An interview method and device based on a video interview system

Abstract

The application discloses an interview method and device based on a video interview system, relates to the technical field of artificial intelligence, and mainly aims at reducing the possibility of artificial review of interview results and improving interview efficiency. The main technical scheme of the application is as follows: obtaining recruitment requirements of a target post, resume information of a candidate and a post evaluation result; generating a current interview question based on a capability item of the recruitment requirements and asking the candidate the question so as to collect, through a video interaction collection module, the answer content of the candidate to the current interview question and facial emotional information when the candidate answers; determining the effectiveness of the current answer according to the answer content and the facial emotional information; determining a target question determination strategy in a question follow-up strategy and a question promotion strategy according to whether the answer effectiveness meets preset effectiveness requirements, and determining a next interview question according to the target question determination strategy until a preset question asking deadline requirement is met; and generating an interview result based on all the interview questions, the answer content and the answer effectiveness.

Description

Technical Field

[0001] This invention relates to the field of artificial intelligence technology, and in particular to an interview method and apparatus based on a video interview system. Background Technology

[0002] As companies continue to expand their recruitment scale, AI-based video interview systems have emerged. However, most mainstream interview systems currently employ a pre-set question bank mechanism: that is, a fixed sequence of questions is pre-configured for different positions, and candidates are asked questions in a predetermined order during the interview, with a recruitment score report generated based on the overall answers after all questions and answers are completed.

[0003] This static questioning model has significant limitations: on the one hand, it lacks the ability to flexibly follow up with questions based on the candidate's real-time responses, making it difficult to delve deeper into their potential strengths or key experiences; on the other hand, it cannot dynamically identify potential logical contradictions or inconsistencies between different answers, leading to superficial and generalized evaluation results that fail to truly reflect the candidate's ability and job suitability. Ultimately, manual review or even a second interview is still required, which undermines the efficiency gains that automated interviewing should bring. Summary of the Invention

[0004] In view of the above problems, the present invention provides an interview method and apparatus based on a video interview system, the main purpose of which is to reduce the possibility of manual review of interview results and improve interview efficiency.

[0005] To solve the above-mentioned technical problems, the present invention proposes the following solution: In a first aspect, the present invention provides an interview method based on a video interview system, the method comprising: Obtain the recruitment requirements for the target position, the candidate's resume information, and the candidate's job assessment results completed in the online assessment system; Based on the competency items in the recruitment requirements, the current interview questions are generated and asked to the candidate, so as to collect the candidate's answers to the current interview questions and facial emotion information during the answering process through the video interaction acquisition module; The validity of the current response is determined based on the content of the response and facial emotion information. Based on whether the validity of the answer meets the preset validity requirements, a target question determination strategy is determined from the question follow-up strategy and question advancement strategy, and the next interview question is determined based on the target question determination strategy until the preset questioning deadline is reached; The interview results are generated based on all interview questions, answers, and the validity of the answers.

[0006] Secondly, the present invention provides an interview device based on a video interview system, the device comprising: The information acquisition unit is used to acquire the recruitment requirements of the target position, the candidate's resume information, and the job assessment results completed by the candidate in the online assessment system. The first question-posing unit is used to generate current interview questions based on the ability items in the recruitment requirements obtained by the information acquisition unit and ask them to the candidate, so as to collect the candidate's answers to the current interview questions and facial emotion information during the answering process through the video interaction acquisition module. The validity assessment unit is used to determine the validity of the current answer based on the answer content and facial emotion information obtained from the first question-asking unit. The second question-posing unit is used to determine the target question-determining strategy from the question follow-up strategy and question advancement strategy based on whether the validity of the answer obtained by the validity evaluation unit meets the preset validity requirements, and to determine the next interview question based on the target question-determining strategy until the preset questioning deadline is reached. The results determination unit is used to generate interview results from all interview questions, answers, and the validity of the answers.

[0007] To achieve the above objectives, according to a third aspect of the present invention, a storage medium is provided, the storage medium including a stored program, wherein, when the program is executed, the device where the storage medium is located executes the interview method based on the video interview system of the first aspect described above.

[0008] To achieve the above objectives, according to a fourth aspect of the present invention, a processor is provided for running a program, wherein the program executes the interview method based on the video interview system described in the first aspect.

[0009] By employing the above technical solution, this invention provides an interview method and apparatus based on a video interview system. It prepares for the interview by acquiring the recruitment requirements of the target position, the candidate's resume information, and the job assessment results. Secondly, instead of relying on a fixed question bank, it dynamically generates initial interview questions based on the competency items in the recruitment requirements and asks them to the candidate. During the candidate's response, a video interaction acquisition module captures the candidate's answers to the current interview questions and their facial emotional information. Furthermore, the validity of the current answer is objectively determined based on the answer content and facial emotional information. Based on this, a target question determination strategy is determined in the question follow-up strategy and question progression strategy according to whether the answer validity meets the preset validity requirements. The next interview question is then determined based on the target question determination strategy until the preset questioning deadline is reached. Finally, the interview result is generated based on all interview questions, answer content, and answer validity. Compared with existing technologies, this mechanism dynamically links the generation of interview questions with the validity of the candidate's answers, avoiding the ability assessment bias that may be caused by asking questions in a fixed order or randomly. The final interview result is based on the actual question and answer content and its validity, improving the objectivity of the assessment and reducing the reliance on manual review to a certain extent.

[0010] The above description is merely an overview of the technical solution of the present invention. In order to better understand the technical means of the present invention and to implement it in accordance with the contents of the specification, and in order to make the above and other objects, features and advantages of the present invention more apparent and understandable, specific embodiments of the present invention are described below. Attached Figure Description

[0011] Various other advantages and benefits will become apparent to those skilled in the art upon reading the following detailed description of preferred embodiments. The accompanying drawings are for illustrative purposes only and are not intended to limit the invention. Furthermore, the same reference numerals denote the same parts throughout the drawings. In the drawings: Figure 1 This invention provides a flowchart of an interview method based on a video interview system according to an embodiment of the present invention. Figure 2 This invention provides a flowchart of another interview method based on a video interview system. Figure 3 This diagram illustrates a block diagram of an interview device based on a video interview system according to an embodiment of the present invention. Figure 4 This invention provides a block diagram of another interview device based on a video interview system. Detailed Implementation

[0012] Exemplary embodiments of the present disclosure will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the disclosure to those skilled in the art.

