Problem generation method, apparatus, and storage medium based on multi-agent collaboration

Abstract

This invention discloses a method, apparatus, and storage medium for question generation based on multi-agent collaboration. The method includes: acquiring job description text and segmenting it into multiple text paragraphs; inputting each text paragraph into multiple tag extraction agents to obtain multiple tag sets; inputting the multiple tag sets into an optimization agent to obtain a target tag set; and generating target question pairs based on the target tag sets, wherein each target question pair includes a question and an answer. This invention, by constructing a multi-agent collaborative mechanism, achieves effective segmentation, parallel processing, and result fusion of job description text, improving the accuracy of tag generation and thus the accuracy of question generation. It solves the problems of missing key information and incomplete tag coverage that often occur during the processing of job description text.

Description

Technical Field

[0001] This invention relates to the field of artificial intelligence technology, and in particular to a problem generation method, apparatus and storage medium based on multi-agent collaboration. Background Technology

[0002] With the rapid development of large language models, leveraging their ability to generate and expand labels has become an effective way to solve the bottleneck of resume analysis in the recruitment field. Compared with traditional manual annotation methods, although manual annotation has high accuracy, it is time-consuming and costly, making it difficult to meet the needs of building large-scale datasets. Existing automatic label generation methods mainly guide large language models to automatically generate high-quality labels based on contextual semantics through carefully designed prompt word engineering.

[0003] While existing large language model-based label generation methods have alleviated the problems of high cost and low efficiency of manual labeling to some extent, they still have significant shortcomings in practical applications. For example, large language models face the problem of distracted attention when processing long texts (job description texts), especially when faced with recruitment texts containing complex job descriptions and technical details. This can easily lead to omissions of key information and incomplete label coverage, making it difficult to meet the requirements of accuracy and professionalism in generating job interview questions, resulting in inaccurate label generation and inaccurate question generation.

[0004] Therefore, there is still an urgent need for a question generation method that can improve the accuracy of tag generation and question generation. Summary of the Invention

[0005] The main objective of this invention is to propose a problem generation method, apparatus, and storage medium based on multi-agent collaboration, which addresses the problem that existing large language models often produce inaccurate tag generation when processing job description text, thus leading to inaccurate problem generation.

[0006] To achieve the above objectives, this invention proposes a problem generation method based on multi-agent cooperation, which includes:

[0007] Obtain the job description text and segment it into multiple text paragraphs;

[0008] The multiple text segments are input into multiple tag extraction agents to obtain multiple tag sets;

[0009] Multiple label sets are input into the optimization agent to obtain the target label set;

[0010] Target question pairs are generated based on the target tag set, wherein the target question pairs include a question and an answer.

[0011] In some embodiments, the plurality of tag extraction agents include a first tag extraction agent and other tag extraction agents; the step of inputting the plurality of text paragraphs into the plurality of tag extraction agents to obtain a plurality of tag sets includes:

[0012] For each of the text segments, the text segment is input into the first label extraction agent to obtain the first predicted label set;

[0013] Input the other predicted label sets obtained from the other label extraction agents into the first label extraction agent;

[0014] The first label extraction agent obtains the label set by correcting the first predicted label set based on other predicted label sets.

[0015] The tag sets obtained by each tag extraction agent are combined to obtain multiple tag sets.

[0016] In some embodiments, inputting multiple label sets into the optimization agent to obtain a target label set includes:

[0017] Input multiple of the aforementioned label sets into the optimized agent;

[0018] The optimized agent performs deduplication processing on multiple tag sets to obtain a deduplicated tag set;

[0019] The target tag set is obtained by filtering the deduplicated tag set according to the preset tag library.

[0020] In some embodiments, prior to obtaining the job description text, the method further includes:

[0021] Crawl Chinese articles from a pre-defined database to obtain the raw article data;

[0022] The original article data is segmented into natural paragraphs to obtain multiple text units;

[0023] Keyword extraction is performed on each of the multiple text units to obtain the keywords corresponding to each text unit;

[0024] Associate the keywords corresponding to each text unit with the tags corresponding to the preset tag library.

[0025] In some embodiments, generating target question pairs based on the target tag set includes:

[0026] Obtain the associated keywords based on each tag in the target tag set;

[0027] Based on each of the keywords, obtain the text units corresponding to each of the keywords;

[0028] By concatenating the text units corresponding to each of the aforementioned keywords, a recombined segment is obtained;

[0029] The recombined segment and prompt template are input into a preset question generation model to obtain the target question pair.

