less than the preset psychological stress threshold, and it is determined that the current interview question is not lower in difficulty than the previous interview question, then perform the step of selecting an interview question whose difficulty is not lower than that of the current interview question as the next interview question and sending it to the interviewee terminal.

[0166] In some embodiments of the present application, as Figure 4 shown, the interviewee has a corresponding interview position, and each interview position has at least one corresponding type of position professional knowledge. Then, the generation module 23 may include:

[0167] A second determination unit 231, configured to, if it is determined that interview questions covering each type of position professional knowledge corresponding to the interview position of the interviewee have been sent to the interviewee, and the number of interview questions for each type of position professional knowledge is not less than the corresponding quantity threshold, determine that the interview process of the interviewee ends.

[0168] In some embodiments of the present application, as Figure 4 shown, the generation module 23 may include:

[0169] A third determination unit 232, configured to determine the similarity between the answer of the interviewee to the current interview question and the preset answer of the current interview question; determine the fitness of the emotion feature data, the language behavior feature data, and the other behavior feature data with respect to the similarity, where the fitness is used to indicate the possibility that the interviewee does not truthfully answer the current interview question based on the professional knowledge he or she has and uses cheating means to answer the interview question; if the fitness is less than the fitness threshold, determine that the interview process of the interviewee ends.

[0170] In some embodiments of the present application, as Figure 4 shown, the generation module 23 may include:

[0171] A fourth determination unit 233, configured to identify abnormal interview behaviors of the interviewee based on the language behavior feature data and the other behavior feature data; if abnormal interview behaviors are identified, determine that the interview process of the interviewee ends.

[0172] In some embodiments of the present application, as Figure 4 shown, the generation module 23 may include:

[0173] A generation unit 234, configured to determine, based on the emotion feature data, the language behavior feature data, and the other behavior feature data corresponding to each interview question in the interview process, the psychological stress feature data of the interviewee answering each interview question, where the psychological stress feature data is used to indicate the stress resistance ability of the interviewee; determine the response evaluation data of each interview question based on the similarity between the preset answer and the answer of the interviewee for each interview question in the interview process, where the response evaluation data is used to indicate the professional technical ability of the interviewee for the interviewed position; generate an interview report of the interviewee based on the psychological stress feature data and the response evaluation data.

[0174] In some embodiments of the present application, as Figure 4 shown, the selection module 22 may include: a first identification unit 225, configured to call a first model to analyze and identify the audio data in the audio-visual data to obtain the voice emotion feature data of the interviewee when answering the current interview question, where the first model is used to identify voice emotion feature data from audio data; call a second model to analyze and identify the first video frame corresponding to the face of the interviewee when answering the current interview question included in the audio-visual data to obtain the micro-expression emotion feature data of the interviewee when answering the current interview question, where the second model is used to identify micro-expression emotion feature data from the video frame of the face; fuse the voice emotion feature data and the micro-expression emotion feature data to obtain the emotion feature data of the interviewee when answering the current interview question.

[0175] In some embodiments of the present application, as Figure 4 shown, the selection module 22 may include: a second recognition unit 226, configured to call a third model, analyze and recognize the audio data in the audio-visual data, and obtain the language behavior feature data of the interviewee when answering the current interview question, where the third model is used to recognize the language behavior feature data from the audio data.

[0176] In some embodiments of the present application, as Figure 4 shown, the selection module 22 may include: a third recognition unit 227, configured to call a fourth model, analyze and recognize the second video frame corresponding to the face of the interviewee when answering the current interview question included in the audio-visual data, and obtain the micro-expression behavior feature data of the interviewee when answering the current interview question, where the fourth model is used to recognize the micro-expression behavior feature data from the video frames of the face; call a fifth model, analyze and recognize the third video frame corresponding to the body of the interviewee when answering the current interview question included in the audio-visual data, and obtain the limb behavior feature data of the interviewee when answering the current interview question, where the fifth model is used to recognize the limb behavior feature data from the video frames of the body; fuse the micro-expression behavior feature data and the limb behavior feature data to obtain other behavior feature data of the interviewee when answering the current interview question.