[0013] The user data, data acquisition, and / or use involved in the embodiments of this application strictly comply with the laws, regulations, and industry standards of relevant countries and regions. The collection and acquisition of data involved in the embodiments of this application are all done in advance by actively prompting or prominently displaying information to inform users and obtaining authorization, or by obtaining full authorization from all parties. The processing, manipulation, forwarding, and use of data involved in the embodiments of this application are all carried out on the premise that the user or relevant party is fully informed and authorized. In implementing the various embodiments of this application, the types of data or information, scope of use, and usage scenarios that may be involved are informed to users or relevant parties and authorization is obtained through appropriate means. The specific methods of notification and authorization may vary according to actual circumstances, and this application is not limited in this regard. The processing of personal information involved in the embodiments of this application is carried out under the premise of having a legal basis (such as obtaining the consent of the personal information subject or being necessary for the performance of a contract), and is only processed within the scope stipulated or agreed. Sensitive personal information such as biometric information, medical and health information, financial account information, and precise location information involved in the embodiments of this application are all processed under the premise of having a specific purpose and sufficient necessity, and with the separate authorization and consent of the user or relevant party. In some embodiments of this application, if the user or related party refuses to process personal information other than the information necessary for the basic functions, it will not affect the use of the basic functions of the embodiments of this application.

[0014] To address the problems of existing video interview systems that rely on fixed question banks and lack dynamic interaction mechanisms, this invention proposes an interview method based on a video interview system. This method abandons the practice of pre-setting a question sequence and instead dynamically generates initial questions based on the competencies required for the job posting. Furthermore, during the candidate's response, the method can simultaneously collect the candidate's answer content and facial emotion information, and judge the validity of the answer in real time based on this information. Subsequently, based on whether the answer's validity meets the preset validity requirements, a target question determination strategy is determined from the question follow-up strategy and question advancement strategy. The next interview question is then generated based on this target question determination strategy, and this process is repeated until the preset questioning deadline is reached. Specifically, when the answer's validity does not meet the preset requirements, a question follow-up strategy is used to generate follow-up questions related to the current question; when the answer's validity meets the preset requirements, a question advancement strategy is used to generate the next question based on competencies not yet covered in the job requirements, thus continuing the interview process. This approach achieves a shift from mechanical question-and-answer to proactive probing, and the final interview result is based on the actual content and validity of the questions and answers, improving the objectivity of the assessment, enhancing the accuracy of automated interviews, and reducing the need for manual review.

[0015] Furthermore, another key difference between this invention and existing technologies lies in the following: existing video interview systems typically execute standardized, mechanical video Q&A processes for all applicants, only achieving preliminary screening and failing to support in-depth recruitment decisions. In contrast, this invention targets high-potential candidates who have passed the initial resume screening and met the job assessment criteria, possessing the prerequisites for entering the in-depth evaluation stage. Simultaneously, this invention conducts the entire interview process based on the recruitment requirements of the target position. All questions are dynamically generated around the clearly defined competency requirements of the position, rather than generalizing questions before the candidate has clearly identified a suitable fit. This design ensures highly targeted interview content and effective evaluation results, extending the interview assessment from the initial screening stage to determining whether candidates who have passed the initial screening are suitable for the interview.

[0016] To further clarify, the implementing entity of this invention can be an intelligent interview platform deployed in the cloud or locally, which integrates a video interview system. Through this system, the platform can interact with candidates via video interviews in the form of a virtual digital human. Furthermore, the platform integrates functional modules such as natural language processing, emotion recognition, dialogue management, and decision reasoning, enabling it to complete the entire interview process collaboratively with the video interview system without human intervention. Specifically, this includes: dynamically generating interview questions based on job requirements; conducting video interaction; collecting multimodal data such as candidate responses and facial expressions; evaluating the effectiveness of responses based on this data; intelligently deciding whether to ask follow-up questions or switch to asking questions about different skills based on the evaluation results; and ultimately generating a structured interview result.

[0017] Next, combined Figure 1 This invention describes an interview method based on a digital human video interview system, and its specific execution steps are as follows: Figure 1 As shown, steps 101-107 are included: 101. Obtain the recruitment requirements for the target position, the candidate's resume information, and the job assessment results completed by the candidate in the online assessment system.

[0018] In the implementation scenario of this invention, the candidate has successfully passed the preliminary manual or automated resume screening stage and completed a standardized job competency assessment matching the target position in the online assessment system designated by the company. Their assessment results meet the preset qualification threshold, thus confirming them as a valid candidate qualified to enter the in-depth interview stage.

[0019] Based on this, the implementing entity (i.e., the intelligent interview platform) can obtain the recruitment requirements for the target position from the enterprise recruitment management system. These requirements cover the core competencies and specific behavioral indicators defined for the position. At the same time, the platform will retrieve the resume information of valid candidates from the candidate database. At this point, the resume information has been transformed into standardized structured data, including educational background, work experience, project experience, skill tags, and other content.

[0020] In addition, the platform can simultaneously obtain the job assessment results of valid candidates for the current target position from the online assessment system. These results include scores for each ability dimension, behavioral tendency analysis, and key answer records. These three types of data will serve as the basic input for subsequently dynamically generating interview questions and evaluating the effectiveness of the answers.

[0021] 102. Generate current interview questions based on the competency items in the recruitment requirements and ask them to the candidates, so as to collect the candidates' answers to the current interview questions and their facial emotional information during the answering process through the video interaction acquisition module.

[0022] 103. Determine the validity of the current response based on the content of the response and facial expression information.

[0023] In this invention, each question must explicitly specify the competency to be assessed. To this end, the system maintains a competency coverage status table based on the recruitment requirements of the target position. This competency coverage status table can be stored as key-value pairs or a relational structure, where the key is the identifier of the competency (e.g., competency ID), and the value is the coverage tag for that competency, which can be "not covered" or "asked." Before the interview begins, the value is "not covered." At the start of the interview, after selecting a competency as the basis for the current question, and before sending the question, the value of the corresponding competency in the competency coverage status table can be updated to "asked."

[0024] In addition, it should be noted that, considering that in actual recruitment scenarios, there are often multiple concurrent or sequential recruitment tasks, the maintenance of the competency coverage status table can be carried out in one of the following two ways: The first approach is to initialize a separate competency coverage status table for each position before initiating the video interview process. This table will only contain the competencies required for that position. Once the recruitment process for that position is complete (e.g., all candidate interviews are finished or the position is closed), the table will be deleted or archived to ensure data isolation and efficient resource management.