[0030] In some embodiments, prior to obtaining the job description text, the method further includes:

[0031] Collect multiple original question pairs generated by a general large model based on the preset tag library;

[0032] Multiple original question pairs are filtered according to preset conditions to obtain several question pairs;

[0033] Save several of the aforementioned question pairs to the question database.

[0034] In some embodiments, generating target question pairs based on the target tag set includes:

[0035] Retrieve the question database based on the target tag set;

[0036] The question pairs in the question library that correspond to the target tag set are identified as the target question pairs.

[0037] In some embodiments, after generating target question pairs based on the target tag set, the method further includes:

[0038] The target problem pair is evaluated to obtain the evaluation results;

[0039] Optimize the target problem pair based on the evaluation results.

[0040] This invention also proposes a problem generation device based on multi-agent cooperation, comprising:

[0041] At least one processor; and,

[0042] A memory communicatively connected to the at least one processor; wherein,

[0043] The memory stores instructions that are executed by the at least one processor to enable the at least one processor to perform any of the above-described multi-agent cooperative problem generation methods.

[0044] The present invention also proposes a storage medium storing a computer program, the computer program including program instructions, which, when executed by a processor, enable the processor to execute the problem generation method based on multi-agent cooperation described above.

[0045] This invention processes job description text through the collaboration of multiple intelligent agents. First, the job description text is segmented into multiple text paragraphs. Then, multiple label extraction agents extract labels from each text paragraph to obtain multiple label sets. Next, an optimization agent optimizes these label sets to obtain a target label set. Finally, target question pairs are generated based on the target label set. By constructing a multi-agent collaborative mechanism, effective segmentation, parallel processing, and result fusion of job description text are achieved, improving the accuracy of label generation and thus the accuracy of question generation. This solves the problem of missing key information and incomplete label coverage that often occurs during the processing of job description text. Attached Figure Description

[0046] Figure 1 This is a flowchart illustrating the problem generation method based on multi-agent collaboration in an embodiment of the present invention.

[0047] Figure 2 This is another flowchart illustrating the problem generation method based on multi-agent collaboration in this invention.

[0048] Figure 3 This is another flowchart illustrating the problem generation method based on multi-agent collaboration in this invention.

[0049] Figure 4 This is another flowchart illustrating the problem generation method based on multi-agent collaboration in this invention.

[0050] Figure 5 This is another flowchart illustrating the problem generation method based on multi-agent collaboration in this invention.

[0051] Figure 6 This is another flowchart illustrating the problem generation method based on multi-agent collaboration in this invention.

[0052] Figure 7 This is another flowchart illustrating the problem generation method based on multi-agent collaboration in this invention.

[0053] Figure 8 This is another flowchart illustrating the problem generation method based on multi-agent collaboration in this invention.

[0054] Figure 9 This is a schematic diagram of the structure of the problem generation device based on multi-agent collaboration involved in the embodiments of the present invention.

[0055] The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0056] The solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0057] 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 positional relationship and movement of each component in a certain specific posture (as shown in the figure). If the specific posture changes, the directional indication will also change accordingly.

[0058] 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.

[0059] Furthermore, the use of terms such as "first" and "second" in this invention is for descriptive purposes only and should not be construed as indicating or implying their relative importance or implicitly specifying the number of technical features indicated. Therefore, a feature defined with "first" or "second" may explicitly or implicitly include at least one of that feature. Additionally, the technical solutions of the various embodiments can be combined with each other, but only on the basis of being achievable by those skilled in the art. When the combination of technical solutions is contradictory or impossible to implement, such a combination of technical solutions should be considered non-existent and not within the scope of protection claimed by this invention.

[0060] To achieve the above objectives, this invention proposes a problem generation method based on multi-agent cooperation, which includes:

[0061] Step S110: Obtain the job description text and segment the job description text into multiple text paragraphs;

[0062] Step S120: Input multiple text paragraphs into multiple label extraction agents to obtain multiple label sets;

[0063] Step S130: Input multiple label sets into the optimization agent to obtain the target label set;

[0064] Step S140: Generate target question pairs based on the target tag set, wherein the target question pairs include questions and answers.

