An interview answer state monitoring method and system

Abstract

The application relates to the technical field of data processing, and discloses an interview answer state monitoring method and system, which analyzes and collates an interview request sent by an interview request end for target recruitment information, generates an interview task list containing interview appointment window time and interview post associated information, acquires idle time periods of a plurality of GPU servers, considers interview appointment window time in the interview task list and linked interview GPU demand information and interview monitoring demand information, solves corresponding interview appointment time periods of the GPU servers to which each interview request end is allocated, and drives each GPU server to execute interview answer state monitoring according to received interview answer video streams. Therefore, the application matches GPU resources with computing power adaptation and delay compliance for each interview request, solves the contradiction between delay caused by computing power overallocation and resource waste caused by low computing power allocation, ensures user experience, and improves the benefits of third-party service institutions as much as possible.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of data processing, and in particular to an interview answer state monitoring method and system. BACKGROUND

[0002] The interview answer state monitoring is a technical process of real-time collection and dynamic analysis of the behavior, language, environment and answer progress of the interviewee in the answering process based on multi-modal perception technology and artificial intelligence analysis. The core is to obtain multi-dimensional data through the interview terminal, including picture data, voice data, behavior data and answer data in the video stream, and then combine the AI monitoring model customized for the post (such as the code operation analysis model for technical positions and the communication appeal analysis model for sales positions) to judge the compliance (such as whether cheating or not focused) and effectiveness (such as logical thinking ability and professional ability matching degree) of the answer state in real time, and output the state analysis results synchronously to provide objective evaluation for the interviewers.

[0003] Since the recruitment enterprises usually do not specially equip the GPU server resources for performing the interview answer state monitoring, the third-party service agencies need to face the auxiliary interview demands of multiple recruitment enterprises in the current industry. In the scenario of having a large number of recruitment interview demands (such as the regular recruitment interview plan of large state-owned enterprises and group enterprises), the recruitment enterprises need to complete a large number of interview tasks in a short time, and the third-party service agencies need to consider the recruitment post demand, recruitment time period demand and recruitment quantity demand of each recruitment enterprise when providing the interview answer state monitoring service, so that the existing technology has the problems of difficulty in dynamically matching the GPU computing power demand, user experience degradation caused by over-provisioning of computing power, resource cost waste caused by low-provisioning of computing power, and high allocation complexity under dynamic demand, which affects the user experience (such as high delay of the monitoring analysis results fed back by the interview answer state monitoring model, which cannot provide real-time analysis and evaluation for the interviewers) and the benefits of the third-party service agencies (low-provisioning of computing power reduces the number of recruitment enterprises served by the third-party service agencies under the premise of limited GPU server resources).

[0004] Therefore, how to improve the intelligent allocation of the GPU server resources of the third-party service agencies when facing the auxiliary interview demands of multiple recruitment enterprises, and improve the benefits of the third-party service agencies as much as possible on the basis of ensuring the user experience, is a technical problem to be solved. SUMMARY

[0005] The present application provides an interview answer state monitoring method and system, which aims to solve at least one of the above technical problems.

[0006] To achieve the above-mentioned purpose, the present application provides an interview answer state monitoring method, which comprises the following steps:

[0007] receive the interview request sent by the interview request end for the target recruitment information, and construct an interview task list according to the interview appointment window time in the interview request and the interview post association information;

[0008] generate interview GPU demand information according to the interview answer state monitoring demand corresponding to the interview post association information, and link the interview GPU demand information to the interview task list as first additional demand information;

[0009] generate interview monitoring demand information according to the interview question content corresponding to the interview post association information, and link the interview monitoring demand information to the interview task list as second additional demand information;

[0010] obtain an interview plan period of the target recruitment information, query the idle period of the plurality of GPU servers in the interview plan period, consider the interview appointment window time in the interview task list and the first and second additional demand information linked thereto, and solve an interview request allocation strategy;

[0011] based on the interview appointment period corresponding to the GPU server to which each interview request end is allocated in the interview request allocation strategy, drive the GPU server to perform interview answer state monitoring on the interview request end.

[0012] Optionally, before the step of receiving the interview request sent by the interview request end for the target recruitment information, and constructing an interview task list according to the interview appointment window time in the interview request and the interview post association information, the method further comprises:

[0013] obtain the target recruitment information sent by the recruitment enterprise, and extract the recruitment interview demand in the target recruitment information; wherein the recruitment interview demand comprises a recruitment interview post, a recruitment quantity of each recruitment interview post, and an interview request delivery period;

[0014] determine a first generation judgment condition of the interview task list based on the recruitment interview post and the recruitment quantity of each recruitment interview post in the recruitment interview demand, and determine a second generation judgment condition of the interview task list based on the interview request delivery period in the recruitment interview demand.

