Recruitment test question generation method and device based on large model and storage medium

Abstract

The invention discloses a recruitment test question generation method and device based on a large model and a storage medium. The method comprises the steps of obtaining a post description text; inputting the post description text into a label labeling model to obtain a target technical label set; querying a preset question bank according to the target technical label set to obtain a candidate question set; and in response to the random extraction instruction, randomly extracting the candidate test question set to obtain a target test question set. According to the technical scheme of the invention, the unstructured post description text is converted into the structured target technical label set, and the standardized target technical label set is allocated to the post description text and the preset question bank through the decoupling design of matching the target technical label set with the preset question bank, so that the expression difference is flattened from the semantic level. The corresponding test questions can be matched only through the technical labels, the computing power resource consumption is reduced, and the question bank resource utilization rate is improved.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and in particular to a method, device and storage medium for generating recruitment test questions based on a large model. Background Art

[0002] With the rapid development of large language models, the use of large language models to automatically generate test questions based on job requirements and the expansion of question banks from unlabeled technical documents in recruitment scenarios have helped solve the problem of missing test questions for emerging technology positions.

[0003] However, in the existing technology, the large language model generates test content word by word based on input instructions, which completely relies on the real-time reasoning ability of the large language model. It is not integrated or collaboratively calculated with the pre-built question bank or other structured data sources, and the speed is slow, resulting in high consumption of computing power resources; or it is integrated or collaboratively calculated with the pre-built question bank or other structured data sources, but traditional retrieval only matches the literal meaning, resulting in a large number of questions that are not explicitly marked with technical points being idle, resulting in low utilization of question bank resources.

[0004] Therefore, there is still an urgent need for a recruitment test question generation method that can reduce computing resource consumption and improve question bank resource utilization. Summary of the Invention

[0005] The main purpose of the present invention is to propose a recruitment test question generation method, device and storage medium based on a large model to solve the problems of high computing power resource consumption and low question bank resource utilization when generating recruitment test questions with existing defective large language models.

[0006] To achieve the above objectives, the present invention proposes a method for generating recruitment test questions based on a large model, the method comprising: Get the job description text; Input the job description text into a label annotation model to obtain a target technical label set; Querying a preset question bank according to the target technical tag set to obtain a candidate question set; In response to the random extraction instruction, the candidate test question set is randomly extracted to obtain a target test question set.

[0007] In some embodiments, the label annotation model includes an industry classification model and a technical label extraction model; inputting the job description text into the label annotation model to obtain a target technical label set includes: Inputting the job description text into the industry classification model to obtain a target industry; Querying the industry-technology label mapping library according to the target industry to obtain a set of candidate technology labels; The job description text and the candidate technical tag set are input into the technical tag extraction model to obtain the target technical tag set.

[0008] In some embodiments, before querying a preset question bank based on the target technical tag set to obtain a candidate question set, the method further includes: Extracting technical tags from the industry-technical tag mapping library to obtain multiple technical tags; Querying a pre-configured knowledge base based on the technical tag to obtain a long text paragraph that explains the principle of the technical tag; Slicing the long text paragraph to obtain multiple unannotated texts; Inputting the prompt project and the unlabeled text into a test question generation model to obtain a test question set; The preset question bank is generated according to the test question set corresponding to each of the technical tags.

[0009] In some embodiments, the prompting project includes principle-question mapping prompts and answer completion prompts; and inputting the prompting project and the unlabeled text into the test question generation model to obtain a test question set includes: Inputting the principle-question mapping prompt and the unlabeled text into the test question generation model to obtain a question set; The answer completion prompt and the question set are input into the test question generation model again to obtain the test question set.

[0010] In some embodiments, inputting the answer completion prompt and the question set into the test question generation model again to obtain the test question set includes: For any question in the question set, inputting the answer completion hint and the question into the question generation model again to obtain multiple candidate answers; Calculating the perplexity of each candidate answer and performing rejection sampling on each candidate answer; Determine the candidate answer with the lowest perplexity as a target answer, and associate the question with the target answer; The test question set is generated according to each of the questions in the question set and the target answers.

