Large model-based resume matching method and device and storage medium
By using a large-model-based resume matching method, which extracts key information and calculates relevance scores using a matching model, the problem of inaccurate matching results in traditional methods is solved, achieving efficient and accurate resume matching.
Patent Information
- Application Number
- CN202511801683.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-12-02
- Publication Date
- 2026-02-27
AI Technical Summary
Existing resume matching methods rely on keyword-based matching technology, which is easily affected by synonyms, polysemous words, and contextual semantics, leading to inaccurate matching results.
A resume matching method based on a large model is adopted. Key information is extracted from job requirement text and resume text through a key information extraction model, and the relevance score is calculated using a resume matching model to determine the optimal resume text.
It improves the accuracy of resume matching by simplifying and integrating long text information, shortening the input length of the matching model, and achieving efficient and accurate matching results.
Smart Images
Figure CN121579709A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of intelligent recruitment technology, and in particular to a resume matching method, apparatus and storage medium based on a large model. Background Technology
[0002] With the widespread adoption of online recruitment platforms and the explosive growth of talent data, resume matching technology, as a core component of talent screening and job recommendation systems, is becoming increasingly important. Resume matching aims to automatically and efficiently calculate the similarity between job seekers' resumes and employers' job requirements, thereby assisting human resource decision-making and improving recruitment efficiency.
[0003] Currently, existing resume matching methods mainly rely on keyword-based matching techniques, such as using term frequency-inverse document frequency or Boolean models for text vectorization, and then comparing the results using metrics like cosine similarity. These methods are simple to implement, computationally inexpensive, and have some practicality in basic scenarios. However, resume texts are typically long texts containing multiple sections such as education, work experience, professional skills, and job expectations. They are information-rich and loosely structured, making traditional keyword matching methods susceptible to the influence of synonyms, polysemous words, and contextual semantics, leading to matching results that deviate from the needs of actual recruitment scenarios.
[0004] Therefore, existing resume matching methods are prone to inaccurate matching results. Summary of the Invention
[0005] The main objective of this application is to propose a resume matching method, apparatus, and storage medium based on a large model, aiming to solve the problem of inaccurate matching results.
[0006] To achieve the above objectives, this application proposes a resume matching method based on a large model, which includes: Retrieve the job requirements text and multiple resume texts to be matched; The job requirement text and multiple resume texts to be matched are input into the key information extraction model to obtain key job information and key resume information. The key job information and multiple key resume information are input into the resume matching model to obtain the relevance scores of the multiple key resume information. Based on the relevance scores, N optimal resume texts are determined from the multiple resume texts to be matched, and the N optimal resume texts are output, where N is a positive integer.
[0007] In some embodiments, before obtaining the job requirement text and multiple resume texts to be matched, the method further includes: Obtain the extraction model configuration parameters input by the user, and configure the initial key information extraction model according to the extraction model configuration parameters; Obtain multiple annotated job requirement texts, wherein the annotated job requirement texts include historical job requirement texts and job annotation information that annotates key information of historical jobs in the historical job requirement texts; Obtain multiple annotated resume texts, wherein the annotated resume texts include historical resume texts and resume annotation information that annotates key information of historical resumes in the historical resume texts; The initial key information extraction model is trained based on multiple labeled job requirement texts and multiple labeled resume texts to obtain the key information extraction model.
[0008] In some embodiments, after training the initial key information extraction model based on multiple labeled job requirement texts and multiple labeled resume texts to obtain the key information extraction model, the method further includes: Input multiple historical job requirement texts and multiple historical resume texts into the key information extraction model to obtain multiple historical job key information and multiple historical resume key information. Obtain the pairing information input by the user, and construct multiple pairing relationships between multiple key information of historical positions and multiple key information of historical resumes based on the pairing information to obtain multiple data pairs, wherein the data pair refers to the data pair of key information of historical positions and key information of historical resumes; The resume matching model is obtained by training a preset mGTE text representation model based on multiple datasets.
[0009] In some embodiments, the step of inputting the job requirement text into a key information extraction model to obtain key job information includes: Input the job requirement text into the key information extraction model; The key information extraction model outputs the educational background requirements, work experience requirements, professional skills requirements, and working conditions provided. The educational background requirements, work experience requirements, professional skills requirements, and working conditions provided are considered as key information for the position.