[0015] Optionally, the step of receiving the interview request sent by the interview request end for the target recruitment information, and constructing an interview task list according to the interview appointment window time in the interview request and the interview post association information, specifically comprises:

[0016] receive the interview request sent by the interview request end for the target recruitment information, and extract the request submission time, the interview appointment window time and the interview post association information of the interview request;

[0017] According to the first generation judgment condition determined by the interview post associated information of the interview request and the recruitment interview post and recruitment quantity in the recruitment interview demand, it is judged whether the interview request receiving quantity of the recruitment interview post matched with the interview post associated information reaches the corresponding recruitment quantity, if yes, it is determined that the interview request meets the first generation judgment condition;

[0018] According to the second generation judgment condition determined by the request submission time of the interview request and the interview request delivery time period in the recruitment interview demand, it is judged whether the request submission time of the interview request is located in the interview request delivery time period, if yes, it is determined that the interview request meets the second generation judgment condition;

[0019] The interview appointment window time in the received interview request meeting the first generation judgment condition and the second generation judgment condition is added to the constructed interview task list according to the interview post associated information.

[0020] Optionally, according to the interview answer state monitoring demand corresponding to the interview post associated information, the interview GPU demand information is generated and linked to the interview task list as the first additional demand information, specifically including:

[0021] Based on the interview post type and the interview post ability requirement set in the interview post associated information, the interview answer state monitoring demand is determined; wherein, the interview answer state monitoring demand includes several interview answer state monitoring projects;

[0022] According to the video analysis test data of several GPU servers for interview answer video analysis, the GPU algorithm requirement parameter and the video analysis result feedback delay of the GPU server candidate set meeting the interview experience requirement when executing the same interview answer state monitoring project are determined;

[0023] Among them, the GPU server candidate set includes several GPU servers whose video analysis result feedback delay parameter in the video analysis test data is lower than the preset delay threshold of the interview experience requirement;

[0024] The GPU server candidate set and the GPU algorithm requirement parameter are summarized to generate the interview GPU demand information, and the interview GPU demand information is linked to the interview task list as the first additional demand information of the corresponding interview request.

[0025] Optionally, based on the interview post type and the interview post ability requirement set in the interview post associated information, the interview answer state monitoring demand step specifically includes:

[0026] query an interview answer state monitoring model database based on the interview post type and the interview post ability requirement set in the interview post association information to obtain an interview answer state monitoring model corresponding to the interview post type and the interview post ability requirement set;

[0027] query a plurality of interview answer state monitoring items covered in the interview answer state monitoring model as interview answer state monitoring requirements corresponding to the interview request.

[0028] Optionally, interview monitoring requirement information is generated based on a plurality of interview question contents corresponding to the interview post association information, and is connected to the interview task list step as second additional requirement information, specifically including:

[0029] extract a post type requirement keyword of the interview post type and the interview post ability requirement set in the interview post association information, and match a plurality of interview question contents with the highest correlation degree with the post type requirement keyword in the corresponding interview question content set;

[0030] According to the interview answer time length requirement recorded in each interview question content, the interview answer estimated total time length of each interview request is estimated, and the interview monitoring requirement information is converted as the second additional requirement information of the corresponding interview request and is linked to the interview task list.

[0031] Optionally, the interview plan period of the target recruitment information is obtained, the idle period of a plurality of GPU servers in the interview plan period is queried, and the interview request allocation strategy step is solved by considering the interview appointment window time in the interview task list and the linked first additional requirement information and second additional requirement information, specifically including:

[0032] Obtain the interview plan period of the target recruitment information, query the idle period of a plurality of GPU servers in the interview plan period, and take the idle period of each GPU server and the GPU computing power parameter it has as the interview answer state monitoring resource information of the GPU server;

[0033] Based on the interview answer state monitoring resource information of each GPU server, considering the interview appointment window time in the interview task list and the linked first additional requirement information and second additional requirement information, taking the time window limit and the GPU computing power limit as the constraint condition, taking the highest GPU computing power proportion uniformity under the premise of the minimum GPU server resource occupation as the optimization target, and using an optimization algorithm to solve the interview appointment period corresponding to the GPU server to which each interview request end is allocated;

[0034] According to the interview appointment period corresponding to the GPU server to which each interview request end is allocated, the interview request allocation strategy of the target recruitment information is generated.

[0035] Optionally, based on the interview answer state monitoring resource information of each GPU server, considering the interview appointment window time in the interview task list and the first additional demand information and the second additional demand information linked, taking the time window limit and the GPU computing power limit as the constraint conditions, taking the highest GPU computing power proportion uniformity under the premise of the minimum GPU server resource occupation as the optimization target, an optimization algorithm is used to solve the corresponding interview appointment period of the GPU server to which each interview request end is allocated, specifically including:

[0036] Based on the GPU computing power parameter and the idle period in the interview answer state monitoring resource information of each GPU server, considering the interview appointment window time of each interview request, the GPU server candidate set, the GPU computing power demand parameter and the interview answer estimated total time length;

[0037] After each interview request end is allocated to the corresponding interview appointment period of the GPU server, the interview request of each interview request end is allocated to the GPU server in the first constraint condition of the corresponding GPU server candidate set of the interview request, the interview appointment period allocated to each interview request falls into the intersection of the corresponding interview appointment window time of the interview request and the idle period of the corresponding GPU server as the second constraint condition, the sum of the GPU computing power demand parameters of all interview requests allocated to each GPU server at each time is not higher than the GPU computing power parameter of the GPU server as the third constraint condition, taking the minimum GPU computing power demand proportion variance of each GPU server allocated to the interview request under the premise of the minimum number of GPU servers allocated to the interview request as the optimization target;

[0038] An heuristic optimization algorithm is used to solve the corresponding interview appointment period of the GPU server to which each interview request end is allocated.