[0025] The second approach is to pre-build a capability coverage status table for all open positions during the system initialization phase. During the interview process, the corresponding table is invoked to track the status based on the specific position applied for by the candidate. Once the entire recruitment process for each position is completed, the status table corresponding to that position is removed from the collection.

[0026] The two maintenance mechanisms mentioned above can be flexibly selected: the first method is suitable for scenarios where the number of positions changes dynamically and lightweight operation is emphasized; the second method is suitable for batch recruitment environments where the position structure is stable and centralized scheduling is required.

[0027] Based on the above explanation, at the start of the current round of questioning, one can first query the competency coverage status table to filter out all competencies in the target position's job requirements that are still in the "uncovered" state as a candidate set. For this candidate set, one of the following two strategies can be used to determine the competencies to be assessed in this round: Strategy 1 (Weight Priority): Based on the competency weighting model for the target position, rank the candidates' competencies and select the competency with the highest weight in the candidate set as the current focus of assessment. This strategy is suitable for positions with clear priority requirements for core competencies (e.g., technical R&D positions prioritize "problem-solving ability," and management positions prioritize "teamwork ability"). It helps to prioritize the verification of key dimensions within a limited number of questions, thereby improving assessment efficiency.

[0028] Strategy Two (Comprehensive Coverage Orientation): Without emphasizing priority, randomly or sequentially select one of the "uncovered" capabilities as the current assessment item. This strategy is suitable for scenarios where the importance of each capability is relatively balanced and the system has enough rounds of questioning to ensure comprehensive coverage, ensuring that all capability dimensions are assessed equally and avoiding the omission of potentially important information due to weight bias.

[0029] Since the cutoff condition of this invention can be either full coverage of the competencies in the recruitment requirements or reaching a preset limit on the number of questions asked, the above two strategies can be dynamically configured according to the actual situation. Regardless of the strategy adopted, it can effectively improve the comprehensiveness and accuracy of the assessment.

[0030] After determining the current skills, the following two technical approaches can be used to generate corresponding interview questions: The first approach is rule-driven: First, the text descriptions of the competency items are segmented and tagged with parts of speech to identify the behavioral elements to be examined (such as "coordinate," "optimize," and "respond") and typical scenario elements (such as "cross-departmental collaboration," "customer complaints," and "urgent delivery"). Then, these elements are embedded into a predefined Behavioral Event Interview (BEI) sentence framework (e.g., "Please describe an experience you had in [scenario] [behavior]"), thereby generating grammatically correct and structurally complete open-ended questions. This method can be implemented using a rule engine or a lightweight sequence labeling model, without relying on large-scale pre-trained language models. It has the advantages of low computational overhead and logical transparency, making it particularly suitable for enterprise recruitment scenarios with high requirements for controllability and compliance.

[0031] The second approach is semantic generation: This method uses contextual embedding techniques based on pre-trained language models (such as Sentence-BERT, RoBERTa, etc.) to semantically encode the competency text, obtaining vector representations. These vector representations are then used as conditional inputs to generate open-ended questions in real-time through natural language generation models (such as finely tuned T5, BART, or dedicated text generators), containing relevant work contexts and the behaviors candidates should take. This method can more accurately capture the semantic connotations of competency items, and the generated questions are more natural and diverse. However, it requires a certain level of computing power and model maintenance, and is suitable for deployment environments with AI infrastructure that prioritize higher levels of human-like interactivity.

[0032] The difference between the two methods lies in their focus: the rule-driven method emphasizes keyword matching and template filling, while the semantic generation method emphasizes end-to-end semantic understanding and generation. In practical applications, the two methods can be flexibly selected or combined based on system resources, security policies, and user experience requirements.

[0033] During the candidate's response, the video interaction acquisition module can simultaneously capture their audio and video stream data. The response content is converted into text using speech recognition technology (such as an end-to-end ASR model or a traditional HMM-DNN hybrid system). Simultaneously, facial regions in the video stream are detected and tracked in real time, extracting temporal facial feature data as the basis for subsequent emotion analysis and nonverbal behavior assessment. Subsequently, based on the obtained response text and facial emotion information, the executing entity comprehensively determines the validity of the current response.

[0034] The validity of a response can be expressed in a quantitative scoring format, such as setting a validity score from 0 to 100. In addition, the validity of a response can also specify which dimensions the response is valid in (such as "relevance meets the standard but the structure is incomplete" or "the behavioral description is clear but the emotional expression is contradictory"). In this case, the preset validity requirements also exist in the form of multi-dimensional judgment rules (for example, requiring "relevance meets the standard and structural integrity is satisfied" to be considered valid).

[0035] Whether a scoring system or a structured assessment is used, the effectiveness evaluation and its corresponding judgment criteria should be consistent in form to ensure that the decision-making logic is clear and executable.

[0036] 104. Based on whether the validity of the answer meets the preset validity requirements, determine the target question determination strategy from the question follow-up strategy and question advancement strategy, and determine the next interview question based on the target question determination strategy, until the preset questioning deadline is reached.

[0037] In this step, to avoid potential biases in ability assessment during the interview process, a target question-setting strategy can be determined based on whether the validity of the candidate's answers meets the preset validity requirements. The next interview question can then be determined based on the target question-setting strategy, until the preset questioning deadline is reached.

[0038] The follow-up question strategy refers to the strategy used to generate follow-up questions related to the current question. These follow-up questions aim to guide candidates to supplement missing information in their answers, clarify vague statements, or verify the authenticity of their responses. In this context, it can manifest as a refinement, supplementation, or clarification of questions targeting the same competency, thereby constructing a logically coherent and progressively developing dialogue sequence.

[0039] During implementation, key elements (such as context, behavior, or result) missing in the current answer can be identified. Then, experiences or answer fragments semantically related to the question can be extracted from resumes or assessment data, and these fragments can be embedded into preset follow-up question patterns to generate targeted follow-up questions.