[0010] In some embodiments, the step of inputting multiple resume texts to be matched into a key information extraction model to obtain multiple resume key information includes: For any of the resume texts to be matched, input the resume texts to be matched into the key information extraction model; Obtain the educational background, work experience, professional skills, and expected working conditions output by the key information extraction model; The educational background, work experience, professional skills, and expected working conditions are used as key information in the resume.
[0011] In some embodiments, inputting the job-related key information and multiple resume-related key information into a resume matching model to obtain relevance scores for multiple resume-related key information includes: For any of the aforementioned key information in the resume, the key information in the job posting and the key information in the resume are input into the resume matching model; Obtain the encoded representations of the key information of the resume and the key information of the job position output by the resume matching model; The relevance score of the key information in the resume is calculated based on the encoded representation.
[0012] In some embodiments, determining N optimal resume texts from a plurality of resume texts to be matched based on the relevance score includes: Based on the key information of each resume, the relevance scores are associated with the text of each resume to be matched; Sort the relevance scores in descending order to obtain a sorted table; The top N relevance scores are determined as N target relevance scores; For any of the target relevance scores, the resume text to be matched associated with the target relevance score is determined as the optimal resume text.
[0013] In some embodiments, outputting N optimal resume texts includes: Based on the N optimal resume texts, the job requirement text, and the target relevance scores of the N optimal resume texts, generate N matching reasons for the optimal resume texts; Output N optimal resume texts and N matching reasons for the optimal resume texts.
[0014] This application further proposes a resume matching 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 that are executed by the at least one processor to enable the at least one processor to perform the large-model-based resume matching method described above.
[0015] This application further proposes a storage medium storing a computer program, the computer program including program instructions that, when executed by a processor, enable the processor to execute the above-described resume matching method based on a large model.
[0016] The technical solution of this application obtains key information about the job requirements and multiple resumes by inputting them into a key information extraction model. Then, it inputs these key information into a resume matching model to obtain relevance scores for the resumes. Finally, it determines several optimal resume texts based on these relevance scores. By extracting perceptual information from the job requirements and multiple resumes, long texts can be simplified and integrated, effectively shortening the input length of the resume matching model. This allows the resume matching model to perform efficient and accurate matching, thereby improving the accuracy of the matching results. Attached Figure Description
[0017] Figure 1 This is a flowchart illustrating an embodiment of the resume matching method based on a large model according to this application; Figure 2 This is a flowchart illustrating another embodiment of the resume matching method based on a large model in this application; Figure 3 This is a flowchart illustrating another embodiment of the resume matching method based on a large model in this application; Figure 4 This is a flowchart illustrating another embodiment of the resume matching method based on a large model in this application; Figure 5 This is a flowchart illustrating another embodiment of the resume matching method based on a large model in this application; Figure 6 This is a flowchart illustrating another embodiment of the resume matching method based on a large model in this application; Figure 7 This is a flowchart illustrating another embodiment of the resume matching method based on a large model in this application; Figure 8 This is a flowchart illustrating another embodiment of the resume matching method based on a large model in this application; Figure 9 This is a schematic diagram of the structure of the resume matching device based on a large model according to the embodiments of this application. Detailed Implementation
[0018] The solutions in the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments in this application, and not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of this application.
[0019] It should be noted that all directional indicators (such as up, down, left, right, front, back, etc.) in the embodiments of this application 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 indicator will also change accordingly.
[0020] It should also be noted that when a component is referred to as "fixed to" or "set on" another component, it can be directly on the other component or an intervening component can be present simultaneously. When a component is referred to as "connected to" another component, it can be directly connected to the other component or an intervening component can be present simultaneously.
[0021] Furthermore, the use of terms such as "first" and "second" in this application 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, features defined as "first" or "second" explicitly or implicitly include at least one of those features. 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. If 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 in this application.
[0022] This application proposes a resume matching method based on a large model, referring to... Figure 1 and Figure 9 , Figure 1 This is a flowchart illustrating an embodiment of the resume matching method based on a large model according to this application. Figure 9 This is a schematic diagram of the structure of the large-model-based resume matching device according to the embodiments of this application. In some embodiments, the large-model-based resume matching method includes: Step S110: Obtain the job requirements text and multiple resume texts to be matched; Step S120: Input the job requirement text and multiple resume texts to be matched into the key information extraction model to obtain the job key information and multiple resume key information. Step S130: Input the key job information and multiple key resume information into the resume matching model to obtain the relevance scores of multiple key resume information; Step S140: Determine N optimal resume texts from multiple resume texts to be matched based on relevance scores, and output the N optimal resume texts, where N is a positive integer.
