Resume recommendation method, resume recommendation system, medium and electronic equipment
By extracting multiple features of resume files and calling corresponding recall strategies, generating a collection of resumes to be recommended and performing quality scores, the problem of inaccurate resume recommendations in the existing technology is solved, and more efficient and accurate resume recommendations are achieved.
Patent Information
- Application Number
- CN202510109602.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-01-23
- Publication Date
- 2025-05-27
AI Technical Summary
The existing resume recommendation system relies on keyword retrieval and simple text matching, which leads to inaccurate recommendations and makes it difficult to deal with massive resumes.
By extracting the entity features, intelligent label features and text content vector features of the resume file, calling their respective resume recall strategies, generating a collection of resumes to be recommended, and scoring and filtering the resume quality.
It improves the accuracy of resume recommendations, avoids excessive homogeneity of recommendation results, meets the resume recommendation needs of different types of positions, and has a high resume recommendation accuracy.
Smart Images
Figure CN120045784A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of data processing, and particularly to a resume recommendation method, a resume recommendation system, a medium, and an electronic device. Background Art
[0002] Previously, enterprises mainly relied on a purely manual method for resume screening. Due to reasons such as different understandings of job requirements and evaluation criteria by recruiters and limited personal processing speed of resume information, it was difficult to process a large number of resumes.
[0003] In recent years, attempts have been made to apply resume recommendation systems to improve resume screening efficiency. Such systems can recommend resumes to recruiters for selection based on job requirements according to the matching degree, saving the time for recruiters to browse all resumes and improving the enterprise recruitment efficiency. Most current resume recommendation systems mainly rely on keyword retrieval and simple text matching methods. Due to the limitations of retrieval, there are problems with inaccurate recommendations. Summary of the Invention
[0004] The purpose of this application is to provide a resume recommendation method, a resume recommendation system, a computer-readable storage medium, and an electronic device, which can improve the accuracy of resume recommendations.
[0005] To solve the above technical problems, this application provides a resume recommendation method, and the specific technical solutions are as follows:
[0006] Obtain a resume file;
[0007] Extract the entity features, intelligent label features, and text content vector features of the resume file;
[0008] According to the entity features, the intelligent label features, and the text content vector features, call their respective corresponding resume recall strategies to obtain a set of resumes to be recommended;
[0009] Perform a resume quality score on the target resumes in the set of resumes to be recommended, and screen out the resumes with a resume quality score greater than a set threshold as recommended resumes.
[0010] Optionally, after obtaining the resume file, it further includes:
[0011] Obtain resume parameters;
[0012] Extract the parameter values in the resume file based on the resume parameters, and establish a mapping relationship with the resume parameters;
[0013] Set resume layout data;
[0014] Generate structured resume data of the resume file according to the mapping relationship and the resume layout data.
[0015] Optionally, the extraction process of the entity features includes:
[0016] Extract the personal information in the resume file;
[0017] Filter the work experience information related to the job information in the resume file;
[0018] Generate entity features based on the personal information and the work experience information.
[0019] Optionally, the extraction process of the intelligent label features includes:
[0020] Call a language model to extract the skill information in the resume file;
[0021] Perform context classification according to the skill information and set classification labels;
[0022] Assign proficiency scores to the assigned labels;
[0023] Use the skill information, classification labels, and the quantity scores as the intelligent label features.
[0024] Optionally, the extraction process of the text content vector features includes:
[0025] Extract the resume file and map the parameter values corresponding to the resume parameters into corresponding vector representations;
[0026] Determine the resume number of the resume file;
[0027] Use the resume number and all the included vector representations as the text content vector features.
[0028] Optionally, according to the entity features, the intelligent label features, and the text content vector features, call their respective corresponding resume recall strategies, and the obtained set of resumes to be recommended includes:
[0029] Obtain resume recommendation requirements;
[0030] For the entity features, call a precise recall strategy to filter out target resumes that meet the entity features corresponding to the resume recommendation requirements as the set of resumes to be recommended;
[0031] For the intelligent label features, determine the job labels corresponding to the resume recommendation requirements, calculate the label similarity with the intelligent label features, and construct a set of resumes to be recommended according to the resume files with label similarity greater than the set similarity threshold;
[0032] For the text content vector features, extract the demand vector features corresponding to the resume recommendation requirements, calculate the feature distance between the demand vector features and the text content vector features, and determine the set of resumes to be recommended according to the feature distance.