[0011] In some embodiments, querying a preset question bank based on the target technical tag set to obtain a candidate question set includes: For any technical tag in the target technical tag set, query the preset question bank according to the technical tag to obtain a sub-candidate question set; The candidate question set is generated according to the sub-candidate question sets corresponding to the technical tags in the target technical tag set.

[0012] In some embodiments, after inputting the job description text and the candidate technical tag set into the technical tag extraction model to obtain the target technical tag set, the method further includes: Assigning a unique identifier to the job description text; A historical tag library is generated according to the unique identifier and the target technical tag set.

[0013] In some embodiments, the generation of recruitment test questions based on the large model further includes: Obtain the job description text again; Determining the unique identifier corresponding to the job description text; Determining whether the historical tag library stores the unique identifier; When the unique identifier is stored in the historical tag library, the corresponding target technical tag set is extracted from the historical tag library, and a preset question bank is searched according to the target technical tag set to obtain a candidate question set. When the unique identifier is not stored in the historical tag library, the step of inputting the job description text into a tag annotation model to obtain a target technical tag set is performed.

[0014] The present invention also proposes a recruitment test question generation device based on a large model, comprising: at least one processor; and, a memory communicatively connected to the at least one processor; wherein, The memory stores instructions to be executed by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute any one of the above-mentioned methods for generating recruitment test questions based on a large model.

[0015] The present invention also proposes a storage medium, which stores a computer program, and the computer program includes program instructions. When the program instructions are executed by a processor, the processor can execute any of the above-mentioned recruitment test question generation methods based on a large model.

[0016] The present invention obtains a job description text, inputs the job description text into a label annotation model, obtains a target technical label set, then queries a preset question bank according to the target technical label set, obtains a candidate question set, and finally responds to a random extraction instruction, randomly extracts the candidate question set, and obtains a target question set; by introducing a job description text labeling middle layer in the question generation process, by converting the unstructured job description text into a structured target technical label set, and through a decoupling design that matches the target technical label set with the preset question bank, according to the allocation of standardized target technical label sets for the job description text and the preset question bank, the expression differences are smoothed out at the semantic level, and only technical labels are needed to match the corresponding questions, thereby reducing computing power resource consumption and improving question bank resource utilization. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 Schematic diagram of the process of generating recruitment test questions based on a large model in an embodiment of the present invention; Figure 2 2 is another flow chart of a method for generating recruitment test questions based on a large model in an embodiment of the present invention; Figure 3 2 is another flow chart of a method for generating recruitment test questions based on a large model in an embodiment of the present invention; Figure 4 2 is another flow chart of a method for generating recruitment test questions based on a large model in an embodiment of the present invention; Figure 5 2 is another flow chart of a method for generating recruitment test questions based on a large model in an embodiment of the present invention; Figure 6 2 is another flow chart of a method for generating recruitment test questions based on a large model in an embodiment of the present invention; Figure 7 2 is another flow chart of a method for generating recruitment test questions based on a large model in an embodiment of the present invention; Figure 8 2 is another flow chart of a method for generating recruitment test questions based on a large model in an embodiment of the present invention; Figure 9 It is a structural diagram of a recruitment test question generation device based on a large model involved in an embodiment of the present invention.

[0018] The purpose, features and advantages of the present invention will be further described with reference to the accompanying drawings and in conjunction with the embodiments. DETAILED DESCRIPTION

[0019] The following will be combined with the accompanying drawings to clearly and completely describe the solutions in the embodiments of the present invention. Obviously, the embodiments described are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making any creative efforts are within the scope of protection of the present invention.

[0020] It should be noted that all directional indications (such as up, down, left, right, front, back, etc.) in the embodiments of the present invention are only used to explain the relative position relationship, movement status, etc. between the various components under a certain specific posture (as shown in the accompanying drawings). If the specific posture changes, the directional indication will also change accordingly.

[0021] It should also be noted that when an element is referred to as being "fixed on" or "disposed on" another element, it may be directly on the other element or there may be an intermediate element. When an element is referred to as being "connected to" another element, it may be directly connected to the other element or there may be an intermediate element.