[0023] In this embodiment, as Figure 1 and Figure 9 As shown, the resume matching method based on a large model can be applied to a resume matching device based on a large model. The resume matching device based on a large model includes at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions that are executed by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to execute the resume matching method based on the large model. In this embodiment, the processor is the main body executing the method steps.
[0024] Understandably, processors can be configured with large models, where large models can be large language models. Large language models refer to ultra-large-scale natural language processing models trained based on deep learning techniques, possessing text understanding, generation, and reasoning capabilities; trained on massive amounts of text data, they can perform tasks such as text extraction and matching.
[0025] During the recruitment process, users can first generate job requirement text based on recruitment needs, and then publish the job requirement text through various means (such as recruitment websites) to obtain resumes from multiple job seekers. After obtaining these resumes, users can treat them as multiple resumes to be matched. At this point, users can input the job requirement text and the multiple resumes to be matched into a large-scale model-based resume matching device. This device will then filter the resumes to obtain N optimal resumes, where N is a positive integer; that is, at least one optimal resume. Once the user inputs the job requirement text and the multiple resumes to be matched into the large-scale model-based resume matching device, the processor within the device can retrieve these information.
[0026] The large language model configured in the processor can include a key information extraction model. This key information extraction model can be pre-trained by the user and then configured on the processor. After the processor receives the job requirement text and multiple resume texts to be matched, it can input these texts into the key information extraction model to obtain the job's key information and the resume's key information. For example, the processor can input the job requirement text into the key information extraction model. After receiving the job requirement text, the key information extraction model can extract the key information from the text, thus obtaining the job's key information, and finally output it. At this point, the processor obtains the job's key information. Similarly, the processor can input multiple resume texts to be matched one by one into the key information extraction model. The key information extraction model can then extract the key information from each resume text individually, thus obtaining multiple resume key information, and finally outputting them. At this point, the processor obtains multiple resume key information. Of course, the processor can also input the job requirement text and multiple resume texts to be matched into the key information extraction model at the same time. The key information extraction model can extract key information from the job requirement text and multiple resume texts to be matched separately, thereby obtaining the key information of the job and the key information of the multiple resumes.
[0027] The large language model configured in the processor can also include a resume matching model. The key information extraction model can be pre-trained by the user and then configured on the processor. After obtaining the job key information and multiple resume key information, the processor can input these into the resume matching model to obtain relevance scores for each resume key information. For example, after the processor inputs the job key information and multiple resume key information into the resume matching model, the model can match each resume key information one by one according to the job key information, thereby obtaining relevance scores for each resume key information, and finally outputting the relevance scores for each resume key information. At this point, the processor can obtain the relevance scores for multiple resume key information.
[0028] After obtaining relevance scores for multiple resume key information, the processor can determine N optimal resume texts from multiple resume texts to be matched based on these relevance scores, and finally output N optimal resume texts. Here, N is a positive integer. For example, after obtaining the relevance scores for multiple resume key information, the processor can sort the relevance scores from high to low, and then select the N top-ranked relevance scores (e.g., first, second, third, etc.). The resume texts corresponding to these N top-ranked relevance scores are then selected as the optimal resume texts, thus obtaining N optimal resume texts. After obtaining N optimal resume texts, the processor can output them. At this point, the user can obtain the N optimal resume texts output by the resume matching device based on a large model.
[0029] The technical solution of this application obtains key information about the job requirements and multiple resumes by inputting them into a key information extraction model. Then, it inputs these key information into a resume matching model to obtain relevance scores for the resumes. Finally, it determines several optimal resume texts based on these relevance scores. By extracting perceptual information from the job requirements and multiple resumes, long texts can be simplified and integrated, effectively shortening the input length of the resume matching model. This allows the resume matching model to perform efficient and accurate matching, thereby improving the accuracy of the matching results.