[0040] For example, when the interview question is "Please give an example of how you promoted cross-departmental collaboration," if the candidate only answers "completed a project with other departments," the system will identify the missing elements of "behavior" and "results," and extract the experience of "leading the marketing and R&D team to implement a user growth plan in 2023" from the resume. It will then embed this experience into a preset sentence and generate follow-up questions: "You mentioned in your resume that you led the marketing and R&D team to implement a user growth plan. How did you coordinate the differences in goals between the two parties at that time? What were the final results?"

[0041] In addition, the current question, the answer, and the associated resume or assessment content can be combined into a contextual input, and follow-up questions can be output through a lightweight natural language generation model. During the generation process, keywords or question types can be constrained to ensure focus.

[0042] For example, when the question is "Describe an experience you had handling a customer complaint," if a candidate's answer is vague, such as "The customer was dissatisfied, and I calmed him down," the system will combine the question, the answer, and the candidate's response to the job assessment question about "actively listening and escalating the issue in a service scenario" into a context. This context will be input into a lightweight generation model, and the system will output a follow-up question: "You mentioned calming down the dissatisfied customer. Could you elaborate on how you understood the customer's needs and took subsequent actions, based on your 'active listening' practice mentioned in the assessment?" The question progression strategy refers to the strategy used to generate interview questions corresponding to the next competency item.

[0043] For example, if the capability item "Resilience" in the coverage status table is "Not Asked," it can be identified as the next capability item. The corresponding description for this capability item is: Maintaining high-efficiency output under high-intensity work conditions. A question can be generated based on the rule-driven method or semantic generation method described above: "Please describe an experience where you maintained high-efficiency output under high-intensity work conditions." 105. Generate interview results based on all interview questions, corresponding answers, and the validity of the answers.

[0044] In this step, all interview questions from each round, their corresponding answers, and the effectiveness of each response can be compiled as the foundational data for generating interview results. Interview results can take several forms: firstly, a comprehensive score can be compared to a pre-set pass / fail threshold to output a "pass" or "fail" conclusion; secondly, an interview report can be presented, including hiring recommendations, a summary of key behavioral evidence (such as citing specific question-and-answer snippets to illustrate a particular ability), and potential risk warnings (such as contradictory answers or abnormal emotional states). These two forms can be used individually or in combination to meet the needs of different recruitment scenarios regarding decision-making efficiency and assessment depth.

[0045] Based on the above Figure 1As can be seen from the implementation method, the interview method based on a video interview system provided by this invention uses the recruitment requirements of the target position, the candidate's resume information, and the job assessment results as pre-interview preparation. Secondly, instead of relying on a fixed question bank, it dynamically generates initial interview questions based on the competency items in the recruitment requirements and asks them to the candidate. During the candidate's response, a video interaction acquisition module collects the candidate's answers to the current interview questions and their facial emotional information. Furthermore, the validity of the current answer is objectively determined based on the answer content and facial emotional information. Based on this, a target question determination strategy is determined in the question follow-up strategy and question progression strategy according to whether the answer validity meets the preset validity requirements. The next interview question is then determined according to the target question determination strategy until the preset questioning deadline is reached. Finally, the interview result is generated based on all interview questions, answer content, and answer validity. Compared with existing technologies, this mechanism dynamically links the generation of interview questions with the validity of the candidate's answers, avoiding the ability assessment bias that may be caused by asking questions in a fixed order or in a random manner. The final interview result is based on the actual question and answer content and its validity, improving the objectivity of the assessment and reducing the reliance on manual review to a certain extent.

[0046] Furthermore, as a response to Figure 1 Further refinement and extension of the illustrated embodiments, this invention also provides another interview method based on a video interview system, such as... Figure 2 As shown, the specific steps are as follows: 201. Obtain the recruitment requirements for the target position, the candidate's resume information, and the job assessment results completed by the candidate in the online assessment system.

[0047] 202. Generate current interview questions based on the competency items in the recruitment requirements and ask them to the candidates, so as to collect the candidates' answers to the current interview questions and their facial emotional information during the answering process through the video interaction acquisition module.

[0048] The implementation methods of steps 201-202 are the same as those of steps 101-102, and can achieve the same technical effect and solve the same technical problem, so they will not be repeated here.

[0049] 203. Determine the validity of the current response based on the content of the response and facial emotion information.

[0050] In this embodiment, semantic similarity calculation methods (such as the Sentence-BERT model) or keyword matching methods can be used to quantify the relevance between the answer content and the current interview question, and output as a score. The semantic similarity calculation method first preprocesses the text of the current interview question and the candidate's answer, including sentence segmentation, removal of irrelevant symbols, and standardization of expressions. Then, the two preprocessed texts are input into the same semantic encoding model (such as Sentence-BERT), which generates fixed-dimensional semantic vectors (e.g., 768-dimensional) for the question and answer respectively. Next, the cosine similarity between these two vectors is calculated, and the result is mapped to a preset score range (e.g., 0.0–1.0 or 0–100 points) as the final relevance score. This score objectively reflects the degree of semantic fit between the answer and the question, providing a quantitative basis for subsequent effectiveness evaluation.

[0051] For example, a semantic similarity score of 0.87 between the answer and the question indicates a high degree of relevance; a score of 0.32 indicates insufficient relevance. This score can be directly used as a quantitative indicator in the evaluation of answer validity, participating in subsequent weighted calculations or threshold determinations.

[0052] In this embodiment of the invention, a pre-built keyword mapping library for capability items can be included. This library uses the capability item's ID as the key and the corresponding keyword as the value. After a candidate generates their answer, the answer text can be segmented and standardized. Then, the number of keywords matched in the answer content is counted to calculate a score, which serves as the final relevance score. This score reflects the degree to which the answer covers the question's intent.

[0053] The scoring process involves counting the number of keywords matched in the answer, then determining the corresponding score based on a preset mapping rule within a specified range. For example, matching 0 keywords earns 0 points, matching 1-2 earns 30 points, matching 3-4 earns 60 points, and matching 5 or more earns 100 points. Finally, this score is used as a relevance score and normalized. This score reflects the degree to which the answer covers the key semantic elements and the question's intent.

[0054] Subsequently, a pre-defined behavioral event model can be used to ensure the structural integrity of the responses. This model, based on the Behavioral Event Interview (BEI) paradigm, defines an effective response as one that must cover four core elements: context, task, action, and outcome. In practice, a rule-matching method is employed: a corresponding set of keywords and typical sentence templates are pre-defined for each element; after segmenting the response text into sentences, keywords and grammatical patterns of each element are matched sentence by sentence, and the number of covered elements is counted to identify missing items (e.g., "no outcome mentioned"). Finally, the structural integrity assessment result is output, for example, using a score between 0 and 1 to represent the degree of coverage, or directly indicating the specific type of missing element, to facilitate subsequent effectiveness evaluation and follow-up questioning decisions.