[0039] Optionally, based on the corresponding interview appointment period of the GPU server to which each interview request end is allocated in the interview request allocation strategy, the GPU server is driven to perform the interview answer state monitoring step on the interview request end, specifically including:

[0040] Based on the corresponding interview appointment period of the GPU server to which each interview request end is allocated in the interview request allocation strategy, an interview answer video stream communication link between each interview request end and the corresponding GPU server in the interview appointment period is established;

[0041] Each GPU server is driven to perform interview answer state feature collection and real-time generation of interview evaluation information of the interview answer state monitoring item according to the received interview answer video stream and using the corresponding interview answer state monitoring model.

[0042] In addition, in order to achieve the above-mentioned purpose, the application also provides an interview answer state monitoring system, comprising:

[0043] An interview request receiving module is configured to receive an interview request sent by an interview request end for a target recruitment information, and construct an interview task list according to the interview appointment window time in the interview request and the interview post association information;

[0044] A first generating module is configured to generate interview GPU demand information as first additional demand information linked to the interview task list according to the interview answer state monitoring demand corresponding to the interview post association information;

[0045] A second generating module is configured to generate interview monitoring demand information as second additional demand information linked to the interview task list according to a plurality of interview question contents corresponding to the interview post association information;

[0046] A strategy solving module is configured to obtain an interview plan period of the target recruitment information, query the idle time periods of a plurality of GPU servers in the interview plan period, consider the interview appointment window time in the interview task list and the first additional demand information and the second additional demand information linked thereto, and solve an interview request distribution strategy;

[0047] A monitoring execution module is configured to drive the GPU servers to perform interview answer state monitoring on the interview request ends based on the interview appointment time period corresponding to the GPU server to which each interview request end is distributed in the interview request distribution strategy.

[0048] The application has the advantages that an interview answer state monitoring method and system are provided, the interview request sent by the interview request end for the target recruitment information is analyzed and arranged, the interview task list containing the interview appointment window time and the interview post association information is generated, the idle time periods of a plurality of GPU servers are obtained, the interview appointment window time in the interview task list and the interview GPU demand information and the interview monitoring demand information linked thereto are considered, the interview appointment time period corresponding to the GPU server to which each interview request end is distributed is solved, and each GPU server performs interview answer state monitoring according to the received interview answer video stream. Thus, the application matches the GPU resources with the computing power, the delay, and the like for each interview request, completely solves the contradiction between the delay caused by the over-provisioning of computing power and the waste of resources caused by the under-provisioning of computing power, reduces the number of GPU servers participating in distribution through the heuristic algorithm, improves the uniformity of GPU computing power, reduces the hardware operation cost of service institutions, provides efficient, accurate, and low-cost interview monitoring servers for recruitment enterprises, and improves the benefits of third-party service institutions as much as possible on the basis of ensuring user experience. BRIEF DESCRIPTION OF DRAWINGS

[0049] Figure 1A flowchart of an interview answer state monitoring method of an embodiment of the present application is shown in FIG. 1.

[0050] Figure 2 A structural diagram of an interview answer state monitoring system of an embodiment of the present application is shown in FIG. 2. DETAILED DESCRIPTION

[0051] In order to make the objectives, technical solutions and advantages of the present application clearer, the present application is further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and do not limit the present application.

[0052] An interview answer state monitoring method is provided in an embodiment of the present application, as shown in FIG. 1. Figure 1 , Figure 1 A flowchart of an interview answer state monitoring method of an embodiment of the present application is shown in FIG. 1.

[0053] In this embodiment, an interview answer state monitoring method includes the following steps:

[0054] S1: receiving an interview request sent by an interview request end for a target recruitment information, and constructing an interview task list according to an interview appointment window time in the interview request and interview post association information;

[0055] S2: generating interview GPU demand information as first additional demand information linked to the interview task list according to an interview answer state monitoring demand corresponding to the interview post association information;

[0056] S3: generating interview monitoring demand information as second additional demand information connected to the interview task list according to a plurality of interview question contents corresponding to the interview post association information;

[0057] S4: obtaining an interview plan period of the target recruitment information, querying idle periods of a plurality of GPU servers in the interview plan period, and considering the interview appointment window time in the interview task list and the first and second additional demand information linked, solving an interview request allocation strategy;

[0058] S5: based on the interview appointment period corresponding to the GPU server to which each interview request end is allocated in the interview request allocation strategy, driving the GPU server to perform interview answer state monitoring on the interview request end.

[0059] It should be noted that in the scenario of large-scale recruitment interview demand (such as regular recruitment interview plan of large state-owned enterprises and group enterprises), the recruitment enterprise needs to complete a large number of interview tasks in a short time, and the third-party service agency needs to consider the recruitment post demand, recruitment time period demand and recruitment quantity demand of each recruitment enterprise when providing interview answer state monitoring service, so that the existing technology has the problems of difficulty in dynamically matching GPU computing power demand, user experience degradation caused by over-provisioning of computing power, waste of resources caused by low-provisioning of computing power, and high allocation complexity under dynamic demand, which affects user experience (for example, the monitoring analysis result feedback of the interview answer state monitoring model has high delay, and cannot provide real-time analysis and evaluation for the interviewer), and also affects the benefit of the third-party service agency (low-provisioning of computing power reduces the number of recruitment enterprises served by the third-party service agency under the premise of limited GPU server resources).