[0030] Reference Figure 2 , Figure 2 This is a flowchart illustrating another embodiment of the resume matching method based on a large model according to this application. In some embodiments, before obtaining the job requirement text and multiple resume texts to be matched, the following steps are also included: Step S150: Obtain the extraction model configuration parameters input by the user, and configure the initial key information extraction model according to the extraction model configuration parameters. Step S151: Obtain multiple annotated job requirement texts, including historical job requirement texts and job annotation information that annotates key information of historical jobs in the historical job requirement texts. Step S152: Obtain multiple annotated resume texts, including historical resume texts and resume annotation information that annotates key information of historical resumes in the historical resume texts; Step S153: Train the initial key information extraction model based on multiple labeled job requirement texts and multiple labeled resume texts to obtain the key information extraction model.
[0031] In this embodiment, as Figure 2As shown, before executing step S110, the processor can first train a key information extraction model. When a user wants to train and configure the key information extraction model for the processor, the user can first collect a large amount of historical job requirement text and a large amount of historical resume text. Then, the user can annotate the key information of historical job requirements in each historical job requirement text and use these annotations as job annotation information; similarly, the user can also annotate the key information of historical resumes in each historical resume text and use these annotations as resume annotation information. The user can integrate the job requirement texts with the job annotation information of each historical job requirement text to obtain multiple annotated job requirement texts; similarly, the user can also integrate the resume texts with the resume annotation information of each historical resume text to obtain multiple annotated resume texts.
[0032] Users can also set extraction model configuration parameters according to their extraction needs, and then input these parameters into the resume matching device based on the large model. At this point, the processor in the resume matching device can obtain the user-input extraction model configuration parameters, and then configure the initial key information extraction model based on these parameters.
[0033] Once the processor has configured the initial key information extraction model, the user can input the previously collected and labeled job requirement texts into the resume matching device based on the large model. At this point, the processor in the large model-based resume matching device can obtain multiple labeled job requirement texts. These labeled job requirement texts include historical job requirement texts and job annotation information that annotates key information from historical job requirements. Key information from historical job requirements may include educational background requirements, work experience requirements, professional skills requirements, and provided working conditions.
[0034] Similarly, users can input multiple annotated resume texts previously collected and labeled into a resume matching device based on a large model. The processor in this device will then generate multiple annotated resume texts. These annotated resume texts include historical resume texts and annotation information that highlights key information from those historical resumes. Key information from historical resumes may include educational background, work experience, professional skills, and desired work conditions.
[0035] The processor can train an initial key information extraction model based on multiple labeled job requirement texts and multiple labeled resume texts, thereby obtaining the key information extraction model.
[0036] Reference Figure 3 , Figure 3This is a flowchart illustrating another embodiment of the resume matching method based on a large model in this application. In some embodiments, after training the initial key information extraction model based on multiple labeled job requirement texts and multiple labeled resume texts to obtain the key information extraction model, the method further includes: Step S160: Input multiple historical job requirement texts and multiple historical resume texts into the key information extraction model to obtain multiple historical job key information and multiple historical resume key information. Step S161: Obtain the pairing information input by the user, and construct a pairing relationship between multiple historical job key information and multiple historical resume key information based on the pairing information to obtain multiple data pairs. Here, a data pair refers to a historical job key information-historical resume key information data pair. Step S162: Train the preset mGTE text representation model based on multiple data to obtain the resume matching model.
[0037] In this embodiment, as Figure 3 As shown, after executing step S153, the processor can further extract key information from multiple historical job requirement texts and multiple historical resume texts based on the trained key information extraction model. After the user has trained the key information extraction model, they can instruct the processor to input multiple historical job requirement texts and multiple historical resume texts into the key information extraction model. Upon receiving the instruction, the processor will input multiple historical job requirement texts and multiple historical resume texts into the key information extraction model, thereby obtaining multiple historical job key information and multiple historical resume key information.
[0038] After obtaining key information from multiple historical job postings and resumes, the processor can output this information. Users can then comprehensively filter and review this data to ensure its quality and relevance. Incomplete or outdated data entries are removed, retaining only the most valuable and representative information. During this process, special attention is paid to data diversity and balance, ensuring the dataset covers different industries, job levels, and work experience backgrounds. To further improve data quality, each "text-key information" pair is meticulously checked to ensure information correspondence and clear, accurate expression. After the user has completed the verification and filtered out high-quality data, they can pair the key information from each historical job posting and resume (e.g., if the key information from a historical job posting is related to a front-end development engineer position, then the key information from the relevant resumes will be paired with that key information from the historical job posting). Once the pairings are complete, matched information can be generated based on these pairings. In addition, users can update and train the key information extraction model based on the verification results to enhance the key information extraction effect of the model.