[0055] The pre-defined behavioral event model can be pre-analyzed by business experts based on the answers of employees who ultimately passed the probationary period in numerous real interviews. This analysis summarizes the typical components of an effective response: what happened (situation), what you were responsible for (task), what you did (action), and the final result (outcome). Then, for each part, common keywords and typical expressions are compiled; for example, "The project was very urgent at the time" is a situation, and "I led the work on..." is an action. This content is then organized into rules, written into the system, and forms a fixed judgment model.

[0056] Subsequently, the facial video stream captured during the face-to-face video interaction can be processed, and a pre-trained facial expression recognition model can be used to extract the emotion category (such as happy, neutral, angry, sad, etc.) and its confidence level for each frame or specific time period. Then, by using temporal aggregation methods (such as taking the dominant emotion during the answer period or a weighted average of emotions at each time point), the candidate's final facial emotional state during the answer to the question can be determined, such as "positive" (corresponding to happy / satisfied) or "negative" (corresponding to angry / anxious / sad).

[0057] Meanwhile, for the response text obtained from speech recognition, a sentiment analysis model (such as a sentiment classifier based on dictionary rules or lightweight BERT) is used to determine the sentiment polarity of the text and output the sentiment tendency of the response content, which is usually divided into "positive", "neutral" or "negative".

[0058] Then, the two sets of results are compared for consistency: if the facial emotional state and the text sentiment are both "positive" or both "negative," they are considered consistent, and the consistency result is set to the first preset value (e.g., 1.0 or "consistent"); if one is positive and the other is negative (e.g., the answer expresses optimism but the face shows obvious negativity), they are considered inconsistent, and the consistency result is set to the second preset value (e.g., 0.0 or "inconsistent"); for cases where either is "neutral," the strategy can be set to "consistent." This consistency result serves as an important dimension of the answer's validity, used to identify risks such as candidate inconsistencies in speech and behavior, emotional suppression, or distorted answers, thereby improving the credibility of the assessment.

[0059] Finally, the validity of the current response can be determined based on relevance, structural integrity, and consistency between emotional state and affective tendency. Specifically, this can be achieved by: First, based on a preset weighting method, the relevance, structural integrity, and consistency results are assigned corresponding first, second, and third weights, respectively. Then, the relevance, structural integrity, and consistency results are weighted and summed to obtain the validity score, which serves as the validity of the response. In this way, instead of relying on a single indicator, semantic relevance, behavioral structural integrity, and consistency are organically integrated, thereby providing more reliable and effective evaluation results.

[0060] Among them, the preset weight allocation method refers to the weight determination method used in the evaluation of the effectiveness of the response, which obtains the initial weight from the preset configuration library according to the target job type, and only reduces the consistency weight when there is inconsistency in emotion or language, while keeping the other weights unchanged and without normalization.

[0061] To further clarify, when determining the first, second, and third weights according to the preset weight allocation method, the corresponding first initial weight (corresponding to relevance), second initial weight (corresponding to structural integrity), and third initial weight (corresponding to emotion-language consistency) can be obtained from the preset weight configuration library based on the target job type. For example, for technical R&D positions, the relevance of the answer to the question is emphasized, and its first initial weight is greater than the second and third initial weights; while for customer service positions, more attention is paid to the consistency of the candidate's words and actions and the authenticity of their emotional expression, and its third initial weight is greater than the first and second initial weights.

[0062] Based on this, if the consistency result is "inconsistent", the third initial weight is adjusted down by a preset ratio (such as a reduction of 30%) to obtain the adjusted third weight, while the first weight and the second weight remain as the first initial weight and the second initial weight, respectively, so that the other weights are not adjusted up to compensate for the reduction of the third weight.

[0063] It is important to note that this processing method differs significantly from existing technologies: In traditional evaluation models, weights typically need to satisfy normalization constraints (i.e., the sum of all weights is 1), and when one weight is adjusted, the remaining weights are redistributed accordingly. However, in this invention, weight adjustment does not require normalization; each dimension independently retains its original dimensions and business meaning, thus more accurately reflecting the actual importance of different evaluation dimensions in a specific position and avoiding evaluation bias caused by forced normalization. If the consistency result is "consistent," the first initial weight, second initial weight, and third initial weight are directly used as the first weight, second weight, and third weight, respectively, without adjustment.

[0064] The reason for adopting this non-normalized weight adjustment mechanism is that when emotional state and linguistic sentiment are inconsistent, it indicates that the candidate may be concealing information, nervous, expressing themselves insincerely, or providing unreliable content in that round of responses. In such cases, the influence of the "consistency" dimension on overall validity should be reduced to reflect the decreased credibility. However, relevance and structural integrity themselves do not become invalid—even if the candidate's emotions are abnormal, whether their answers are closely related to the question and whether they contain complete behavioral events remains an independent and valid assessment criterion. Forcibly normalizing the weights (e.g., automatically increasing the first and second weights after reducing the third weight) would artificially amplify the originally reasonable relevance or structural scores, distorting the true contribution of each dimension and thus affecting the objectivity of the assessment.

[0065] 204. If the validity of the answer does not meet the preset validity requirements, the follow-up question strategy will be determined as the target question determination strategy, and the next interview question will be determined based on the follow-up question strategy. The follow-up question strategy is to combine the resume information or job assessment results with the content related to the current interview question to generate the corresponding follow-up question as the next interview question.

[0066] In this step, if the validity of the answer does not meet the preset validity requirements, a follow-up question will be generated based on the resume information or job assessment results that are relevant to the current interview question, and will be used as the next interview question.

[0067] In this step, joint semantic analysis methods can be used to collaboratively analyze the current interview question and the candidate's answer to identify the key elements that are missing from the answer in the behavioral event structure (such as the STAR model) relative to the question requirements. The missing elements include one or more of the task, action, and result.

[0068] In practice, the interview questions and answers are first aligned, and the answers are judged based on the preset behavioral event template to determine whether the answers fully cover the behavioral dimensions implied by the questions. If any missing elements are found (for example, the question asks for a description of "how to resolve the conflict", but the answer only states "there is a conflict" without mentioning "what measures were taken" or "the final result"), then the missing elements are marked.