[0060] To solve the above problems, the embodiment generates an interview task list containing interview appointment window time and interview post associated information by analyzing and sorting the interview request sent by the interview request end for the target recruitment information, and then obtains the idle time period of a plurality of GPU servers, considers the interview appointment window time in the interview task list and the linked interview GPU demand information and interview monitoring demand information, solves the interview appointment period corresponding to each GPU server allocated to each interview request end, and drives each GPU server to execute interview answer state monitoring according to the received interview answer video stream. Thus, the present application matches the GPU resources with appropriate computing power and delay for each interview request, solves the contradiction between delay caused by over-provisioning of computing power and resource waste caused by low-provisioning of computing power, ensures user experience, and improves the benefit of the third-party service agency as much as possible.

[0061] In the preferred embodiment, before the step of receiving the interview request sent by the interview request end for the target recruitment information and constructing the interview task list according to the interview appointment window time and the interview post associated information in the interview request, it further includes:

[0062] S101: obtaining the target recruitment information sent by the recruitment enterprise, and extracting the recruitment interview demand in the target recruitment information; wherein the recruitment interview demand includes recruitment interview post, recruitment quantity of each recruitment interview post and interview request delivery time period;

[0063] S102: determining a first generation judgment condition of the interview task list based on the recruitment interview post and the recruitment quantity of each recruitment interview post in the recruitment interview demand, and determining a second generation judgment condition of the interview task list based on the interview request delivery time period in the recruitment interview demand.

[0064] In this embodiment, the system first acquires the target recruitment information sent by the recruitment enterprise, extracts the core recruitment interview requirements therefrom, including the recruitment position name, the planned recruitment quantity of each position, the effective delivery period of the interview request, and then sets two judgment conditions based on the extracted requirements: the first generation judgment condition (associated with the position and the recruitment quantity, used for screening requests that do not exceed the quantity) and the second generation judgment condition (associated with the request delivery period, used for screening requests that do not exceed the time), which sets the standard for the subsequent precise construction of the task list. Therefore, by clearly defining the generation boundary and screening basis of the interview task list, the invalid requests that exceed the recruitment quantity and the delivery period are avoided from entering the subsequent process, the invalid consumption of resources is reduced, and the effectiveness and precision of the task list are improved.

[0065] On this basis, the interview request sent by the interview request end for the target recruitment information is received, and the interview task list is constructed according to the interview appointment window time in the interview request and the interview position association information, specifically including:

[0066] S111: receiving a plurality of interview requests sent by the interview request end for the target recruitment information, extracting the request submission time, the interview appointment window time and the interview position association information of the interview request;

[0067] S112: determining whether the interview request receiving quantity of the recruitment interview position matched with the interview position association information reaches the corresponding recruitment quantity according to the first generation judgment condition determined by the interview position association information of the interview request and the recruitment interview position and the recruitment quantity in the recruitment interview requirement, if yes, determining that the interview request meets the first generation judgment condition;

[0068] S113: determining whether the request submission time of the interview request is located in the interview request delivery period according to the second generation judgment condition determined by the request submission time of the interview request and the interview request delivery period in the recruitment interview requirement, if yes, determining that the interview request meets the second generation judgment condition;

[0069] S114: adding the interview appointment window time and the interview position association information in the interview request received at the same time meeting the first generation judgment condition and the second generation judgment condition to the constructed interview task list.

[0070] In this embodiment, after receiving the requests of multiple interview request terminals, the key information of each request is extracted, and then the requests are double-checked according to the first and second generated judgment conditions: checking whether the number of requests corresponding to the post associated information does not exceed the recruitment quantity (meeting the first condition), and checking whether the request submission time is within the delivery period (meeting the second condition). Finally, the requests that meet both conditions are added to the interview task list only with the appointment window time and the post associated information. In this way, the invalid requests are accurately filtered through double condition checking, ensuring that the interview task list only contains valid requests within the quantity limit and within the time validity period, avoiding the occupation of GPU resources and scheduling effort by invalid requests, and improving the subsequent resource allocation efficiency.

[0071] In the preferred embodiment, according to the interview answer state monitoring demand corresponding to the interview post associated information, the interview GPU demand information is generated and linked to the interview task list as the first additional demand information. Specifically, it includes:

[0072] S21: Determine the interview answer state monitoring demand based on the interview post type and the interview post ability requirement set in the interview post associated information; wherein the interview answer state monitoring demand includes a plurality of interview answer state monitoring projects;

[0073] S22: Determine the GPU algorithm requirement parameter and the GPU server candidate set that meets the interview experience requirement in terms of video analysis result feedback delay when performing the same interview answer state monitoring project, according to the video analysis test data of a plurality of GPU servers for interview answer video analysis;

[0074] Among them, the GPU server candidate set includes a plurality of GPU servers whose video analysis result feedback delay parameter in the video analysis test data is lower than the preset delay threshold of the interview experience requirement;

[0075] S23: Aggregate the GPU server candidate set and the GPU algorithm requirement parameter to generate interview GPU demand information, and link the interview GPU demand information as the first additional demand information corresponding to the interview request to the interview task list.