[0022] In addition, the descriptions of "first", "second", etc. in the present invention are for descriptive purposes only and should not be understood as indicating or implying their relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined as "first" or "second" may explicitly or implicitly include at least one of such features. In addition, the technical solutions between the various embodiments can be combined with each other, but this must be based on the fact that they can be implemented by ordinary technicians in this field. When the combination of technical solutions is contradictory or cannot be implemented, it should be deemed that such combination of technical solutions does not exist and is not within the scope of protection required by the present invention.

[0023] To achieve the above objectives, the present invention proposes a method for generating recruitment test questions based on a large model. The method for generating recruitment test questions based on a large model includes: Step S110, obtaining a job description text; Step S120: Input the job description text into the label annotation model to obtain a target technical label set; Step S130, querying a preset question bank based on the target technical tag set to obtain a candidate question set; Step S140 , responding to the random extraction instruction, randomly extracting the candidate test question set to obtain the target test question set.

[0024] In this embodiment, referring to Figure 1 and Figure 9The large-model-based recruitment question generation method can be applied to a large-model-based recruitment question generation device. The large-model-based recruitment question generation device includes at least one processor and a memory communicatively connected to the at least one processor. The large-model-based recruitment question generation method can be stored in the memory; the processor can call the large-model-based recruitment question generation method from the memory to execute the large-model-based recruitment question generation method. In this embodiment, the processor is the main body for executing the method steps.

[0025] It is understood that the processor can be configured with a large model, which can be a large language model. A large language model refers to an ultra-large-scale natural language processing model trained using deep learning technology, capable of text understanding, generation, and reasoning. Trained on massive amounts of text data, it can perform tasks such as text generation, classification, and question-answering. Job description text refers to the descriptive text of job responsibilities and job requirements published during a company's recruitment process.

[0026] When a user needs to use the large-model-based recruitment question generator to generate recruitment questions, they can input a job description into the large-model-based recruitment question generator. For example, the user may be a recruiter, and the job description they input could include a description of the job responsibilities and job requirements (for example, "Backend Development Engineer, proficient in Java, MySQL, and Spring Boot framework, with at least three years of development experience"). At this point, the processor can retrieve the job description.

[0027] After the processor obtains the job description text, it can process the job description text. The processor inputs the job description text into the label annotation model to obtain the target technical label set. Among them, the large model can include a label annotation model, and the label annotation model can be pre-trained by the user. For example: the user can collect multiple job description texts, and then manually annotate the job description texts to train the label annotation model. When the processor obtains the job description text, it can input the job description text into the label annotation model. After the label annotation model obtains the job description text, it can process the job description text to output the target technical label set (for example: the label annotation model processes "backend development engineer, required to be proficient in Java, MySQL and Spring Boot framework, with more than 3 years of development experience" to obtain "Java, MySQL, Spring Boot"). At this point, the processor can obtain the target technical label set.

[0028] The processor may also be configured with a preset question bank. This preset question bank can be user-configured. For example, a user may pre-configure various questions and answers related to technical tags and save them to a database, thereby obtaining a preset question bank. The processor may then query the preset question bank based on the target set of technical tags to obtain a set of candidate test questions (for example, the processor may query the preset question bank based on "Java, MySQL, Spring Boot" to obtain test questions related to "Java, MySQL, SpringBoot," thereby obtaining a set of candidate test questions).

[0029] The candidate question set may contain many questions, and the user may only need to extract a portion of the candidate question set. In this case, the user may input a random extraction instruction. The processor may then receive the random extraction instruction and respond to the random extraction instruction by randomly extracting the candidate question set to obtain a target question set (for example, if the candidate question set contains 5 questions related to "Java," 3 questions related to "MySQL," and 2 questions related to "Spring Boot"; the random extraction instruction may be to randomly extract 3 questions from the 5 questions related to "Java," 1 question from the 3 questions related to "MySQL," and 1 question from the 2 questions related to "Spring Boot," thereby obtaining a target question set containing 5 questions). The random extraction is a random extraction without replacement.