[0065] In this embodiment, refer to Figure 1 and Figure 9The problem generation method based on multi-agent cooperation can be applied to a problem generation device based on multi-agent cooperation. The problem generation device based on multi-agent cooperation includes at least one processor and a memory communicatively connected to the at least one processor. The problem generation method based on multi-agent cooperation can be stored in the memory; the processor can invoke the problem generation method based on multi-agent cooperation stored in the memory to execute the problem generation method based on multi-agent cooperation. In this embodiment, the processor is the executing entity for all method steps.

[0066] Understandably, the processor is configured with multiple agents, including multiple label extraction agents and optimization agents. The number of label extraction agents needs to be greater than or equal to the number of text paragraphs so that each text paragraph can correspond to at least one label extraction agent.

[0067] When a user needs to generate job interview questions using a multi-agent collaborative question generation device, they can input job description text into the device. For example, the user can be a recruiter, and the job description text they input could describe the job duties, responsibilities, and required skills. The processor can then retrieve the job description text.

[0068] After the processor receives the job description text, it will segment the text into multiple paragraphs. For example, the job description text can be segmented by equal tokens, with each paragraph limited to no more than 512 tokens. Alternatively, the total number of tokens in the job description text can be divided by 512 tokens to determine the number of paragraphs.

[0069] After the processor segments the text into multiple paragraphs, it can input each paragraph into a separate label extraction agent; that is, one label extraction agent is assigned to each text paragraph. Each label extraction agent extracts labels based on its assigned text paragraph, resulting in its own extracted label set. Each label extraction agent then outputs its extracted label set. At this point, the processor obtains multiple label sets.

[0070] After the processor obtains multiple label sets, it can input them together into the optimizing agent. The optimizing agent can optimize these multiple label sets, for example, by removing duplicates. After optimizing the multiple label sets, the optimizing agent obtains the target label set, which it then outputs. At this point, the processor can obtain the target label set.

[0071] After obtaining the target label set, the processor can generate target question pairs based on the target label set, where each target question pair includes a question and an answer.

[0072] This embodiment utilizes a multi-agent collaborative approach to process job description text. First, the job description text is segmented into multiple paragraphs. Then, multiple tag extraction agents extract tags from each paragraph, resulting in multiple tag sets. Next, an optimization agent optimizes these tag sets to obtain a target tag set. Finally, target question pairs are generated based on the target tag set. By constructing a multi-agent collaborative mechanism, effective segmentation, parallel processing, and result fusion of job description text are achieved, improving the accuracy of tag generation and consequently, the accuracy of question generation. This addresses the issues of missing key information and incomplete tag coverage that often occur during the processing of job description text.

[0073] In some embodiments, the multiple tag extraction agents include a first tag extraction agent and other tag extraction agents; the aforementioned inputting multiple text segments into the multiple tag extraction agents to obtain multiple tag sets includes:

[0074] Step S150: For each text segment, input the text segment into the first label extraction agent to obtain the first predicted label set;

[0075] Step S151: Input the other predicted label sets obtained by the other label extraction agents into the first label extraction agent;

[0076] Step S152: Obtain the first label extraction agent to correct the first predicted label set based on other predicted label sets, and obtain the resulting label set;

[0077] Step S153: Collect the tag sets obtained by each tag extraction agent to obtain multiple tag sets.

[0078] In this embodiment, refer to Figure 2 When executing step S120, the processor may exchange the label sets of each label extraction agent. Multiple label extraction agents may include a first label extraction agent and other label extraction agents, wherein the first label extraction agent and the other label extraction agents are the same label extraction agent. The first label extraction agent and the other label extraction agents are relative to text segments. For each text segment, the label extraction agent input to that text segment is the first label extraction agent for that text segment; the label extraction agents input to other text segments are other label extraction agents for that text segment.

[0079] For example, consider three text paragraphs: paragraph a, paragraph b, and paragraph c; and three tag extraction agents: tag extraction agent a, tag extraction agent b, and tag extraction agent c. Paragraph a is input into tag extraction agent a, paragraph b into tag extraction agent b, and paragraph c into tag extraction agent c. For paragraph a, tag extraction agent a is the first tag extraction agent, and tag extraction agents b and c are the other tag extraction agents. Similarly, for paragraph b, tag extraction agent b is the first tag extraction agent, and tag extraction agents a and c are the other tag extraction agents.