[0177] In some embodiments of the present application, as Figure 4 shown, the interview device provided in this embodiment may further include: a fine-tuning module 24, configured to collect the target audio-visual data of the interviewee in a natural state; based on the target audio-visual data, fine-tune the target model so that the fine-tuned target model is applicable to the interviewee; the target model includes at least one of the following: the first model, the second model, the third model, the fourth model, and the fifth model.

[0178] In some embodiments of the present application, as Figure 4 shown, the interview device provided in this embodiment may further include: a first adding module 25, configured to generate the emotional change trend data of the interviewee based on the emotional feature data corresponding to each interview question in the interview process, and add the emotional trend data to the interview report generated by the generation module 23.

[0179] In some embodiments of the present application, as Figure 4As shown in the figure, the interview device provided in this embodiment may further include: a second adding module 26, configured to determine the psychological stress characteristic data corresponding to the interviewee's response to each interview question based on the emotion characteristic data, language behavior characteristic data, and other behavior characteristic data corresponding to each interview question in the interview process; generate the psychological stress change trend data of the interviewee based on the psychological stress characteristic data corresponding to each interview question, and add the psychological stress change trend data to the interview report generated by the generating module 23; if it is evaluated that there is abnormal psychological stress characteristic data in the psychological stress characteristic data corresponding to the interviewee's response to each interview question, mark the abnormal psychological stress characteristic data in the interview report generated by the generating module 23.

[0180] In some embodiments of the present application, as Figure 4 As shown in the figure, the interview device provided in this embodiment may further include: a third adding module 27, configured to determine the self-characteristic data of the interviewee based on the emotion characteristic data, language behavior characteristic data, and other behavior characteristic data corresponding to each interview question in the interview process, where the self-characteristic data includes at least one of the following: personality characteristic data and ability characteristic data, and the ability characteristic data is used to indicate one of the following abilities of the interviewee: emotion stability ability, communication and expression ability, self-control ability, strain ability, on-site reaction ability, and teamwork ability; add the self-characteristic data to the interview report generated by the generating module 23.

[0181] In some embodiments of the present application, as Figure 4 As shown in the figure, the interview device provided in this embodiment may further include: a prompting module 28, configured to identify the low-confidence state of the interviewee based on the emotion characteristic data, the language behavior characteristic data, and the other behavior characteristic data; if it is identified that the interviewee is in a low-confidence state, send an encouraging prompt to the interviewee's terminal.

[0182] In some embodiments of the present application, as Figure 4 As shown in the figure, the interview device provided in this embodiment may further include: a summarizing module 29, configured to, if the current interview question is a sensitive question that requires the interviewee to provide his / her own true situation, judge whether there is a behavior of concealing the true situation when the interviewee answers the sensitive question based on the emotion characteristic data, the language behavior characteristic data, and the other behavior characteristic data, and if so, after the interview process ends, summarize the behavior of the interviewee concealing the true situation for the sensitive question into the interview report generated by the generating module 23.

[0183] In the interview device provided in the embodiments of the present application, the detailed explanations adopted during the operation of each functional module can be referred to the corresponding detailed explanations of the above interview method embodiments, which will not be elaborated here.

[0184] Further, an embodiment of the present application further provides a computer-readable storage medium, where the storage medium includes a stored program, and when the program runs, it controls the device where the storage medium is located to execute the above-mentioned interview method.

[0185] Further, an embodiment of the present application further provides an electronic device, which includes: a memory for storing a program; a processor coupled to the memory for running the program to execute the above-mentioned interview method.

[0186] Further, an embodiment of the present application further provides a computer program product, including computer program / instructions, which implement the above-mentioned interview method when executed by a processor.

[0187] In the above embodiments, the descriptions of the respective embodiments have their own emphases. For parts not detailed in a certain embodiment, reference may be made to the relevant descriptions of other embodiments.

[0188] It can be understood that the relevant features in the above methods and devices can be referred to each other. In addition, the "first", "second", etc. in the above embodiments are used to distinguish the respective embodiments, and do not represent the advantages and disadvantages of the respective embodiments.