[0030] After obtaining the target test set, the processor can also separate the questions and answers in the target test set to obtain a question set and an answer set, and then generate a test paper from the question set and a reference answer set from the answer set. At this point, the recruiter can print out the test paper and the reference answer set and directly provide the test paper to the applicant. After the applicant completes the test paper, the reference answer set can be used to mark the test paper.

[0031] Through this embodiment, the job description text is obtained, and then the job description text is input into the label annotation model to obtain the target technical label set, and then the preset question bank is queried according to the target technical label set to obtain the candidate question set, and finally the random extraction instruction is responded to, the candidate question set is randomly extracted to obtain the target question set; by introducing a job description text labeling middle layer in the question generation process, by converting the unstructured job description text into a structured target technical label set, and through the decoupling design of matching the target technical label set with the preset question bank, according to the assignment of standardized target technical label sets to the job description text and the preset question bank, the expression differences are smoothed out at the semantic level, and only technical labels are needed to match the corresponding questions, thereby reducing computing power resource consumption and improving question bank resource utilization.

[0032] In some embodiments, the label annotation model includes an industry classification model and a technical label extraction model; the aforementioned input of the job description text into the label annotation model to obtain the target technical label set includes: Step S150: Input the job description text into the industry classification model to obtain the target industry; Step S151: query the industry-technology label mapping library according to the target industry to obtain a set of candidate technology labels; Step S152: Input the job description text and the candidate technical tag set into a technical tag extraction model to obtain a target technical tag set.

[0033] In this embodiment, referring to Figure 2 When the processor executes step S120, it can process the job description text twice. The label annotation model can include an industry classification model and a technical label extraction model. The processor can first input the job description text into the industry classification model. The industry classification model can then analyze the job description text to determine the industry corresponding to the job description text, thereby obtaining the target industry, and then output the target industry. At this point, the processor can obtain the target industry. For example: the processor inputs the job description text (back-end development engineer, required to be proficient in Java, MySQL and Spring Boot framework, with more than 3 years of development experience) and instructions (to determine the major industry category) into the industry classification model, triggering the reasoning of the industry classification model, thereby obtaining the target industry (Internet-back-end development).

[0034] The processor can also be configured with a user-defined industry-technology tag mapping library. Once the processor determines the target industry, it can query the library based on the target industry to obtain candidate technology tag sets related to the target industry. For example, the processor can query the library based on the target industry (Internet - Backend Development) to obtain candidate technology tag sets related to the target industry (Java, MySQL, Spring Boot, Redis, Kafka, Docker, Jenkins, Git, Linux, Nginx, Tomcat, PostgreSQL).

[0035] After the processor obtains the candidate technical tag set, it can also input the job description text and the candidate technical tag set into the technical tag extraction model. The technical tag extraction model can then extract the corresponding technical tags from the candidate technical tag set based on the job description text, thereby obtaining the target technical tag set, and finally output the target technical tag set. At this point, the processor can obtain the target technical tag set. For example: the processor inputs the job description text (backend development engineer, required to be proficient in Java, MySQL, and Spring Boot framework, with more than three years of development experience), the candidate technical tag set (Java, MySQL, SpringBoot, Redis, Kafka, Docker, Jenkins, Git, Linux, Nginx, Tomcat, PostgreSQL), and the instruction (to determine which technical tags in the candidate technical tag set are relevant to the job description text) into the technical tag extraction model, triggering the technical tag extraction model's reasoning and obtaining the target technical tag set (Java, MySQL, Spring Boot).

[0036] In some embodiments, before querying the preset question bank based on the target technical tag set to obtain the candidate question set, the method further includes: Step S160: extracting technical labels from the industry-technology label mapping library to obtain multiple technical labels; Step S161: querying a pre-configured knowledge base based on the technical tag to obtain a long text paragraph that explains the principle of the technical tag; Step S162, slicing the long text paragraph to obtain multiple unannotated texts; Step S163, inputting the prompt project and the unlabeled text into the test question generation model to obtain a test question set; Step S164: Generate a preset question bank based on the question sets corresponding to each technical tag.