[0039] Users can input pairing information into the processor, which then retrieves this information. After obtaining the pairing information, the processor can construct multiple pairing relationships between key information from historical job positions and key information from historical resumes, resulting in multiple data pairs. Each data pair refers to a pair of key information from a historical job position and key information from a historical resume. Specifically, one key information from a historical job position can form one data pair with one key information from a historical resume; conversely, one key information from a historical job position can also form multiple data pairs with different key information from different historical resumes.
[0040] After the processor obtains multiple data pairs, it can train the preset mGTE text representation model based on the multiple data pairs to obtain the resume matching model.
[0041] For example, users can first configure the original mGTE text representation model according to actual needs, and then train the original mGTE text representation model using rotational position encoding, gradually expanding the maximum context length with a two-stage progressive training strategy. Rotational position encoding supports length extrapolation, meaning that shorter sequences are used during training, while longer sequences can be processed during inference. During training, the mGTE text representation model is trained in the first stage with a rotation angle cardinality of 10000 and a length of 2048 input words; in the second stage, the rotation angle cardinality is set to 160000 and training continues on a length of 8192 input words, thus obtaining the preset mGTE text representation model. After obtaining the preset mGTE text representation model, it can be trained on multiple datasets to obtain a resume matching model. Throughout the training process, this progressive approach allows the model to gradually adapt to long text inputs, avoiding the instability and resource consumption caused by directly training long sequences.
[0042] Reference Figure 4 , Figure 4 This is a flowchart illustrating another embodiment of the resume matching method based on a large model in this application. In some embodiments, the aforementioned input of job requirement text into a key information extraction model to obtain key job information includes: Step S170: Input the job requirement text into the key information extraction model; Step S171: Obtain the key information extraction model output, including educational background requirements, work experience requirements, professional skills requirements, and working conditions provided. Step S172: The educational background requirements, work experience requirements, professional skills requirements, and working conditions provided are taken as key job information.
[0043] In this embodiment, as Figure 4 As shown, when the processor inputs the job requirement text into the key information extraction model in step S120 to obtain the key job information, it can obtain the educational background requirements, work experience requirements, professional skill requirements, and provided working conditions. The processor can input the job requirement text into the key information extraction model. After obtaining the job requirement text, the key information extraction model can extract the key information from the job requirement text, thereby obtaining the educational background requirements, work experience requirements, professional skill requirements, and provided working conditions, and finally output the educational background requirements, work experience requirements, professional skill requirements, and provided working conditions of the job requirement text.
[0044] After the key information extraction model outputs the extracted key information, the processor can obtain the educational background requirements, work experience requirements, professional skills requirements, and working conditions provided in the job requirement text. The processor can also use these four key information as job-related key information. This embodiment focuses on these four core dimensions—educational background requirements, work experience requirements, professional skills requirements, and working conditions—to improve subsequent matching accuracy.
[0045] Reference Figure 5 , Figure 5 This is a flowchart illustrating another embodiment of the resume matching method based on a large model in this application. In some embodiments, the aforementioned input of multiple resume texts to be matched into a key information extraction model yields multiple resume key information, including: Step S180: For any resume text to be matched, input the resume text to be matched into the key information extraction model. Step S181: Obtain key information from the model output, including educational background, work experience, professional skills, and expected working conditions. Step S182: Include educational background, work experience, professional skills, and expected working conditions as key information in the resume.
[0046] In this embodiment, as Figure 5 As shown, when the processor performs step S120, it inputs multiple resume texts to be matched into the key information extraction model and obtains multiple resume key information, including education experience, work experience, professional skills, and expected working conditions.
[0047] For any resume text to be matched, the processor can input the resume text into the key information extraction model. After obtaining the resume text, the key information extraction model can extract the key information from the resume text, thereby obtaining the educational background, work experience, professional skills, and expected working conditions of the resume text, and finally output the educational background, work experience, professional skills, and expected working conditions of the resume text.