[0033] Optionally, after obtaining the resume recommendation requirements, it further includes:
[0034] Determine the weights corresponding to each resume parameter in the resume recommendation requirements;
[0035] Prioritize screening the resume files that meet the resume parameters with higher weights as the resumes to be recommended in the set of resumes to be recommended.
[0036] This application also provides a resume recommendation system, including:
[0037] A resume acquisition module for acquiring resume files;
[0038] A feature extraction module for extracting the entity features, intelligent label features, and text content vector features of the resume files;
[0039] A resume recall module for calling the respective corresponding resume recall strategies according to the entity features, the intelligent label features, and the text content vector features to obtain a set of resumes to be recommended;
[0040] A resume recommendation module for performing a resume quality score on the target resumes in the set of resumes to be recommended, and screening the resumes with a resume quality score greater than a set threshold as the recommended resumes.
[0041] This application also provides a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, the steps of the method described above are implemented.
[0042] This application also provides an electronic device, including a memory and a processor, where a computer program is stored in the memory, and when the processor calls the computer program in the memory, the steps of the method described above are implemented.
[0043] This application provides a resume recommendation method, including: acquiring resume files; extracting the entity features, intelligent label features, and text content vector features of the resume files; calling the respective corresponding resume recall strategies according to the entity features, the intelligent label features, and the text content vector features to obtain a set of resumes to be recommended; performing a resume quality score on the target resumes in the set of resumes to be recommended, and screening the resumes with a resume quality score greater than a set threshold as the recommended resumes.
[0044] After obtaining the resume file, the present application extracts the feature information of the resume file from different perspectives, and adopts corresponding recall strategies for each type of feature information, which can make the established recommendation results more diverse, avoid the over-homogenization of the recommendation results, and at the same time, through the recall complementarity of multiple recall strategies, it can avoid the poor applicability of a single recall strategy to certain types of positions, has high flexibility, can meet the resume recommendation needs of different types of positions, and has a high resume recommendation accuracy rate.
[0045] The present application also provides a resume recommendation system, a computer-readable storage medium, and an electronic device, which have the above beneficial effects and will not be elaborated here. BRIEF DESCRIPTION OF THE DRAWINGS
[0046] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the drawings in the following description are only the embodiments of the present application. For those of ordinary skill in the art, other drawings can be obtained according to the provided drawings without creative efforts.
[0047] Figure 1 It is a flowchart of a resume recommendation method provided by an embodiment of the present application;
[0048] Figure 2 It is a schematic structural diagram of a resume recommendation system provided by an embodiment of the present application. DETAILED DESCRIPTION OF THE EMBODIMENTS
[0049] To make the objectives, technical solutions, and advantages of the embodiments of the present application clearer, the following will clearly and completely describe the technical solutions in the embodiments of the present application with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are some, but not all, of the embodiments of the present application. Based on the embodiments of the present application, all other embodiments obtained by those of ordinary skill in the art without creative efforts fall within the scope of protection of the present application.
[0050] Please refer to Figure 1 , Figure 1 It is a flowchart of a resume recommendation method provided by an embodiment of the present application
[0051] S101: Obtain a resume file;
[0052] S102: Extract the entity features, intelligent label features, and text content vector features of the resume file;
[0053] S103: According to the entity features, the intelligent label features, and the text content vector features, call their respective corresponding resume recall strategies to obtain a set of resumes to be recommended;
[0054] S104: Perform a resume quality score on the target resumes in the to-be-recommended resume set, and screen out the resumes with a resume quality score greater than the set threshold as the recommended resumes.
[0055] Here, there is no limitation on how to obtain the resume file. It can be obtained through forms such as scheduled tasks and long-running tasks to obtain existing resume files or newly added resume files in the enterprise resume talent pool.
[0056] After that, the entity features, intelligent label features, and text content vector features of the resume file can be extracted.
[0057] For entity features, the following process can be adopted:
[0058] First step: Extract the personal information in the resume file;
[0059] Second step: Screen out the work experience information related to the position information in the resume file;
[0060] Third step: Generate entity features based on the personal information and the work experience information.
[0061] When extracting entity features, the large language model can be utilized as it has demonstrated excellent capabilities in tasks such as semantic understanding, text generation, and knowledge answering. It can perform key information extraction and structured processing on the resume that has been converted into text format, including but not limited to personal information (name, contact information, etc.), educational background (school name, educational level, study time, etc.), work experience (company name, position, working time, and duty description, etc.). Through the structured processing of this information, the system can generate a standardized resume format for subsequent analysis and processing. Due to the diversity of resume formats and content, it becomes a challenging task to automatically parse the key information in the resume, especially the work experience information. To improve the accuracy and standardization of parsing, a database containing common industry position types can be constructed. Preferably, during parsing, the standardized position category information is provided to the large model as context, enabling the large model to parse the candidate's work and project experiences according to the provided position categories to obtain standardized results.