[0037] In this embodiment, referring to Figure 3 Before executing step S130, the processor also needs to generate a preset question bank. The industry-technology tag mapping library can include multiple technology tags, and the processor can generate corresponding test question sets based on these technology tags. The processor first extracts the technology tags from the industry-technology tag mapping library to obtain multiple technology tags.

[0038] Users can pre-configure a knowledge base. For example, they can collect a large amount of data related to a job and save it to the knowledge base. For example, they can search for job-related data on a public website and save the retrieved data to the knowledge base. After the processor obtains multiple technical tags, it can query them one by one. The processor can query the pre-configured knowledge base based on the technical tags and then retrieve a long text paragraph from the knowledge base that explains the principles behind the technical tags.

[0039] After the processor obtains the long text paragraph, it can perform slicing processing on the long text paragraph to slice the long text paragraph into multiple unannotated texts.

[0040] After the processor obtains multiple unlabeled texts, it can process the unlabeled texts one by one. The processor inputs the prompt project and the unlabeled text into the test question generation model to obtain a test question set. The prompt project can be a prompt word, which can be used to prompt the test question generation model to output related content. For example, the prompt project can generate three test questions based on the unlabeled text. After the processor inputs the unlabeled text and the three test questions generated based on the unlabeled text into the test question generation model, the test question generation model can output three test questions (test question set). At this point, the processor can obtain three test questions (test question set).

[0041] The processor can generate a preset question bank based on the test question sets corresponding to each technical tag, that is, collect the test question sets generated by each unlabeled text of each technical tag to generate the preset question bank.

[0042] In some embodiments, the prompting process includes principle-question mapping prompts and answer completion prompts. The aforementioned prompting process and unlabeled text are input into the test question generation model to obtain a test question set, including: Step S170 , inputting the principle-question mapping prompt and the unlabeled text into the test question generation model to obtain a question set; Step S171: input the answer completion prompt and the question set into the test question generation model again to obtain the test question set.

[0043] In this embodiment, referring to Figure 4When executing step S163, the processor may process the unannotated text twice. The prompt process may include principle-question mapping prompts and answer completion prompts. The processor may first input the principle-question mapping prompts and the unannotated text into the test question generation model to obtain a question set. For example, the unannotated text may be an explanation of the principles of Java, and the principle-question mapping prompt may be to generate three questions based on the Java principle explanation. After the processor inputs the Java principle explanation and the three questions generated based on the Java principle explanation into the test question generation model, the test question generation model can output three questions (question set). At this point, the processor can obtain three questions (question set).

[0044] After receiving the question set, the processor can also complete the answers. The processor can input the answer completion prompt and the question set back into the test question generation model to generate the test question set. For example, the answer completion prompt can be the answer to the completed question. After the processor inputs the three questions and the answers to the completed questions back into the test question generation model, the test question generation model can output the three questions and their corresponding answers (the test question set). At this point, the processor will obtain the three questions and their corresponding answers (the test question set).

[0045] In some embodiments, the aforementioned answer completion prompt and question set are inputted again into the test question generation model to obtain a test question set, including: Step S180: For any question in the question set, the answer completion hint and the question are input into the question generation model again to obtain multiple candidate answers; Step S181, calculate the perplexity of each candidate answer and perform rejection sampling on each candidate answer; Step S182: determine the candidate answer with the lowest perplexity as the target answer, and associate the question with the target answer; Step S183: Generate a test question set based on the questions and target answers in the question set.

[0046] In this embodiment, referring to Figure 5 When executing step S171, the processor may first generate multiple candidate answers and then select the candidate answer with the lowest perplexity. For any question in the question set, the processor may re-enter the answer completion prompt and the question into the question generation model to obtain multiple candidate answers. For example, if the question is a subjective question, the processor may re-enter the answer completion prompt and the question into the question generation model to obtain multiple candidate answers.

[0047] After the processor obtains multiple candidate answers, it performs rejection sampling on each candidate and calculates the perplexity of each candidate. For example, the processor can first set a perplexity threshold, such as 50. The perplexity of each candidate answer is then calculated, where lower perplexity indicates a more consistent answer. Rejection sampling is then performed, retaining only candidates with a perplexity below 50.