[0069] Subsequently, retrieve relevant experience segments from the candidate's resume information that relate to the competencies involved in the current question (e.g., match project descriptions in work experience by keywords related to competency items), and extract their answer records for the corresponding competency dimensions from the job assessment results (e.g., the selection of situational judgment questions or self-description content).

[0070] Finally, the retrieved experience information or assessment answers are used as factual anchors. Combined with the identified missing elements, pre-set follow-up questions are embedded (such as "You mentioned participating in the XX project. What specific actions were taken to address the existing conflicts?") to generate semantically coherent and context-connected supplementary questions. These supplementary questions not only focus on filling information gaps but also form a logically progressive dialogue sequence with the original questions, thereby enabling a deeper exploration of the candidate's abilities.

[0071] Therefore, missing elements can serve as a basis for identifying the information dimensions that need to be supplemented, and ensure that the generated follow-up questions are targeted, that is, missing elements can directly point to key behavioral details that the candidate has not fully elaborated.

[0072] In addition, the pre-set follow-up question format dynamically fills the extracted answer records and missing elements into a language template with a pre-set logic and syntax structure, thereby automatically generating a follow-up question.

[0073] 205. If the validity of the answer meets the preset validity requirements, the question progression strategy will be determined as the target question determination strategy, and the next interview question will be determined based on the question progression strategy. The question progression strategy is to generate the next interview question based on the ability items not covered in the recruitment requirements.

[0074] In some examples of this step, the skills that have not yet been asked about are filtered out from all the skills included in the job requirements to form a set of target skills; then, a skill is randomly selected from this set as the current target skill.

[0075] Next, a context embedding method based on a pre-trained language model (such as Sentence-BERT or RoBERTa) is used to semantically encode the text description of the ability item to obtain a vector representation; then the vector is input into a natural language generation model (such as a finely tuned T5 or BART) to generate the next interview question.

[0076] By randomly selecting uncovered capability items, we ensure that all capability dimensions are examined equally and unbiasedly in the process, avoiding blind spots in evaluation caused by fixed order or subjective priorities; at the same time, by using semantic encoding and generative models, we ensure that the content of the questions closely matches the connotation of the capability items.

[0077] In other examples, it is also possible to filter out the key competencies defined in the job requirements that have not yet been asked about, and select one competency whose importance is higher than a preset value as the current target competency.

[0078] 206. For each competency item in the recruitment requirements, the validity scores of all answers related to the competency item are aggregated and calculated to obtain the comprehensive score for the competency item; 208. The interview results are obtained based on the comprehensive scores of each ability item.

[0079] In steps 206 and 207, for each competency item in the recruitment requirements, the validity scores of the answers to all corresponding interview questions (including initial questions and possible follow-up questions) can be aggregated (e.g., by taking the average, weighted average, or maximum value) to obtain the comprehensive score for that competency item. Subsequently, based on the job weights pre-set for each competency item in the recruitment requirements, the comprehensive scores of all competency items are weighted and summed to calculate the total interview score. If the total interview score exceeds the preset passing threshold, the interview result is determined to be "passed"; otherwise, it is "failed".

[0080] It needs to be clarified that the "response validity score" aggregated here does not merely reflect whether the response is complete or the language is fluent, but rather incorporates an assessment of the substantive content of the competency performance during the validity determination process. Specifically, in determining response validity as described above, relevance analysis not only determines whether the response is irrelevant but also identifies whether the response contains key behaviors that match the target competency; structural integrity analysis examines whether specific actions and results reflecting the competency level are presented; and emotion-language consistency helps determine whether the stated behavior is credible.

[0081] Therefore, a highly effective response is essentially a carrier of evidence of ability that simultaneously meets the conditions of "content relevance, structural completeness, specific behavior, and credible expression." Since the calculation of the effectiveness score incorporates a quantitative assessment of these ability performance elements, its value is designed to be positively correlated with the strength of the candidate's actual ability demonstration. Precisely because of this, if a candidate lacks practical experience or behavioral support related to the target ability, it will be difficult to provide a satisfactory response; their response will be insufficient in dimensions such as relevance, completeness, or consistency, ultimately leading to a low effectiveness score.

[0082] Furthermore, as a response to the above Figure 1In addition to the implementation of the method shown, this embodiment of the invention also provides an interview device based on a video interview system, used for the above-mentioned... Figure 1 The method shown is implemented accordingly. This device embodiment corresponds to the foregoing method embodiment. For ease of reading, this device embodiment will not repeat the details of the foregoing method embodiment, but it should be clear that the device in this embodiment can implement all the contents of the foregoing method embodiment. Figure 3 As shown, the device includes: Information acquisition unit 301 is used to acquire the recruitment requirements of the target position, the candidate's resume information, and the job assessment results completed by the candidate in the online assessment system; The first question-posing unit 302 is used to generate current interview questions based on the ability items in the recruitment requirements obtained by the information acquisition unit 301 and ask them to the candidate, so as to collect the candidate's answers to the current interview questions and facial emotion information during the answering process through the video interaction acquisition module. The validity assessment unit 303 is used to determine the validity of the current answer based on the answer content and facial emotion information obtained from the first question-asking unit 302; The second question-posing unit 304 is used to determine whether the validity of the answer obtained by the validity evaluation unit 303 meets the preset validity requirements, determine the target question-determining strategy in the question follow-up strategy and question advancement strategy, and determine the next interview question according to the target question-determining strategy until the preset questioning deadline requirement is reached. Result determination unit 305 is used to generate interview results from all interview questions, answers, and answer validity.