[0080] For each text segment, the processor inputs it into a first label extraction agent. The first label extraction agent extracts labels based on the text segment, obtaining a first predicted label set. The first label extraction agent then saves the first predicted label set and outputs it. At this point, the processor obtains the first predicted label set. Of course, multiple text segments can be input into multiple label extraction agents simultaneously, allowing the processor to obtain other predicted label sets output by the other label extraction agents.

[0081] After obtaining the first predicted label set and other predicted label sets, the processor inputs the first predicted label set into other label extraction agents, and also inputs the other predicted label sets into the first label extraction agent. The first label extraction agent then supplements or corrects the first predicted label set based on the other predicted label sets, thus obtaining a complete label set, which is finally output. At this point, the processor obtains the label set obtained by the first label extraction agent based on the supplementation or correction of the first predicted label set by the other predicted label sets.

[0082] The processor can combine the tag sets obtained by the agents from each tag extraction to obtain multiple tag sets.

[0083] For example: If text segment 'a' is input into the 'a' label extraction agent, the 'a' label extraction agent will extract the predicted label set for 'a' from the text segment 'a'. Similarly, if text segment 'b' is input into the 'b' label extraction agent, the 'b' label extraction agent will extract the predicted label set for 'b' from the text segment 'b'. Likewise, if text segment 'c' is input into the 'c' label extraction agent, the 'c' label extraction agent will extract the predicted label set for 'c' from the text segment 'c'.

[0084] For text segment 'a', the tag extraction agent 'a' is the first tag extraction agent, while the tag extraction agents 'b' and 'c' are the other tag extraction agents. The predicted tag set 'a' is the first predicted tag set, and the predicted tag sets 'b' and 'c' are the other predicted tag sets. At this point, the predicted tag sets 'b' and 'c' are input into the tag extraction agent 'a', and the predicted tag set 'a' is also input into the tag extraction agents 'b' and 'c'. The tag extraction agent 'a' can then supplement or correct the predicted tag set 'a' based on the predicted tag sets 'b' and 'c', thus obtaining the final tag set.

[0085] Similarly, the label extraction agent b can supplement or correct the predicted label set b based on the predicted label set a and the predicted label set c, thus obtaining the corresponding label set. The label extraction agent c can supplement or correct the predicted label set c based on the predicted label set a and the predicted label set b, thus obtaining the corresponding label set. Finally, the processor combines the label sets obtained by each label extraction agent to obtain multiple label sets.

[0086] In this embodiment, each label extraction agent can refer to the output of other label extraction agents to further supplement or correct its own generated predicted label set, thereby improving the completeness and consistency of the label set.

[0087] In some embodiments, the aforementioned inputting multiple label sets into the optimization agent to obtain a target label set includes:

[0088] Step S160: Input multiple label sets into the optimization agent;

[0089] Step S161: Obtain the deduplicated tag set obtained by the optimized agent after deduplicating multiple tag sets;

[0090] Step S162: Filter the deduplicated tag set according to the preset tag library to obtain the target tag set.

[0091] In this embodiment, refer to Figure 3 When executing step S130, the processor can perform deduplication processing on multiple label sets. The processor first inputs the obtained multiple label sets into the optimizing agent. Then, the optimizing agent performs deduplication processing on the multiple label sets. For example, the optimizing agent will integrate the multiple label sets into a large label set, and then perform deduplication on the identical labels in this large label set to obtain a deduplicated label set. Finally, the optimizing agent outputs the deduplicated label set. At this point, the processor obtains the deduplicated label set.

[0092] After the processor obtains the deduplicated tag set, it can filter the deduplicated tag set according to a preset tag library to obtain the target tag set. The preset tag library can be user-defined or generated from historical tags output by the optimization agent. The processor can compare the deduplicated tag set according to the preset tag library, filter out multiple tags that meet the specifications, and thus obtain the target tag set.

[0093] In some embodiments, prior to obtaining the job description text, the process further includes:

[0094] Step S170: Crawl Chinese articles from a preset database to obtain raw article data;

[0095] Step S171: Perform natural paragraph segmentation on the original article data to obtain multiple text units;

[0096] Step S172: Extract keywords from multiple text units to obtain the keywords corresponding to each text unit;

[0097] Step S173: Associate the keywords corresponding to each text unit with the tags corresponding to the preset tag library.