[0048] After the processor obtains the perplexity of each candidate answer, it can sort the perplexity of each candidate answer in descending order to determine the candidate answer with the lowest perplexity, and then determine the candidate answer with the lowest perplexity as the target answer; finally, the processor can also associate the question with the target answer.

[0049] After the processor obtains all the questions and target answers in the question set, it can generate a test question set based on the questions and target answers in the question set.

[0050] In some embodiments, the aforementioned step of querying a preset question bank based on the target technical tag set to obtain a candidate question set includes: Step S190: for any technical tag in the target technical tag set, query the preset question bank based on the technical tag to obtain a sub-candidate question set; Step S191 , generating a candidate question set according to the sub-candidate question sets corresponding to the technical tags in the target technical tag set.

[0051] In this embodiment, referring to Figure 6 When executing step S130, the processor may query the preset question bank for each technical tag in the target technical tag set. For any technical tag in the target technical tag set, the processor may query the preset question bank based on the technical tag to obtain a sub-candidate question set. For example, the processor may query the preset question bank for each technical tag in the target technical tag set to obtain a sub-candidate question set corresponding to each technical tag.

[0052] After the processor obtains the sub-candidate question sets corresponding to each technical label in the target technical label set, it can generate a candidate question set based on the sub-candidate question sets corresponding to each technical label in the target technical label set. For example, the processor can generate a candidate question set by combining the sub-candidate question sets corresponding to each technical label.

[0053] In some embodiments, after inputting the job description text and the candidate technical tag set into the technical tag extraction model to obtain the target technical tag set, the following steps may be further performed: Step S200, assigning a unique identifier to the job description text; Step S201 : Generate a historical tag library based on the unique identifier and the target technical tag set.

[0054] In this embodiment, referring to Figure 7 After executing step S152, the processor may also generate a historical tag library. The processor may assign a unique identifier to the job description text. Identical job description texts may be assigned the same unique identifier. The processor may assign a unique identifier to the job description text when obtaining the job description text.

[0055] After the processor assigns a unique identifier to the job description text, it can generate a historical tag library based on the unique identifier and the target technical tag set. That is, each time the processor generates a new target technical tag set, it can save it in the historical tag library, thereby updating the historical tag library.

[0056] In some embodiments, the aforementioned recruitment test question generation based on the large model further includes: Step S210, obtaining the job description text again; Step S211, determining a unique identifier corresponding to the job description text; Step S212, determining whether the historical tag library stores a unique identifier; Step S213: when the historical tag library stores a unique identifier, extract the corresponding target technical tag set from the historical tag library, and perform a step of querying the preset question bank based on the target technical tag set to obtain a candidate question set; Step S214 , when the historical tag library does not store the unique identifier, execute the step of inputting the job description text into the tag annotation model to obtain the target technical tag set.

[0057] In this embodiment, referring to Figure 8 When the processor obtains the job description text again, it can determine whether the historical tag library is saved. The user can continuously use the recruitment test question generation device based on the large model to generate recruitment test questions. At this time, the processor can obtain the job description text again.

[0058] After the processor obtains the job description text, it can assign a unique identifier to the job description text. Among them, the same job description text can be assigned the same unique identifier. For example: if the job description text currently obtained is the same as the job description text obtained previously, the processor can assign the unique identifier assigned to the job description text previously obtained to the job description text currently obtained. If the job description text currently obtained is different from the job description text previously obtained, the processor will assign a new unique identifier to the job description text currently obtained. After the processor assigns unique identifiers to the job description texts, it can directly determine the unique identifier corresponding to the job description text.

[0059] Once the processor determines the unique identifier for the job description, it can determine whether the historical tag library contains the unique identifier. For example, if the same job description is submitted multiple times (e.g., in batch imports or repeated user submissions), the processor can directly access the historical tag library to quickly obtain the target technical tag set, significantly reducing response time.