[0189] Those skilled in the art can clearly understand that for the convenience and brevity of description, the specific working processes of the systems, devices, and units described above can refer to the corresponding processes in the foregoing method embodiments, and will not be elaborated herein.

[0190] The algorithms and displays provided herein are not inherently related to any particular computer, virtual system, or other device. Various general-purpose systems can also be used in conjunction with the teachings herein. The structure required to construct such systems is obvious from the above description. In addition, the present application is not directed to any specific programming language. It should be understood that the content of the present application described herein can be implemented using various programming languages, and the descriptions of specific languages above are for disclosing the preferred embodiments of the present application.

[0191] In addition, the memory may include non-permanent memory in a computer-readable medium, in the form of random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash memory (flash RAM), and the memory includes at least one storage chip.

[0192] Those skilled in the art should understand that the embodiments of the present application can be provided as a method, a system, or a computer program product. Therefore, the present application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present application can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk memory, CD-ROM, optical memory, etc.) that contain computer-usable program code.

[0193] The present application is described with reference to the flowcharts and / or block diagrams of methods, apparatuses (systems), and computer program products according to the embodiments of the present application. It should be understood that each flow and / or block in the flowchart and / or block diagram, as well as the combination of flows and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing devices to generate a machine, such that the instructions executed by the processor of the computer or other programmable data processing devices generate means for implementing the functions specified in Figure 1 one flow or multiple flows and / or blocks Figure 1 one block or multiple blocks.

[0194] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to work in a specific manner, such that the instructions stored in the computer-readable memory generate a manufactured article including instruction means that implement the functions specified in Figure 1 one flow or multiple flows and / or blocks Figure 1 one block or multiple blocks.

[0195] These computer program instructions can also be loaded onto a computer or other programmable data processing device, such that a series of operation steps are executed on the computer or other programmable device to generate a computer-implemented process, and thus the instructions executed on the computer or other programmable device provide steps for implementing the functions specified in Figure 1 one flow or multiple flows and / or blocks Figure 1 one block or multiple blocks.

[0196] In a typical configuration, a computing device includes one or more processors (CPUs), an input / output interface, a network interface, and a memory.

[0197] The memory may include non-permanent memory in the form of computer-readable media, random access memory (RAM), and / or non-volatile memory, such as read-only memory (ROM) or flash memory (flash RAM). The memory is an example of computer-readable media.

[0198] Computer readable media include permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information. Information can be computer readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, magnetic tape magnetic disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer readable media does not include temporary computer readable media (transitory media), such as modulated data signals and carrier waves.

[0199] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.

[0200] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment or an embodiment in combination with software and hardware. Moreover, the present application may adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that contain computer-usable program code.

[0201] The above are only embodiments of the present application and are not intended to limit the present application. For those skilled in the art, the present application may have various changes and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included within the scope of the claims of the present application.

Claims

1. An interview method, characterized in that: The method comprises: During the interview process, collect the interviewee’s audio and video data; Each time an interview question is sent to the interviewer's terminal, the collected audio and video data are analyzed to obtain emotional characteristic data, language behavior characteristic data, and other behavior characteristic data other than the language behavior characteristic data of the interviewer in response to the current interview question, and the next interview question is selected based on the emotional characteristic data, the language behavior characteristic data, and the other behavior characteristic data and sent to the interviewer's terminal; If it is determined that the interview process of the interviewee is finished, an interview report of the interviewee is generated based on the emotional characteristic data, language behavior characteristic data, other behavior characteristic data and answers corresponding to the interviewee's responses to each interview question during the interview process.

2. The method according to claim 1, characterized in that The next interview question is selected based on the emotion feature data, the language behavior feature data, and the other behavior feature data and sent to the interviewer terminal, including: Determining the psychological stress characteristic data of the interviewee based on the emotion characteristic data, the language behavior characteristic data and the other behavior characteristic data; Determining the psychological state information of the interviewee based on the psychological stress characteristic data, wherein the psychological state information is used to indicate the psychological stress level of the interviewee; If the psychological stress level indicated by the psychological state information is not less than a preset psychological stress threshold, selecting an interview question with a lower difficulty than the current interview question as the next interview question and sending it to the interviewer terminal; If the psychological stress level indicated by the psychological state information is less than the preset psychological stress threshold, an interview question with a difficulty not less than the current interview question is selected as the next interview question and sent to the interviewer terminal.