[0048] After the key information extraction model outputs the extracted key information, the processor can obtain the educational background, work experience, professional skills, and expected work conditions from the resume text to be matched. The processor can also use the educational background, work experience, professional skills, and expected work conditions from the resume text to be matched as key information in the resume.
[0049] After extracting all the resume texts to be matched, the processor can obtain multiple key resume information. This embodiment focuses the key information on four core dimensions: education experience, work experience, professional skills, and expected working conditions, thereby improving the accuracy of subsequent matching.
[0050] Reference Figure 6 , Figure 6 This is a flowchart illustrating another embodiment of the resume matching method based on a large model according to this application. In some embodiments, the aforementioned input of key job information and multiple key resume information into the resume matching model to obtain relevance scores for multiple key resume information includes: Step S190: For any key information in a resume, input the key information of the job posting and the key information in the resume into the resume matching model; Step S191: Obtain the encoded representations of key resume information and key job information output by the resume matching model; Step S192: Calculate the relevance score of key information in the resume based on the coding representation.
[0051] In this embodiment, as Figure 6 As shown, when executing step S130, the processor can match key information from the resume one by one. For any key information in the resume, the processor can input the key information of the job position and the key information of the resume into the resume matching model. The key information of the job position includes educational background requirements, work experience requirements, professional skills requirements, and the working conditions provided; the key information of the resume includes educational background, work experience, professional skills, and expected working conditions. The resume matching model can encode the educational background requirements, work experience requirements, professional skills requirements, and the working conditions provided in the key information of the job position, as well as the key information of the resume including educational background, work experience, professional skills, and expected working conditions, to obtain vector representations of each part. Then, the corresponding vector representations are calculated using the cosine formula to obtain their corresponding cosine similarity values (e.g., the vector representations of educational background requirements and educational experience are calculated using the cosine formula to obtain the cosine similarity value between educational background requirements and educational experience; similarly, the cosine similarity values between work experience requirements and work experience, between professional skills requirements and professional skills, and between the working conditions provided and the expected working conditions are obtained), and the cosine similarity value is used as the encoded representation.
[0052] The resume matching model outputs encoded representations of key resume information and key job information. The processor can then obtain these encoded representations. The processor can also calculate relevance scores for the key resume information based on these encoded representations. For example, it can obtain a relevance score by weighted summing of the cosine similarity values across four dimensions. After matching and calculating all the key resume information, the processor can obtain multiple relevance scores for each piece of information.
[0053] The formula for calculating the relevance score includes: ; ; ; ; in, This indicates the first key information in the job posting. Vector representation of key information. Operators that represent vector representations This indicates the first key information in the job posting. One key piece of information In It includes educational background requirements, work experience requirements, professional skills requirements, and working conditions provided, where 0 represents educational background requirements, 1 represents work experience requirements, 2 represents professional skills requirements, and 3 represents working conditions provided.
[0054] in, This indicates the first key information in the resume. Vector representation of key information. Operators that represent vector representations This indicates the first key information in the resume. One key piece of information In It includes educational background, work experience, professional skills, and expected working conditions, where 0 represents educational background, 1 represents work experience, 2 represents professional skills, and 3 represents expected working conditions.
[0055] in, This indicates the first key information in the job posting. The vector representation of the key information and the first key information in the resume The cosine similarity value of the vector representation of key information. This represents the operator for calculating the cosine formula.
[0056] in, Relevance score, Indicates the first indivual The weights are equal to 1.
[0057] Reference Figure 7 , Figure 7 This is a flowchart illustrating another embodiment of the resume matching method based on a large model in this application. In some embodiments, the aforementioned determination of N optimal resume texts from multiple resume texts to be matched based on relevance scores includes: Step S200: Based on the key information of each resume, associate each relevance score with each resume text to be matched; Step S201: Sort the relevance scores in descending order to obtain a sorting table; Step S202: Determine the top N relevance scores as the N target relevance scores; Step S203: For any target relevance score, determine the resume text to be matched associated with the target relevance score as the optimal resume text.
[0058] In this embodiment, as Figure 7 As shown, when the processor determines N optimal resume texts from multiple resume texts to be matched based on relevance scores in step S140, it can first associate each relevance score with each resume text to be matched based on the key information of each resume. The processor can determine each resume text to be matched corresponding to each resume key information based on the key information of each resume, and then associate the relevance score of each resume key information with each resume text to be matched.