[0098] In this embodiment, refer to Figure 4 Before executing step S110, the processor needs to crawl Chinese articles from a preset database to obtain raw article data. The processor can access the preset database using a preset interface to crawl Chinese articles from it. These Chinese articles can be job-related, such as articles detailing the technical aspects of the job. These job-related Chinese articles are then aggregated to obtain the raw article data. Finally, the raw article data is segmented into natural paragraphs to obtain multiple text units.

[0099] After the processor obtains multiple text units, it can extract keywords from each unit, identifying the keywords corresponding to each text unit. Then, it associates the keywords of each text unit with tags from a pre-defined tag library. For example, the keyword and tag can be associated only if they match, or if they have the same meaning.

[0100] For example, the processor first crawls Chinese articles from a preset database, cleans the articles, and then uses the cleaned articles as the original article data. Next, the original article data is segmented into multiple independent text units according to natural paragraphs. Then, keywords are extracted from each text unit, and the top three high-frequency keywords are extracted and associated with corresponding tags from a preset tag library using regular expressions.

[0101] In some embodiments, the aforementioned generation of target question pairs based on the target tag set includes:

[0102] Step S180: Obtain the associated keywords based on each tag in the target tag set;

[0103] Step S181: Obtain the text unit corresponding to each keyword based on each keyword;

[0104] Step S182: Concatenate the text units corresponding to each keyword to obtain the recombined segment;

[0105] Step S183: Input the recombined segment and prompt template into the preset question generation model to obtain the target question pair.

[0106] In this embodiment, refer to Figure 5 When executing step S140, the processor can generate target question pairs based on Chinese articles in a preset database. First, the processor obtains associated keywords based on each tag in the target tag set; then, it obtains the text units corresponding to each keyword. Next, it concatenates the text units corresponding to each keyword to obtain a reconstructed segment; finally, it inputs the reconstructed segment and prompt template into a preset question generation model, which then outputs the target question pairs based on the reconstructed segment and prompt template. At this point, the processor obtains the target question pairs. When a user needs multiple target question pairs to form a test paper, the above steps can be repeated to obtain more target question pairs.

[0107] The preset question generation model has a limit on the length of the character input (e.g., a character length limit of 4096 characters), which determines the preset length. For example, the processor can sequentially concatenate the text units corresponding to each keyword to obtain a reconstructed segment. If the total length of the reconstructed segment reaches or approaches the preset length, it will not continue concatenating; instead, it will concatenate the remaining text units corresponding to the keywords for the next reconstructed segment. The processor can assign the same prompt template to each reconstructed segment. The prompt template can be used to guide the preset question generation model to generate structured target question pairs. The prompt template can include explicit instructions (e.g., based on the content of the reconstructed segment, please generate a technically in-depth question and answer).

[0108] In some embodiments, prior to obtaining the job description text, the process further includes:

[0109] Step S190: Collect multiple original question pairs generated by the general large model based on the preset tag library;

[0110] Step S191: Filter multiple original question pairs according to preset conditions to obtain several question pairs;

[0111] Step S192: Save several question pairs to the question database.

[0112] In this embodiment, refer to Figure 6 Before executing step S110, the processor can also collect multiple original question pairs generated by the general large model according to the preset tag library. The processor first collects multiple original question pairs generated by the general large model according to the preset tag library, then filters the multiple original question pairs according to preset conditions, thereby selecting several question pairs that meet the preset conditions, and then saves several question pairs to the question library.

[0113] For example, the processor first collects a large number of raw question pairs generated by a general large model based on tags in a preset tag library. Then, it filters out the parts with higher scores, standardized formats, and accurate content from the large number of raw question pairs to obtain several question pairs. Finally, it saves these question pairs to a question library. Each question pair in the question library is associated with the tag that generated that question pair.

[0114] In some embodiments, the aforementioned generation of target question pairs based on the target tag set includes:

[0115] Step S200: Retrieve the question database based on the target tag set;

[0116] Step S201: Determine the question pairs in the question database that correspond to the target tag set as the target question pairs.

[0117] In this embodiment, refer to Figure 7 During step S140, the processor can also retrieve target question pairs from the question database. The processor can search the question database based on tags in the target tag set, and then determine the question pairs in the question database that correspond to the tags in the target tag set as target question pairs. Specifically, when the target tag set includes multiple tags, each tag can find a corresponding question pair; in this case, each question pair can be a target question pair.