3. The method according to claim 2, characterized in that The method further includes: if the psychological stress level indicated by the psychological state information is not less than a preset psychological stress threshold, and it is determined that the similarity between the interviewee's answer to the current interview question and the preset answer to the current interview question is not less than the similarity threshold, then selecting an interview question with a difficulty not less than the current interview question as the next interview question and sending it to the interviewee terminal; if the psychological stress level indicated by the information is not less than the preset psychological stress threshold, and it is determined that the similarity between the interviewee's answer to the current interview question and the preset answer to the current interview question is less than the similarity threshold, then executing the step of selecting an interview question with a difficulty less than the current interview question as the next interview question and sending it to the interviewee terminal; and / or, The method also includes: if the psychological stress level indicated by the psychological state information is less than the preset psychological stress threshold, and it is determined that the current interview question is less difficult than the previous interview question, then selecting an interview question with a difficulty not less than the previous interview question as the next interview question and sending it to the interviewer terminal; if the psychological stress level indicated by the psychological state information is less than the preset psychological stress threshold, and it is determined that the current interview question is not less difficult than the previous interview question, then executing the step of selecting an interview question with a difficulty not less than the current interview question as the next interview question and sending it to the interviewer terminal.

4. The method according to claim 1, characterized in that: The interviewer has a corresponding interview position, and each interview position has at least one corresponding position professional knowledge type. Then, the method further includes: if it is determined that interview questions covering each position professional knowledge type corresponding to the interviewer's interview position have been sent to the interviewer terminal, and the number of interview questions of each position professional knowledge type is not less than the corresponding number threshold, then determining that the interview process of the interviewer is finished; and / or, The method further includes: determining the similarity between the interviewee's answer to the current interview question and the preset answer to the current interview question; determining the degree of fit between the emotional characteristic data, the language behavior characteristic data, and the other behavior characteristic data and the similarity, wherein the degree of fit is used to indicate the possibility that the interviewee does not truly answer the current interview question based on his / her professional knowledge and uses cheating means to answer the interview question; if the degree of fit is less than a threshold value of the degree of fit, determining that the interview process of the interviewee is over; and / or, The method further includes: identifying abnormal interview behavior of the interviewee based on the language behavior feature data and the other behavior feature data; if abnormal interview behavior is identified, determining that the interview process of the interviewee is finished.

5. The method according to claim 1, characterized in that Generate an interview report of the interviewer based on the emotional characteristic data, language behavior characteristic data, other behavior characteristic data and answers corresponding to each interview question answered by the interviewer during the interview, including: Determine the psychological stress characteristic data corresponding to the interviewee's answer to each interview question based on the emotional characteristic data, language behavior characteristic data and other behavior characteristic data corresponding to each interview question in the interview process, wherein the psychological stress characteristic data is used to indicate the interviewee's stress resistance ability; Determining response evaluation data for each interview question based on the similarity between the preset answer to each interview question in the interview process and the interviewee's answer, wherein the response evaluation data is used to indicate the professional and technical ability of the interviewee for the interviewed position; An interview report of the interviewer is generated based on the psychological stress characteristic data and the response evaluation data.