[0076] On this basis, based on the interview post type and the interview post ability requirement set in the interview post associated information, the interview answer state monitoring demand is determined, specifically including:

[0077] S211: Query the interview answer state monitoring model that meets the monitoring requirements corresponding to the interview post type and the interview post ability requirement set in the interview post associated information in the interview answer state monitoring model database based on the interview post type and the interview post ability requirement set in the interview post associated information;

[0078] S22: Query a plurality of interview answer state monitoring items covered in the interview answer state monitoring model as interview answer state monitoring requirements corresponding to the interview request.

[0079] In this embodiment, based on the job type and job capability requirement set in the interview job associated information, the interview answer state monitoring requirements required by the job are determined, and then based on the historical GPU server video analysis test data, two key parameters when executing these monitoring items are determined: GPU computing power requirement parameter (computing power consumed to complete monitoring) and GPU server candidate set satisfying interview experience (for example, GPU number with video analysis feedback delay ≤100 ms), finally the candidate set and the computing power parameter are summarized as interview GPU requirement information and linked to the list entry of the corresponding interview task as the first additional requirement information.

[0080] It should be noted that based on the job type and job capability requirement set in the interview job associated information, in the preset interview answer state monitoring model database, the monitoring model completely matched with the job type and the capability requirement is queried, and then all specific interview answer state monitoring items contained in the matched monitoring model are extracted, which are used as the final monitoring requirements of the interview request, realizing accurate matching of monitoring requirements and jobs.

[0081] For example, the post associated information of the "backend development position" shows that "Java code capability evaluation" is required, and the corresponding monitoring requirements are "code operation recognition" and "syntax correctness analysis". The historical test data shows that 24 TFLOPS computing power is required to execute these two monitoring items, and only the feedback delay of NVIDIA A100 model GPU (number G1-G15) meets the standard. The system generates GPU requirement information (computing power 24 TFLOPS, candidate set G1-G15) accordingly and links it to the interview task entry of the job.

[0082] Therefore, the embodiment accurately defines the GPU resource requirement for each interview task, avoids the problems of "insufficient computing power leading to delay" or "excessive computing power leading to waste" in subsequent allocation, and provides a basis for accurate scheduling of GPU resources.

[0083] In the preferred embodiment, the interview monitoring requirement information is generated according to the plurality of interview question contents corresponding to the interview job associated information, and is connected to the interview task list step as the second additional requirement information, which specifically includes:

[0084] S31: Extract the job type requirement keywords of the interview job type and the interview job capability requirement set in the interview job associated information, and match the plurality of interview question contents with the highest correlation degree with the job type requirement keywords in the corresponding interview question content set;

[0085] S32: According to the interview answer time length requirement recorded in each interview question content, the interview answer estimated total time length of each interview request is estimated, and the interview answer estimated total time length is converted into interview monitoring demand information as second additional demand information of the corresponding interview request and linked to the interview task list.

[0086] In this embodiment, the core keywords of the post type and the post ability requirement set (such as "communication ability" of the functional post) are extracted from the interview post associated information, then a number of interview questions with the highest association degree with these keywords (such as "scene communication question" and "team cooperation question" matched with the functional post) are matched in the preset interview question content set, finally the interview answer estimated total time length of the interview request is estimated according to the answer time length requirement recorded in each matched question, and the time length is converted into interview monitoring demand information as second additional demand information linked to the corresponding interview task list entry.

[0087] For example, the keywords of "human resource post" (functional post) are "communication and coordination, labor law knowledge", which are matched with "employee dispute handling question" (time length 20 minutes) and "labor contract interpretation question" (time length 15 minutes), and the system estimates the total interview time length to be 35 minutes, and "monitoring time length 35 minutes" is linked to the interview task entry of the post as second additional demand information.

[0088] Therefore, the embodiment provides a basis for subsequent query of GPU idle period and matching of idle window with suitable time length by specifying the monitoring time length required by each interview task, so as to avoid the task interruption problem caused by the fact that the GPU idle period is shorter than the interview time length.

[0089] In the preferred embodiment, the interview plan period of the target recruitment information is obtained, the idle periods of a number of GPU servers in the interview plan period are queried, the interview appointment window time and the first additional demand information and the second additional demand information linked in the interview task list are considered, and the interview request distribution strategy is solved.

[0090] S41: The interview plan period of the target recruitment information is obtained, the idle periods of a number of GPU servers in the interview plan period are queried, and the idle period and the GPU computing power parameter of each GPU server are taken as the interview answer state monitoring resource information of the GPU server;

[0091] S42: based on the interview answer state monitoring resource information of each GPU server, considering the interview appointment window time in the interview task list and the first additional demand information and the second additional demand information linked, taking the time window limit and the GPU computing power limit as the constraint condition, taking the highest GPU computing power proportion uniformity under the premise of the minimum GPU server resource occupation as the optimization target, and using an optimization algorithm to solve the interview appointment period corresponding to the GPU server allocated to each interview request end;

[0092] S43: generating an interview request allocation strategy of the target recruitment information according to the interview appointment period corresponding to the GPU server allocated to each interview request end.