[0059] The processor can also sort the relevance scores in descending order to obtain a sorted table. Descending order sorts from highest to lowest; the processor sorts the relevance scores from highest to lowest to obtain the sorted table.
[0060] The processor can also determine the top N relevance scores as N target relevance scores based on a sorting table. Here, N is a positive integer. N can be user-defined, such as N being 1, 2, or 3.
[0061] For any given target relevance score, the processor can determine the optimal resume text as the one associated with that score. For example, when N is 1, the processor can obtain 1 optimal resume text; when N is 2, the processor can obtain 3 optimal resume texts; when N is 3, the processor can obtain 3 optimal resume texts, and so on.
[0062] Reference Figure 8 , Figure 8 This is a flowchart illustrating another embodiment of the resume matching method based on a large model in this application. In some embodiments, the aforementioned output of N optimal resume texts includes: Step S210: Based on the target relevance scores of the N optimal resume texts, the job requirement text, and the N optimal resume texts, generate the matching reasons for the N optimal resume texts. Step S211: Output the N optimal resume texts and the matching reasons for the N optimal resume texts.
[0063] In this embodiment, as Figure 8As shown, when the processor executes step S140 to output N optimal resume texts, it can first generate N matching reasons for each optimal resume text. The processor can generate N matching reasons for each optimal resume text based on the N optimal resume texts, the job requirement text, and the target relevance scores of the N optimal resume texts. Then, it outputs the N optimal resume texts and their matching reasons.
[0064] For example, after obtaining N optimal resume texts, the processor integrates these N optimal resume texts, job requirement texts, and target relevance scores into an information-rich reasoning context, which is then input into a reasoning generation model. Guided by specific prompts, the reasoning generation model simulates the chain-like thinking process of expert review, performing multi-dimensional, evidence-oriented reasoning, and ultimately outputting structured, explanatory matching reasons. This process transforms a single target relevance score into a highly transparent and logically sound judgment criterion, further enhancing the credibility and interpretability of the recommendations for the N optimal resume texts, and providing auxiliary information for users' accurate decision-making.
[0065] The technical solution of this application obtains key information about the job requirements and multiple resumes by inputting them into a key information extraction model. Then, it inputs these key information into a resume matching model to obtain relevance scores for the resumes. Finally, it determines several optimal resume texts based on these relevance scores. By extracting perceptual information from the job requirements and multiple resumes, long texts can be simplified and integrated, effectively shortening the input length of the resume matching model. This allows the resume matching model to perform efficient and accurate matching, thereby improving the accuracy of the matching results.
[0066] This application further proposes a resume matching device based on a large model, referring to... Figure 9 , Figure 9 This is a schematic diagram of the structure of the large-model-based resume matching device according to the embodiments of this application. In some embodiments, the large-model-based resume matching device includes: At least one processor; and, 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 resume matching method based on the large model described above.
[0067] In this embodiment, refer to Figure 9The large-model-based resume matching device in this application embodiment can be a processor capable of running a large-model-based resume matching method; there is at least one processor. Figure 9 As shown, the resume matching device based on a large model 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 enable 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 non-volatile memory, such as a disk drive. Optionally, the memory 1005 may also be a storage device independent of the aforementioned processor 1001.
[0068] Those skilled in the art will understand that Figure 9 The structure of the large-model-based resume matching device shown in the figure does not constitute a limitation on the large-model-based resume matching device. It may include more or fewer components than shown, or combine certain components, or have different component arrangements.
[0069] 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.
[0070] exist Figure 9 In the large-model-based resume matching 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, it implements the steps of the large-model-based resume matching method described above.
[0071] This application further proposes a storage medium storing a computer program, which includes program instructions that, when executed by a processor, enable the processor to execute the aforementioned resume matching method based on a large model.
[0072] In this embodiment, the resume matching method based on a large model can be configured as a computer program, which can be stored in a storage medium. When the storage medium is connected to a host computer, the processor of the host computer can execute the resume matching method based on a large model stored in the storage medium.