[0060] When the historical tag library stores a unique identifier, the processor can extract the corresponding target technical tag set from the historical tag library and execute the steps of querying the preset question bank based on the target technical tag set to obtain the candidate question set. For example, if the historical tag library stores a unique identifier, the processor can directly extract the target technical tag set, quickly proceeding to the next step and improving execution efficiency.

[0061] If the historical tag library does not contain a unique identifier, the processor must execute the steps of inputting the job description text into the tagging model to obtain the target technical tag set. For example, if the historical tag library contains a unique identifier, the processor must execute the entire tagging process. After executing the entire tagging process, the processor can also update the historical tag library with the target technical tag set and unique identifier. This allows the processor to quickly obtain the target technical tag set the next time the job description text is retrieved.

[0062] The present invention is not limited to generating target question pairs corresponding to job descriptions; by adjusting the input description text, the present invention is also applicable to other scenarios that require automatic generation of structured question pairs based on description text, such as educational assessment, technical training, and professional competency assessment. For example, in the education industry, the job description text can be replaced with a course syllabus, in which case exercises can be generated.

[0063] The present invention obtains a job description text, inputs the job description text into a label annotation model, obtains a target technical label set, then queries a preset question bank according to the target technical label set, obtains a candidate question set, and finally responds to a random extraction instruction, randomly extracts the candidate question set, and obtains a target question set; by introducing a job description text labeling middle layer in the question generation process, by converting the unstructured job description text into a structured target technical label set, and through a decoupling design that matches the target technical label set with the preset question bank, according to the allocation of standardized target technical label sets for the job description text and the preset question bank, the expression differences are smoothed out at the semantic level, and only technical labels are needed to match the corresponding questions, thereby reducing computing power resource consumption and improving question bank resource utilization.

[0064] The apparatus for generating recruitment test questions based on a large model according to an embodiment of the present invention may be a processor capable of running a method for generating recruitment test questions based on a large model; the processor may include at least one. Figure 9As shown, the large-scale model-based recruitment test question generation device may include: a processor 1001 (e.g., a CPU), a network interface 1004, a user interface 1003, a memory 1005, and a communication bus 1002. The communication bus 1002 is used to enable communication between these components. The user interface 1003 may include a display and an input unit, such as a keyboard. Optionally, the user interface 1003 may also include a standard wired interface or a wireless interface. The network interface 1004 may optionally include a standard wired interface or a wireless interface (e.g., a Wi-Fi interface). The memory 1005 may be a high-speed RAM memory or a non-volatile memory, such as a disk storage device. The memory 1005 may also be a storage device independent of the aforementioned processor 1001.

[0065] Those skilled in the art will understand that Figure 9 The structure of the recruitment test question generation device based on a large model shown in the figure does not constitute a limitation of the recruitment test question generation device based on a large model, and may include more or fewer components than shown in the figure, or combine certain components, or arrange the components differently.

[0066] like Figure 9 As shown, the memory 1005 as a computer storage medium may include an operating system, a network communication module, a user interface module and a computer program.

[0067] exist Figure 9 In the large-model-based recruitment test question generation device shown, the network interface 1004 is mainly used to connect to the backend server and communicate data with the backend server; the user interface 1003 is mainly used to connect to the client (user end) and communicate data with the client; and the processor 1001 can be used to call the computer program stored in the memory 1005. When the computer program is called and executed by the processor 1001, the steps of the above-mentioned large-model-based recruitment test question generation method are implemented.

[0068] The present invention also proposes a computer device, which includes a memory, a processor, and a computer program stored in the memory and running on the processor. When the processor executes the computer program, it can execute any of the above-mentioned large-model-based recruitment test question generation methods.

[0069] The present invention also proposes a storage medium, which stores a computer program, and the computer program includes program instructions. When the program instructions are executed by a processor, the processor can execute any of the above-mentioned recruitment test question generation methods based on a large model.

[0070] The above description is only a partial or preferred embodiment of the present invention. Neither the text nor the drawings can limit the scope of protection of the present invention. All equivalent structural transformations made by using the contents of the present invention description and drawings under the overall concept of the present invention, or direct / indirect application in other related technical fields are included in the scope of protection of the present invention.