[0093] On this basis, based on the interview answer state monitoring resource information of each GPU server, considering the interview appointment window time in the interview task list and the first additional demand information and the second additional demand information linked, taking the time window limit and the GPU computing power limit as the constraint condition, taking the highest GPU computing power proportion uniformity under the premise of the minimum GPU server resource occupation as the optimization target, and using an optimization algorithm to solve the interview appointment period corresponding to the GPU server allocated to each interview request end step, specifically comprising:

[0094] S421: based on the GPU computing power parameter possessed by each GPU server and the idle period in the interview answer state monitoring resource information, considering the interview appointment window time of each interview request, the GPU server candidate set, the GPU computing power demand parameter and the interview answer estimated total time length;

[0095] S422: taking the interview request of each interview request end being allocated to the interview appointment period corresponding to the GPU server as the first constraint condition, the interview request being allocated to the interview appointment period falling into the intersection of the interview appointment window time corresponding to the interview request and the idle period of the corresponding GPU server as the second constraint condition, and the sum of the GPU computing power demand parameters of all interview requests allocated to each GPU server at each time being not higher than the GPU computing power parameter possessed by the GPU server as the third constraint condition, taking the minimum GPU computing power demand proportion variance of each GPU server allocated to the interview request under the premise of the minimum number of GPU servers allocated to the interview request as the optimization target;

[0096] S423: using a heuristic optimization algorithm to solve the interview appointment period corresponding to the GPU server allocated to each interview request end.

[0097] In this embodiment, the overall interview plan period of the target recruitment information is first obtained, the idle period of all GPU servers in this period is queried, and the idle period and the GPU computing parameter are integrated into the interview monitoring resource information of the GPU server. Based on these resource information, combined with the reservation window time in the interview task list, the first additional demand information (GPU demand), and the second additional demand information (time length demand), with the two conditions as constraints: time window limit (the allocation period needs to be in the intersection of the reservation window and the GPU idle period), GPU computing power limit (the total allocation task computing power is less than or equal to the upper limit of the GPU), with the optimization goal of the minimum GPU server resource occupation (less server use) and the highest GPU computing power proportion uniformity (avoiding single GPU overload), the optimization algorithm is used to solve the GPU server and the corresponding interview appointment period allocated to each interview request, and finally the complete interview request allocation strategy is generated according to the solving result.

[0098] Specifically, considering the interview reservation window time in the interview task list and the linked interview GPU demand information and interview monitoring demand information, with the time window limit and the GPU computing power limit as the constraint conditions, with the optimization goal of the highest GPU computing power proportion uniformity under the premise of the minimum GPU server resource occupation, the optimization algorithm is used to solve the interview appointment period of the GPU server allocated to each interview request end, and each GPU server is driven to perform the interview answer state feature collection and interview evaluation information real-time generation of the interview answer state monitoring project according to the received interview answer video stream and the corresponding interview answer state monitoring model.

[0099] Therefore, under the premise of meeting the time and computing power constraints, the embodiment realizes the optimization scheduling of the GPU resources: both reduces the number of required GPU servers (reduces the cost) and ensures the uniform use of each GPU computing power (avoids overload or idling), and improves the resource utilization efficiency and scheduling rationality.

[0100] In the preferred embodiment, based on the interview appointment period of the GPU server allocated to each interview request end in the interview request allocation strategy, the GPU server is driven to perform the interview answer state monitoring step on the interview request end, specifically including:

[0101] S51: Based on the interview appointment period of the GPU server allocated to each interview request end in the interview request allocation strategy, the interview answer video stream communication link between each interview request end and the corresponding GPU server in the interview appointment period is established;

[0102] S52: Each GPU server is driven to perform the interview answer state feature collection and interview evaluation information real-time generation of the interview answer state monitoring project according to the received interview answer video stream and the corresponding interview answer state monitoring model.

[0103] In this embodiment, the automation, real-time and post-specific of interview monitoring and evaluation are realized, and the state collection and evaluation generation can be completed without manual intervention, which not only improves the interview monitoring efficiency, but also provides objective data evaluation basis for interviewers, ensures stable video stream transmission and avoids monitoring interruption.

[0104] Therefore, the application matches the GPU resources with adaptive computing power and delay compliance for each interview request, completely solves the contradiction between delay caused by over-provisioning of computing power and resource waste caused by low-provisioning of computing power, reduces the number of GPU servers participating in distribution through heuristic algorithm, improves the uniformity of GPU computing power, reduces the hardware operation cost of service agencies, and provides efficient, accurate and low-cost interview monitoring servers for recruitment enterprises, which improves the benefits of third-party service agencies as much as possible on the basis of ensuring user experience.

[0105] Reference Figure 2 , Figure 2 The structure diagram of the interview answer state monitoring system of the embodiment of the application is shown.

[0106] As shown in Figure 2 , the interview answer state monitoring system provided by the embodiment of the application comprises:

[0107] The interview request receiving module 10 is configured to receive the interview request sent by the interview request end for the target recruitment information, and construct an interview task list according to the interview appointment window time in the interview request and the interview post association information;

[0108] The first generation module 20 is configured to generate interview GPU demand information as first additional demand information linked to the interview task list according to the interview answer state monitoring demand corresponding to the interview post association information;

[0109] The second generation module 30 is configured to generate interview monitoring demand information as second additional demand information connected to the interview task list according to the interview question content corresponding to the interview post association information;

[0110] The strategy solving module 40 is configured to obtain the interview plan period of the target recruitment information, query the idle period of the plurality of GPU servers in the interview plan period, consider the interview appointment window time in the interview task list and the first additional demand information and the second additional demand information linked, and solve the interview request distribution strategy;

[0111] The monitoring execution module 50 is configured to drive the GPU server to perform the interview answer state monitoring on the interview request end based on the interview appointment period corresponding to the GPU server to which each interview request end is allocated in the interview request distribution strategy.