[0118] In some embodiments, after generating target question pairs based on the target label set, the method further includes:

[0119] Step S210: Evaluate the target problem pair and obtain the evaluation results;

[0120] Step S211: Optimize the target problem pair based on the evaluation results.

[0121] In this embodiment, refer to Figure 8After executing step S140, the processor will further evaluate and optimize the target problem pair. The processor can also be configured with three types of agents: an evaluation agent, an optimization agent, and a control agent. The processor uses the control agent to invoke the evaluation agent to evaluate the target problem pair and obtain the evaluation result; then it invokes the optimization agent to optimize the target problem pair based on the evaluation result.

[0122] For example, the evaluation agent is primarily responsible for conducting a comprehensive quality analysis of the target problem pair from multiple dimensions. Its evaluation content covers difficulty level, practical value, correctness of answers, and reasonableness of questions. The difficulty assessment uses a three-level classification system, dividing the questions in the target problem pair into high, medium, and low levels, facilitating subsequent screening based on the needs of different job levels. The practicality assessment focuses on whether the questions accurately reflect the applicant's abilities and are closely related to the core responsibilities of the position. The correctness assessment ensures that the question semantics are clear and unambiguous, and that the reference answer accurately corresponds to the requirements of the question. The reasonableness assessment determines whether the question setting is relevant to actual work scenarios, avoiding fictitious content detached from the business context.

[0123] Based on the evaluation results of the assessment agent, the optimization agent restructures the target problem pair accordingly. The optimization agent not only corrects grammatical errors and logical flaws but also refines the expression of the target problem pair to better conform to industry terminology and expression conventions. Simultaneously, the optimization process preserves the key features of the original target problem pair to maintain consistency in intent and ensure the effectiveness of information transmission between agents.

[0124] The control agent, acting as the central coordinator between the evaluation and optimization agents, is responsible for managing the communication sequence and interaction rhythm between agents. It employs a polling scheduling strategy to control message flow and sets a maximum interaction limit of six rounds to control computational resource consumption while ensuring optimization effectiveness. In each round of interaction, the evaluation agent provides new evaluation results, and the optimization agent iteratively optimizes based on these results until a preset quality standard is reached or the maximum number of rounds is exceeded.

[0125] This invention is not limited to generating target question pairs corresponding to job description text; it can also be applied by adjusting the input description text. It is also applicable to other scenarios requiring the automatic generation of structured question pairs based on description text, such as educational assessments, technical training, and vocational competency evaluations. For example, in the education industry, job description text can be replaced with a course syllabus, which can then generate practice questions.

[0126] This invention processes job description text through the collaboration of multiple intelligent agents. First, the job description text is segmented into multiple text paragraphs. Then, multiple label extraction agents extract labels from each text paragraph to obtain multiple label sets. Next, an optimization agent optimizes these label sets to obtain a target label set. Finally, target question pairs are generated based on the target label set. By constructing a multi-agent collaborative mechanism, effective segmentation, parallel processing, and result fusion of job description text are achieved, improving the accuracy of label generation and thus the accuracy of question generation. This solves the problem of missing key information and incomplete label coverage that often occurs during the processing of job description text.

[0127] The problem generation device based on multi-agent cooperation in this invention can be a processor capable of running a problem generation method based on multi-agent cooperation; there is at least one processor. Figure 9 As shown, the problem generation device based on multi-agent collaboration may include: a processor 1001 (e.g., CPU), a network interface 1004, a user interface 1003, a memory 1005, and a communication bus 1002. The communication bus 1002 is used to establish communication between these components. The user interface 1003 may include a display screen 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 (such as a Wi-Fi interface). The memory 1005 may be high-speed RAM or stable non-volatile memory, such as a disk drive. Optionally, the memory 1005 may also be a storage device independent of the aforementioned processor 1001.

[0128] Those skilled in the art will understand that Figure 9 The structure of the problem generation device based on multi-agent collaboration shown in the figure does not constitute a limitation on the problem generation device based on multi-agent collaboration. It may include more or fewer components than shown, or combine certain components, or have different component arrangements.

[0129] like Figure 9 As shown, the memory 1005, which serves as a computer storage medium, may include an operating system, a network communication module, a user interface module, and computer programs.