6. The method according to claim 1, characterized in that Analyzing the collected audio and video data to obtain emotional characteristic data of the interviewee's answer to the current interview question, including: calling a first model to analyze and identify the audio data in the audio and video data to obtain the voice emotional characteristic data of the interviewee when answering the current interview question, the first model is used to identify the voice emotional characteristic data from the audio data; calling a second model to analyze and identify the first video frame corresponding to the face of the interviewee when answering the current interview question included in the audio and video data to obtain the micro-expression emotional characteristic data of the interviewee when answering the current interview question, the second model is used to identify the micro-expression emotional characteristic data from the facial video frame; fusing the voice emotional characteristic data and the micro-expression emotional characteristic data to obtain the emotional characteristic data of the interviewee when answering the current interview question; and / or, Analyzing the collected audio and video data to obtain the language behavior feature data of the interviewee when answering the current interview question, including: calling a third model to analyze and identify the audio data in the audio and video data to obtain the language behavior feature data of the interviewee when answering the current interview question, wherein the third model is used to identify the language behavior feature data from the audio data; and / or, Analyze the collected audio and video data to obtain other behavioral feature data of the interviewee's answer to the current interview question, including: calling the fourth model to analyze and identify the second video frame corresponding to the interviewee's face when answering the current interview question included in the audio and video data, to obtain the micro-expression behavior feature data of the interviewee when answering the current interview question, and the fourth model is used to identify the micro-expression behavior feature data from the facial video frame; call the fifth model to analyze and identify the third video frame corresponding to the interviewee's body when answering the current interview question included in the audio and video data, to obtain the limb behavior feature data of the interviewee when answering the current interview question, and the fifth model is used to identify the limb behavior feature data from the body video frame; fuse the micro-expression behavior feature data and the limb behavior feature data to obtain other behavioral feature data of the interviewee when answering the current interview question.

7. The method according to any one of claims 1 to 6, characterized in that The method further comprises: generating emotional change trend data of the interviewee based on the emotional characteristic data corresponding to each interview question in the interview process, and adding the emotional trend data to the interview report; and / or, The method further includes: determining the psychological stress characteristic data corresponding to each interview question answered by the interviewee based on the emotional characteristic data, language behavior characteristic data and other behavior characteristic data corresponding to each interview question in the interview process; generating the psychological stress change trend data of the interviewee based on the psychological stress characteristic data corresponding to each interview question, and adding the psychological stress change trend data to the interview report; if it is assessed that there is abnormal psychological stress characteristic data in the psychological stress characteristic data corresponding to each interview question answered by the interviewee, marking the abnormal psychological stress characteristic data in the interview report; and / or, The method further comprises: determining the interviewee's own characteristic data based on the emotional characteristic data, language behavior characteristic data and other behavior characteristic data corresponding to each interview question in the interview process, wherein the own characteristic data comprises at least one of the following: personality characteristic data and ability characteristic data, wherein the ability characteristic data is used to indicate the interviewee's ability: emotional stability ability, communication and expression ability, self-control ability, adaptability, on-the-spot reaction ability, and teamwork ability; and adding the own characteristic data to the interview report; and / or, The method further includes: identifying the low-confidence state of the interviewee based on the emotional characteristic data, the language behavior characteristic data and the other behavior characteristic data; if the interviewee is identified to be in a low-confidence state, sending an encouraging reminder to the interviewee terminal; and / or, The method also includes: if the current interview question is a sensitive question that requires the interviewee to provide his or her true situation, then based on the emotional characteristic data, the language behavior characteristic data and the other behavior characteristic data, it is determined whether the interviewee has concealed the true situation when answering the sensitive question; if so, after the interview process is over, the interviewee's behavior of concealing the true situation on the sensitive question is summarized in the interview report.

8. An interview device, characterized in that: The device comprises: The collection module is used to collect the audio and video data of the interviewer during the interview process; A selection module is used for, each time an interview question is sent to the interviewer terminal, analyzing the collected audio and video data to obtain the emotional characteristic data, language behavior characteristic data and other behavior characteristic data other than the language behavior characteristic data of the interviewer in response to the current interview question, and selecting the next interview question based on the emotional characteristic data, the language behavior characteristic data and the other behavior characteristic data and sending it to the interviewer terminal; A generation module is used to generate an interview report of the interviewer based on the emotional characteristic data, language behavior characteristic data, other behavior characteristic data and answers corresponding to each interview question answered by the interviewer during the interview process if it is determined that the interview process of the interviewer is over.

9. A computer-readable storage medium, characterized in that: The storage medium includes a stored program, wherein when the program is executed, the device where the storage medium is located is controlled to execute the interview method described in any one of claims 1 to 7.

10. An electronic device, characterized in that: The electronic device comprises: Memory, used to store programs; A processor, coupled to the memory, is configured to run the program to execute the interview method according to any one of claims 1 to 7.

Images