[0073] The above description is only a part or preferred embodiment of this application. Neither the text nor the drawings should limit the scope of protection of this application. All equivalent structural transformations made using the content of this application's specification and drawings under the overall concept of this application, or direct / indirect applications in other related technical fields, are included within the scope of protection of this application.
Claims
1. A resume matching method based on a large model, characterized in that, The resume matching method based on the large model includes: Retrieve the job requirements text and multiple resume texts to be matched; The job requirement text and multiple resume texts to be matched are input into the key information extraction model to obtain key job information and key resume information. The key job information and multiple key resume information are input into the resume matching model to obtain the relevance scores of the multiple key resume information. Based on the relevance scores, N optimal resume texts are determined from the multiple resume texts to be matched, and the N optimal resume texts are output, where N is a positive integer.
2. The resume matching method based on a large model according to claim 1, characterized in that, Before obtaining the job requirements text and multiple resume texts to be matched, the process also includes: Obtain the extraction model configuration parameters input by the user, and configure the initial key information extraction model according to the extraction model configuration parameters; Obtain multiple annotated job requirement texts, wherein the annotated job requirement texts include historical job requirement texts and job annotation information that annotates key information of historical jobs in the historical job requirement texts; Obtain multiple annotated resume texts, wherein the annotated resume texts include historical resume texts and resume annotation information that annotates key information of historical resumes in the historical resume texts; The initial key information extraction model is trained based on multiple labeled job requirement texts and multiple labeled resume texts to obtain the key information extraction model.
3. The resume matching method based on a large model according to claim 2, characterized in that, After training the initial key information extraction model based on multiple labeled job requirement texts and multiple labeled resume texts to obtain the key information extraction model, the process further includes: Input multiple historical job requirement texts and multiple historical resume texts into the key information extraction model to obtain multiple historical job key information and multiple historical resume key information. Obtain the pairing information input by the user, and construct multiple pairing relationships between multiple key information of historical positions and multiple key information of historical resumes based on the pairing information to obtain multiple data pairs, wherein the data pair refers to the data pair of key information of historical positions and key information of historical resumes; The resume matching model is obtained by training a preset mGTE text representation model based on multiple datasets.
4. The resume matching method based on a large model according to claim 1, characterized in that, The step of inputting the job requirement text into the key information extraction model to obtain key job information includes: Input the job requirement text into the key information extraction model; The key information extraction model outputs the educational background requirements, work experience requirements, professional skills requirements, and working conditions provided. The educational background requirements, work experience requirements, professional skills requirements, and working conditions provided are considered as key information for the position.
5. The resume matching method based on a large model according to claim 4, characterized in that, The process involves inputting multiple resume texts to be matched into a key information extraction model to obtain multiple key resume information, including: For any of the resume texts to be matched, input the resume texts to be matched into the key information extraction model; Obtain the educational background, work experience, professional skills, and expected working conditions output by the key information extraction model; The educational background, work experience, professional skills, and expected working conditions are used as key information in the resume.
6. The resume matching method based on a large model according to claim 1, characterized in that, The step of inputting the key job information and multiple key resume information into a resume matching model to obtain relevance scores for multiple key resume information includes: For any of the aforementioned key information in the resume, the key information in the job posting and the key information in the resume are input into the resume matching model; Obtain the encoded representations of the key information of the resume and the key information of the job position output by the resume matching model; The relevance score of the key information in the resume is calculated based on the encoded representation.
7. The resume matching method based on a large model according to claim 1, characterized in that, The step of determining N optimal resume texts from multiple resume texts to be matched based on the relevance score includes: Based on the key information of each resume, the relevance scores are associated with the text of each resume to be matched; Sort the relevance scores in descending order to obtain a sorted table; The top N relevance scores are determined as N target relevance scores; For any of the target relevance scores, the resume text to be matched associated with the target relevance score is determined as the optimal resume text.
8. The resume matching method based on a large model according to claim 7, characterized in that, The output includes N optimal resume texts, including: Based on the N optimal resume texts, the job requirement text, and the target relevance scores of the N optimal resume texts, generate N matching reasons for the optimal resume texts; Output N optimal resume texts and N matching reasons for the optimal resume texts.
9. A resume matching 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 that are executed by the at least one processor to enable the at least one processor to perform the resume matching 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 that, when executed by a processor, enable the processor to perform the resume matching method based on a large model as described in any one of claims 1 to 8.