[0130] exist Figure 9In the multi-agent collaborative problem generation device shown, the network interface 1004 is mainly used to connect to the backend server and communicate with the backend server; the user interface 1003 is mainly used to connect to the client (user end) and communicate 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, it implements the steps of the multi-agent collaborative problem generation method described above.

[0131] The present invention also proposes a computer device, the computer device including a memory, a processor and a computer program stored in the memory and running on the processor, wherein when the processor executes the computer program, it is capable of executing the problem generation method based on multi-agent cooperation as described above.

[0132] The present invention also proposes a storage medium storing a computer program, the computer program including program instructions, which, when executed by a processor, enable the processor to execute the problem generation method based on multi-agent cooperation described above.

[0133] The above description is only a part or preferred embodiment of the present invention. Neither the text nor the drawings should limit the scope of protection of the present invention. All equivalent structural transformations made using the content of the present invention specification and drawings under the overall concept of the present invention, or direct / indirect applications in other related technical fields, are included within the scope of protection of the present invention.

Claims

1. A problem generation method based on multi-agent cooperation, characterized in that, The problem generation method based on multi-agent cooperation includes: Obtain the job description text and segment it into multiple text paragraphs; The multiple text segments are input into multiple tag extraction agents to obtain multiple tag sets; Multiple label sets are input into the optimization agent to obtain the target label set; Target question pairs are generated based on the target tag set, wherein the target question pairs include a question and an answer; The multiple tag extraction agents include a first tag extraction agent and other tag extraction agents; the step of inputting multiple text paragraphs into the multiple tag extraction agents to obtain multiple tag sets includes: For each of the text segments, the text segment is input into the first label extraction agent to obtain the first predicted label set; Input the other predicted label sets obtained from the other label extraction agents into the first label extraction agent; The first label extraction agent obtains the label set by correcting the first predicted label set based on other predicted label sets. The tag sets obtained by each tag extraction agent are combined to obtain multiple tag sets; Among them, the first tag extraction agent and the other tag extraction agents are the same tag extraction agents; The step of generating target question pairs based on the target label set includes: Obtain the associated keywords based on each tag in the target tag set; Based on each of the keywords, obtain the text units corresponding to each of the keywords; By concatenating the text units corresponding to each of the aforementioned keywords, a recombined segment is obtained; The recombined segment and prompt template are input into a preset question generation model to obtain the target question pair.

2. The problem generation method based on multi-agent cooperation according to claim 1, characterized in that, The step of inputting multiple label sets into the optimization agent to obtain the target label set includes: Input multiple of the aforementioned label sets into the optimized agent; The optimized agent performs deduplication processing on multiple tag sets to obtain a deduplicated tag set; The target tag set is obtained by filtering the deduplicated tag set according to the preset tag library.

3. The problem generation method based on multi-agent cooperation according to claim 2, characterized in that, Before obtaining the job description text, the process also includes: Crawl Chinese articles from a pre-defined database to obtain the raw article data; The original article data is segmented into natural paragraphs to obtain multiple text units; Keyword extraction is performed on each of the multiple text units to obtain the keywords corresponding to each text unit; Associate the keywords corresponding to each text unit with the tags corresponding to the preset tag library.

4. The problem generation method based on multi-agent cooperation according to claim 2, characterized in that, Before obtaining the job description text, the process also includes: Collect multiple original question pairs generated by a general large model based on the preset tag library; Multiple original question pairs are filtered according to preset conditions to obtain several question pairs; Save several of the aforementioned question pairs to the question database.

5. The problem generation method based on multi-agent cooperation according to claim 4, characterized in that, The step of generating target question pairs based on the target label set further includes: Retrieve the question database based on the target tag set; The question pairs in the question library that correspond to the target tag set are identified as the target question pairs.

6. The problem generation method based on multi-agent cooperation according to any one of claims 1 or 5, characterized in that, After generating the target question pair based on the target label set, the method further includes: The target problem pair is evaluated to obtain the evaluation results; Optimize the target problem pair based on the evaluation results.

7. A problem generation device based on multi-agent cooperation, 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 that are executed by the at least one processor to enable the at least one processor to perform the problem generation method based on multi-agent cooperation as described in any one of claims 1 to 6.

8. A storage medium, characterized in that, The storage medium stores a computer program, which includes program instructions that, when executed by a processor, enable the processor to perform the problem generation method based on multi-agent cooperation as described in any one of claims 1 to 6.

Images