[0112] Other embodiments or specific implementations of the interview answer state monitoring system of the present application can refer to the above-mentioned method embodiments, and will not be repeated here.

[0113] It can be understood that, in the description of the present specification, the description of the terms "an embodiment", "another embodiment", "other embodiments", or "first embodiment to Nth embodiment" means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present application. In the present specification, the illustrative description of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any one or more embodiments or examples in a suitable manner.

[0114] It should be noted that, in this paper, the terms "include", "contain" or any other variants thereof are intended to cover non-exclusive inclusion, so that the process, method, article or system including a series of elements not only includes those elements, but also includes other elements not explicitly listed, or includes elements inherent to such process, method, article or system. Without more limitations, the element defined by the statement "including a" does not exclude the existence of other identical elements in the process, method, article or system including the element.

[0115] The above is only the preferred embodiment of the present application, and does not limit the patent scope of the present application, and any equivalent structure or equivalent process transformation using the content of the specification and drawings of the present application, or direct or indirect application in other related technical fields, are also included in the patent protection scope of the present application.

Claims

1. A method for monitoring interview answering status, characterized in that, The method includes the following steps: Receive interview requests sent by the interview request client for the target recruitment information, and construct an interview task list based on the interview appointment window time and the job posting information in the interview request; Based on the interview answer status monitoring requirements corresponding to the interview position association information, generate interview GPU requirement information, and link it to the interview task list as the first additional requirement information. Based on the content of several interview questions corresponding to the interview position association information, generate interview monitoring requirement information, and link it to the interview task list as a second additional requirement information; Obtain the interview schedule time slots for the target recruitment information, query the idle time slots of several GPU servers within the interview schedule time slots, and consider the interview appointment window time and the first and second additional requirement information of the link in the interview task list to solve the interview request allocation strategy. Based on the interview request allocation strategy, each interview request client is assigned a GPU server corresponding to the agreed interview time period, which drives the GPU server to monitor the interview answering status of the interview request client.

2. The interview answering status monitoring method as described in claim 1, characterized in that, Before the step of receiving an interview request from the interview requester targeting the job posting and constructing an interview task list based on the interview appointment window time and job posting information in the interview request, the process also includes: Obtain target recruitment information sent by recruiting companies, and extract the recruitment interview requirements from the target recruitment information; wherein, the recruitment interview requirements include the recruitment interview positions, the number of positions to be filled for each recruitment interview position, and the time period for submitting interview requests; Based on the job positions and the number of openings for each job position in the recruitment and interview requirements, a first condition for generating the interview task list is determined. Based on the interview request submission period in the recruitment and interview requirements, a second condition for generating the interview task list is determined.

3. The interview answering status monitoring method as described in claim 2, characterized in that, The steps for receiving interview requests for the target job posting from the interview request client, and constructing an interview task list based on the interview appointment window time and the job posting information in the interview request, specifically include: Receive interview requests for target job information from several interview request clients, and extract the request submission time, interview appointment window time and job-related information of the interview requests; Based on the first generation judgment condition determined by the interview position association information of the interview request and the recruitment interview position and recruitment quantity in the recruitment interview requirements, it is determined whether the number of interview requests received for recruitment interview positions that match the interview position association information reaches the corresponding recruitment quantity. If so, it is determined that the interview request meets the first generation judgment condition. Based on the second generation judgment condition determined by the request submission time of the interview request and the interview request submission period in the recruitment interview requirements, it is determined whether the request submission time of the interview request is within the interview request submission period. If so, it is determined that the interview request meets the second generation judgment condition. The interview appointment window time and the associated information of the interview position in the received interview requests that simultaneously meet the first generation judgment condition and the second generation judgment condition are added to the constructed interview task list.

4. The interview answering status monitoring method as described in claim 1, characterized in that, Based on the interview answer status monitoring requirements corresponding to the interview position association information, the process of generating interview GPU requirement information and linking it to the interview task list as the first additional requirement information specifically includes: Based on the job type and competency requirement set in the job-related information, the interview answer status monitoring requirements are determined; wherein, the interview answer status monitoring requirements include several interview answer status monitoring items; Based on video analysis test data from several GPU servers used for interview answer video analysis, a candidate set of GPU servers was determined that meet the requirements of interview experience in terms of GPU computing power optimization parameters and video analysis result feedback latency when performing the same interview answer status monitoring project. The GPU server candidate set includes several GPU servers whose video analysis result feedback latency parameters in the video analysis test data are lower than the preset latency threshold required for the interview experience. The candidate set of GPU servers and the GPU computing power quantitative requirement parameters are summarized to generate interview GPU requirement information. This interview GPU requirement information is then linked to the interview task list as the first additional requirement information for the corresponding interview request.