[0083] Furthermore, as a response to the above Figure 2 In addition to the implementation of the method shown, this embodiment of the invention also provides another interview device based on a video interview system, used for the above-mentioned... Figure 2 The method shown is implemented accordingly. This device embodiment corresponds to the foregoing method embodiment. For ease of reading, this device embodiment will not repeat the details of the foregoing method embodiment, but it should be clear that the device in this embodiment can implement all the contents of the foregoing method embodiment. Figure 4 As shown, the device includes: Information acquisition unit 301 is used to acquire the recruitment requirements of the target position, the candidate's resume information, and the job assessment results completed by the candidate in the online assessment system; The first question-posing unit 302 is used to generate current interview questions based on the ability items in the recruitment requirements obtained by the information acquisition unit 301 and ask them to the candidate, so as to collect the candidate's answers to the current interview questions and facial emotion information during the answering process through the video interaction acquisition module. The validity assessment unit 303 is used to determine the validity of the current answer based on the answer content and facial emotion information obtained from the first question-asking unit 302; The second question-posing unit 304 is used to determine whether the validity of the answer obtained by the validity evaluation unit 303 meets the preset validity requirements, determine the target question-determining strategy in the question follow-up strategy and question advancement strategy, and determine the next interview question according to the target question-determining strategy until the preset questioning deadline requirement is reached. Result determination unit 305 is used to generate interview results from all interview questions, answers, and answer validity.

[0084] In one optional implementation, the effectiveness evaluation unit 303 is specifically used for: The relevance between the answer content and the current interview question is determined using a relevance determination method; The structural integrity of the response content is determined using a pre-defined behavioral event model; The facial emotion information is analyzed using a facial expression recognition model to obtain the candidate's emotional state; The sentiment analysis model is used to analyze the response content to obtain the sentiment tendency of the response content; The validity of the current response is determined based on the correlation, structural integrity, and consistency between the emotional state and the sentiment tendency.

[0085] In one optional implementation, when the validity assessment unit 303 determines the validity of the current response based on the relevance, structural integrity, and consistency between emotional state and affective tendency, it is specifically used to: According to the preset weighting method, the correlation, structural integrity and consistency results are assigned corresponding first weights, second weights and third weights respectively. The relevance, structural integrity, and consistency results are weighted and summed to obtain a validity score as the validity of the response. The consistency result is set to a first preset value when consistent and a second preset value when inconsistent.

[0086] In one optional implementation, when the validity evaluation unit 303 assigns corresponding first weights, second weights, and third weights to the correlation, structural integrity, and consistency results according to a preset weight allocation method, it is specifically used for: Based on the job type of the target position, the corresponding first initial weight, second initial weight, and third initial weight are obtained from the preset weight configuration library. Among them, the first initial weight of technical R&D positions is greater than the second initial weight and the third initial weight, and the third initial weight of customer service positions is greater than the first initial weight and the second initial weight. If the consistency result is inconsistent, the weight value of the third initial weight is reduced by a preset ratio to obtain the third weight, and the first initial weight is used as the first weight and the second initial weight is used as the second weight. If the consistency result is consistent, then the first initial weight, the second initial weight, and the third initial weight are respectively used as the first weight, the second weight, and the third weight.

[0087] In one optional implementation, the second problem-posing unit 304 includes: The unqualified question raising module 3041 is used to determine the question follow-up strategy as the target question determination strategy if the validity of the answer does not meet the preset validity requirements, and to determine the next interview question according to the question follow-up strategy. The question follow-up strategy is to combine the resume information or job assessment results related to the current interview question to generate corresponding follow-up questions as the next interview question. The target question generation module 3042 is used to determine the question progression strategy as the target question determination strategy if the validity of the answer meets the preset validity requirements, and to determine the next interview question according to the question progression strategy. The question progression strategy is to generate the next interview question based on the ability items not covered in the recruitment requirements.

[0088] In one optional implementation, the non-compliance issue raising module 3041 is specifically used for: The joint semantic analysis method is used to jointly analyze the current interview questions and the candidates' answers to identify the missing elements in the structural integrity of the answers relative to the interview questions. The missing elements include tasks, actions, and results. Retrieve experience information related to the competency items corresponding to the current interview question from the resume information, and extract answer records related to the corresponding competency items from the job assessment results; Based on the experience information and the answer record, and combined with the missing elements, additional questions corresponding to the current interview question are generated. The additional questions and the current interview question form a semantically coherent dialogue sequence.

[0089] In one optional implementation, the compliance problem raising module 3042 is specifically used for: Among all the competency items in the recruitment requirements, determine the set of target competency items that were not asked about; Randomly select one target capability from the set of target capability items as the current target capability item; A context embedding method based on a pre-trained language model is used to semantically encode the text description of the current target capability item to obtain the corresponding vector representation; Based on this vector representation, a natural language generation model is used to generate corresponding interview questions. The interview questions include descriptions of the work situation related to the target competency and descriptions of the behaviors that the candidate should take.

[0090] In one optional implementation, the result determination unit 305 is specifically used for: For each competency item in the recruitment requirements, the validity scores of all answers related to the competency item are aggregated and calculated to obtain the comprehensive score for the competency item; Based on the weights assigned to each competency item in the recruitment requirements, the overall scores of all competency items are weighted and summed to obtain a total interview score. If the total interview score exceeds the preset passing threshold, the interview result is determined to be passed.

[0091] Furthermore, embodiments of the present invention also provide a storage medium for storing a computer program, wherein the computer program, when running, controls the device where the storage medium is located to execute the above-described... Figure 1-2 The interview method based on a video interview system described in the article.

[0092] Furthermore, embodiments of the present invention also provide a processor for running a program, wherein the program executes the above-described... Figure 1-2 The interview method based on a video interview system described in the article.

[0093] In the above embodiments, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions in other embodiments.

[0094] It is understood that the relevant features in the above methods and apparatus can be referenced interchangeably. Furthermore, the terms "first," "second," etc., in the above embodiments are used to distinguish between embodiments and do not represent the superiority or inferiority of any particular embodiment.

[0095] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0096] 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 required structure for constructing such systems is apparent from the above description. Furthermore, this invention is not directed to any particular programming language. It should be understood that the contents of the invention described herein can be implemented using various programming languages, and the above description of specific languages ​​is for the purpose of disclosing the best mode of implementation of the invention.

[0097] In addition, the memory may include non-permanent memory in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM, and the memory includes at least one memory chip.

[0098] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0099] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0100] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0101] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0102] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0103] Memory may include non-persistent memory in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. Memory is an example of computer-readable media.

[0104] Computer-readable media includes both permanent and non-permanent, removable and non-removable media that can store information using any method or technology. Information can be computer-readable instructions, data structures, modules of programs, 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 technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic magnetic disk storage or other magnetic storage devices, or any other non-transferable medium that can be used to store information accessible by a computing device. As defined herein, computer-readable media does not include transient computer-readable media, such as modulated data signals and carrier waves.