Claims

1. A recruitment test question generation method based on a large model, characterized in that: The method for generating recruitment test questions based on the large model includes: Get the job description text; Input the job description text into a label annotation model to obtain a target technical label set; Querying a preset question bank according to the target technical tag set to obtain a candidate question set; In response to the random extraction instruction, the candidate test question set is randomly extracted to obtain a target test question set.

2. The method for generating recruitment test questions based on a large model according to claim 1, characterized in that: The label annotation model includes an industry classification model and a technical label extraction model; the position description text is input into the label annotation model to obtain a target technical label set, including: Inputting the job description text into the industry classification model to obtain a target industry; Querying the industry-technology label mapping library according to the target industry to obtain a set of candidate technology labels; The job description text and the candidate technical tag set are input into the technical tag extraction model to obtain the target technical tag set.

3. The method for generating recruitment test questions based on a large model according to claim 2, characterized in that: Before querying the preset question bank according to the target technical tag set to obtain the candidate question set, the method further includes: Extracting technical tags from the industry-technical tag mapping library to obtain multiple technical tags; Querying a pre-configured knowledge base based on the technical tag to obtain a long text paragraph that explains the principle of the technical tag; Slicing the long text paragraph to obtain multiple unannotated texts; Inputting the prompt project and the unlabeled text into a test question generation model to obtain a test question set; The preset question bank is generated according to the test question set corresponding to each of the technical tags.

4. The method for generating recruitment test questions based on a large model according to claim 3, characterized in that: The prompting project includes principle-question mapping prompts and answer completion prompts; The prompt project and the unlabeled text are input into the test question generation model to obtain a test question set, including: Inputting the principle-question mapping prompt and the unlabeled text into the test question generation model to obtain a question set; The answer completion prompt and the question set are input into the test question generation model again to obtain the test question set.

5. The method for generating recruitment test questions based on a large model according to claim 4 is characterized in that: The step of inputting the answer completion prompt and the question set into the test question generation model again to obtain the test question set includes: For any question in the question set, inputting the answer completion hint and the question into the question generation model again to obtain multiple candidate answers; Calculating the perplexity of each candidate answer and performing rejection sampling on each candidate answer; Determine the candidate answer with the lowest perplexity as a target answer, and associate the question with the target answer; The test question set is generated according to each of the questions in the question set and the target answers.

6. The method for generating recruitment test questions based on a large model according to claim 5, characterized in that: The step of querying a preset question bank according to the target technical tag set to obtain a candidate question set includes: For any technical tag in the target technical tag set, query the preset question bank according to the technical tag to obtain a sub-candidate question set; The candidate question set is generated according to the sub-candidate question sets corresponding to the technical tags in the target technical tag set.

7. The method for generating recruitment test questions based on a large model according to claim 2, characterized in that: After inputting the job description text and the candidate technical tag set into the technical tag extraction model to obtain the target technical tag set, the method further includes: Assigning a unique identifier to the job description text; A historical tag library is generated according to the unique identifier and the target technical tag set.

8. The method for generating recruitment test questions based on a large model according to claim 7, characterized in that: The generation of recruitment test questions based on the large model also includes: Obtain the job description text again; Determining the unique identifier corresponding to the job description text; Determining whether the historical tag library stores the unique identifier; When the unique identifier is stored in the historical tag library, the corresponding target technical tag set is extracted from the historical tag library, and a preset question bank is searched according to the target technical tag set to obtain a candidate question set. When the unique identifier is not stored in the historical tag library, the step of inputting the job description text into a tag annotation model to obtain a target technical tag set is performed.

9. A recruitment test question generation device based on a large model, characterized in that: include: at least one processor; as well as, a memory communicatively connected to the at least one processor; wherein, The memory stores instructions executed by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the recruitment test question generation method based on a large model as described in any one of claims 1 to 8.

10. A storage medium, characterized in that: The storage medium stores a computer program, which includes program instructions. When the program instructions are executed by a processor, the processor is enabled to execute the recruitment test question generation method based on a large model according to any one of claims 1 to 8.

Images