5. The interview answering status monitoring method as described in claim 4, characterized in that, Based on the job type and competency requirement set in the job-related information, the steps for monitoring interview answering status are determined, specifically including: Based on the interview job type and interview job competency requirement set in the interview job association information, query the interview answer status monitoring model database for an interview answer status monitoring model that meets the monitoring requirements corresponding to the interview job type and interview job competency requirement set. Query several interview answer status monitoring items covered in the interview answer status monitoring model, and use them as the interview answer status monitoring requirements for the corresponding interview request.

6. The interview answering status monitoring method as described in claim 4, characterized in that, Based on the interview questions corresponding to the job postings, generate interview monitoring requirements information, which is then linked to the interview task list as a second set of supplementary requirements information. This process specifically includes: Extract the job type and job competency requirement keywords from the job type association information, and match the most relevant interview questions to the corresponding interview question content set. Based on the interview answering time requirements recorded in the content of each interview question, the estimated total interview answering time for each interview request is estimated. The estimated total interview answering time is then converted into interview monitoring requirement information and linked to the interview task list as the second additional requirement information for the corresponding interview request.

7. The interview answering status monitoring method as described in claim 6, characterized in that, The steps for obtaining the interview schedule time slots for the target recruitment information, querying the idle time slots of several GPU servers within the interview schedule time slots, considering the interview appointment window time and the first and second additional demand information of the links in the interview task list, and solving the interview request allocation strategy include: Obtain the interview schedule time of the target recruitment information, query the idle time of several GPU servers within the interview schedule time, and use the idle time of each GPU server and the GPU computing power optimization parameters as the GPU server's interview answering status monitoring resource information; Based on the interview answer status monitoring resource information of each GPU server, considering the interview appointment window time and the first and second additional requirement information of the link in the interview task list, with time window limit and GPU computing power limit as constraints, and with the highest uniformity of GPU computing power ratio under the premise of minimizing GPU server resource consumption as the optimization objective, the optimization algorithm is used to solve the interview appointment time corresponding to the GPU server allocated to each interview request end. Based on the agreed interview time slot corresponding to the GPU server allocated to each interview request client, an interview request allocation strategy for the target recruitment information is generated.

8. The interview answering status monitoring method as described in claim 7, characterized in that, Based on the interview answering status monitoring resource information of each GPU server, considering the interview appointment window time and the first and second additional requirement information of the link in the interview task list, and with time window limits and GPU computing power limits as constraints, and with the optimization objective of maximizing the uniformity of GPU computing power ratio under the premise of minimizing GPU server resource consumption, an optimization algorithm is used to solve the steps corresponding to the interview appointment time slot of each interview request client allocated to the GPU server. Specifically, this includes: Based on the GPU computing power optimization parameters of each GPU server and the idle time period in the interview answer status monitoring resource information, the interview appointment window time, GPU server candidate set, GPU computing power optimization requirements parameters and the estimated total interview answer time are considered for each interview request. The first constraint is that after each interview request is assigned to the corresponding GPU server and the corresponding interview time slot, the interview request is assigned to the GPU server within the candidate set of the GPU server corresponding to the interview request. The second constraint is that the interview time slot assigned to each interview request falls within the intersection of the interview appointment window time corresponding to the interview request and the idle time slot of the corresponding GPU server. The third constraint is that the sum of the GPU computing power quantitative demand parameters of all interview requests assigned to each GPU server at each time is not higher than the GPU computing power quantitative parameters of the GPU server. The optimization objective is to minimize the variance of the GPU computing power demand ratio of each GPU server assigned to an interview request under the premise of minimizing the number of GPU servers assigned to interview requests. A heuristic optimization algorithm was used to determine the GPU server allocated to each interview request client for the corresponding agreed interview time slot.

9. The interview answering status monitoring method as described in claim 1, characterized in that, Based on the interview request allocation strategy, each interview request client is assigned a GPU server corresponding to a pre-agreed interview time slot. The GPU server is then driven to perform an interview answering status monitoring step for the interview request client, specifically including: Based on the agreed interview time slot corresponding to the GPU server allocated to each interview request client in the interview request allocation strategy, a video stream communication link for the interview answering between each interview request client and the corresponding GPU server is established within the agreed interview time slot. Each GPU server is driven to collect interview answer status features and generate interview evaluation information in real time based on the received interview answer video stream and the corresponding interview answer status monitoring model.

10. An interview answering status monitoring system, characterized in that, include: The interview request receiving module is used to receive interview requests sent by the interview request client for the target recruitment information, and to build an interview task list based on the interview appointment window time and the job position association information in the interview request; The first generation module is used to generate interview GPU requirement information based on the interview answer status monitoring requirements corresponding to the interview position association information, and link it to the interview task list as the first additional requirement information. The second generation module is used to generate interview monitoring requirement information based on the content of several interview questions corresponding to the interview position association information, and connect it to the interview task list as a second additional requirement information. The strategy solving module is used to obtain the interview plan time period of the target recruitment information, query the idle time periods of several GPU servers within the interview plan time period, and solve the interview request allocation strategy by considering the interview appointment window time and the first and second additional demand information of the link in the interview task list. The monitoring and execution module is used to drive the GPU server to monitor the interview answering status of the interview request client based on the agreed interview time period corresponding to the GPU server allocated to each interview request client in the interview request allocation strategy.

Images