[0105] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element.

[0106] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0107] The above are merely embodiments of this application and are not intended to limit the scope of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of the claims of this application.

Claims

1. An interview method based on a video interview system, characterized in that, The method includes: Obtain the recruitment requirements for the target position, the candidate's resume information, and the candidate's job assessment results completed in the online assessment system; Based on the competency items in the recruitment requirements, the current interview questions are generated and asked to the candidate, so as to collect the candidate's answers to the current interview questions and facial emotion information during the answering process through the video interaction acquisition module; The validity of the current response is determined based on the content of the response and facial emotion information. Based on whether the validity of the answer meets the preset validity requirements, a target question determination strategy is determined from the question follow-up strategy and question advancement strategy, and the next interview question is determined based on the target question determination strategy until the preset questioning deadline is reached; The interview results are generated based on all interview questions, answers, and the validity of the answers.

2. The method according to claim 1, characterized in that, Based on the content of the response and facial emotion information, the validity of the current response is determined, including: The relevance between the answer content and the current interview question is determined using a relevance determination method; The structural integrity of the response content is determined using a pre-defined behavioral event model; The facial emotion information is analyzed using a facial expression recognition model to obtain the candidate's emotional state; The sentiment analysis model is used to analyze the response content to obtain the sentiment tendency of the response content; The validity of the current response is determined based on the correlation, structural integrity, and consistency between the emotional state and the sentiment tendency.

3. The method according to claim 2, characterized in that, The validity of the current response is determined based on the results of relevance, structural integrity, and consistency between emotional state and affective tendency, including: According to the preset weighting method, the correlation, structural integrity and consistency results are assigned corresponding first weights, second weights and third weights respectively. The relevance, structural integrity, and consistency results are weighted and summed to obtain a validity score as the validity of the response. The consistency result is set to a first preset value when consistent and a second preset value when inconsistent.

4. The method according to claim 3, characterized in that, According to a preset weighting method, the correlation, structural integrity, and consistency results are assigned corresponding first, second, and third weights, including: Based on the job type of the target position, the corresponding first initial weight, second initial weight, and third initial weight are obtained from the preset weight configuration library. Among them, the first initial weight of technical R&D positions is greater than the second initial weight and the third initial weight, and the third initial weight of customer service positions is greater than the first initial weight and the second initial weight. If the consistency result is inconsistent, the weight value of the third initial weight is reduced by a preset ratio to obtain the third weight, and the first initial weight is used as the first weight and the second initial weight is used as the second weight. If the consistency result is consistent, then the first initial weight, the second initial weight, and the third initial weight are respectively used as the first weight, the second weight, and the third weight.

5. The method according to claim 1, characterized in that, Based on whether the validity of the answer meets the preset validity requirements, a target question-setting strategy is determined from the question follow-up strategy and question progression strategy. The next interview question is then determined based on the target question-setting strategy until the preset questioning deadline is reached, including: If the validity of the answer does not meet the preset validity requirements, the follow-up question strategy is determined as the target question determination strategy, and the next interview question is determined according to the follow-up question strategy. The follow-up question strategy is to combine the resume information or job assessment results related to the current interview question to generate a corresponding follow-up question as the next interview question. If the validity of the answer meets the preset validity requirements, the question progression strategy is determined as the target question determination strategy, and the next interview question is determined according to the question progression strategy. The question progression strategy is to generate the next interview question based on the ability items not covered in the recruitment requirements.

6. The method according to claim 5, characterized in that, Based on the resume information or job assessment results, and considering the relevant content from the current interview question, generate corresponding follow-up questions as the next interview question, including: The joint semantic analysis method is used to jointly analyze the current interview questions and the candidates' answers to identify the missing elements in the structural integrity of the answers relative to the interview questions. The missing elements include tasks, actions, and results. Retrieve experience information related to the competency items corresponding to the current interview question from the resume information, and extract answer records related to the corresponding competency items from the job assessment results; Based on the experience information and the answer record, and combined with the missing elements, additional questions corresponding to the current interview question are generated. The additional questions and the current interview question form a semantically coherent dialogue sequence.

7. The method according to claim 5, characterized in that, Generate the next interview question based on the skills not covered in the job requirements, including: Among all the competency items in the recruitment requirements, determine the set of target competency items that were not asked about; Randomly select one target capability from the set of target capability items as the current target capability item; A context embedding method based on a pre-trained language model is used to semantically encode the text description of the current target capability item to obtain the corresponding vector representation; Based on this vector representation, a natural language generation model is used to generate corresponding interview questions. The interview questions include descriptions of the work situation related to the target competency and descriptions of the behaviors that the candidate should take.

8. The method according to claim 1, characterized in that, The interview results are generated based on all interview questions, answers, and the validity of the answers, including: For each competency item in the recruitment requirements, the validity scores of all answers related to the competency item are aggregated and calculated to obtain the comprehensive score for the competency item; Based on the weights assigned to each competency item in the recruitment requirements, the overall scores of all competency items are weighted and summed to obtain a total interview score. If the total interview score exceeds the preset passing threshold, the interview result is determined to be passed.

9. An interview device based on a video interview system, characterized in that, The device includes: The information acquisition unit is used to acquire the recruitment requirements of the target position, the candidate's resume information, and the job assessment results completed by the candidate in the online assessment system. The first question-posing unit is used to generate current interview questions based on the ability items in the recruitment requirements obtained by the information acquisition unit and ask them to the candidate, so as to collect the candidate's answers to the current interview questions and facial emotion information during the answering process through the video interaction acquisition module. The validity assessment unit is used to determine the validity of the current answer based on the answer content and facial emotion information obtained from the first question-asking unit. The second question-posing unit is used to determine the target question-determining strategy from the question follow-up strategy and question advancement strategy based on whether the validity of the answer obtained by the validity evaluation unit meets the preset validity requirements, and to determine the next interview question based on the target question-determining strategy until the preset questioning deadline is reached. The results determination unit is used to generate interview results from all interview questions, answers, and the validity of the answers.

10. A storage medium, characterized in that, The storage medium includes a stored program, wherein, when the program is executed, it controls the device where the storage medium is located to perform the interview method based on the video interview system as described in any one of claims 1 to 8.

11. A processor, characterized in that, The processor is used to run a program, wherein the program executes the interview method based on the video interview system as described in any one of claims 1 to 8.

Images