[0062] Specifically, a feasible design of the prompt text (the Prompt text is the text or instruction input by the user when interacting with the artificial intelligence model to guide the model to generate a specific output) is as follows:
[0063] "Please extract the key information based on the candidate's resume.
[0064] Instruction:
[0065] 1. Extract the candidate's personal information;
[0066] 2. Extract the educational background information of the candidates;
[0067] 3. Extract the work experience information of the candidates according to the provided job categories;
[0068] Context;
[0069] The following is the resume of the candidate:
[0070] {Resume};
[0071] The work experience must be selected from the following categories
[0072] {Job category};
[0073] Output:
[0074] When replying to me, it must be output in the following JSON format:
[0075] {Output format};
[0076] ...”
[0077] For the intelligent label feature, its extraction process can be:
[0078] Step 1: Call the language model to extract the skill information in the resume file;
[0079] Step 2: Classify the context according to the skill information and set classification labels;
[0080] Step 3: Assign proficiency scores to the assigned labels;
[0081] Step 4: Use the skill information, classification labels and the quantity scores as the intelligent label features.
[0082] The intelligent label features can utilize the powerful natural language processing capabilities of the large language model to automatically extract skill information from the resume file, set classification labels, and improve the efficiency and accuracy of the recruitment process. For the resume that has been converted into text format, utilize the context understanding ability of the large language model to identify the skill-related information in the resume, extract skill phrases from the resume file, classify and label them according to the context, and assign proficiency scores to each extracted skill label. Thereafter, the extracted skill labels and their proficiency scores can be output and stored in a structured format.
[0083] For the text content vector features, its extraction process can include:
[0084] Step 1: Extract the resume file and map the parameter values corresponding to the resume parameters into corresponding vector representations;
[0085] Step 2: Determine the resume number of the resume file;
[0086] Step 3: Use the resume number and all the corresponding vector representations it contains as the vector feature of the text content.
[0087] For example, the resume text included in the enterprise talent pool can be read, and the text content including but not limited to work experience, internship experience, former job responsibilities, and experience in projects participated in can be extracted from each resume one by one. The above-mentioned text content extracted from this resume is saved in segments for subsequent text vectorization operations.
[0088] Apply the pre-trained text embedding model text-embedding-large-3 to the text content saved in segments one by one to map it into the corresponding vector representation, and extract this vector representation. Here, the pre-trained text embedding model can also be replaced with models such as BERT and RoBERTa. Here, the vector representation is actually a low-dimensional dense digital vector predicted and generated by the pre-trained text embedding model according to the text input. For the mapping of unstructured text content, this vector representation has mathematical properties. Assuming that the text content input into the model has similar meanings, the vector representations predicted and generated by the model are more similar, which is represented as a closer distance between vectors in terms of data.
[0089] After that, the vector representation and the resume number to which the vector belongs can be stored in the vector database for subsequent vector retrieval and resume original text backtracking.
[0090] When obtaining the set of resumes to be recommended, the resume recommendation requirements can be obtained first, and then the corresponding retrieval strategies can be adopted according to different types of features. The user recommendation request can be divided into two parts: structured request and unstructured request. The request refers to including standardized job requirements, including the range of job types, work experience requirements, educational background requirements, age requirements, etc. The unstructured request includes job descriptions and job requirements in text form. For this part of the content, the large language model can extract a set of skill requirements from it and label them for subsequent label retrieval. The specific process can be as follows:
[0091] For the entity features, call the precise recall strategy to screen target resumes that meet the entity features corresponding to the resume recommendation requirements as the set of resumes to be recommended. At this time, according to the standardized job requirements provided by the user, matching resume documents can be quickly and accurately queried. As one of the resume recommendation strategies, according to the set screening conditions, match with the standardized information of each resume in the resume pool to quickly and accurately find resumes that fully or partially meet the conditions. Further, for resumes that partially meet the conditions, the user can customize the weights of different conditions, calculate the matching degree with the position according to the weights, and perform sorting to select the most suitable one. The precise retrieval module can quickly locate target resumes that meet specific screening conditions in a large amount of resume data and provide a list of candidates most likely to meet the recruitment requirements through predefined matching degree calculation rules, greatly improving the efficiency and quality of the recruitment work.
[0092] For the intelligent label features, determine the position labels corresponding to the resume recommendation requirements, calculate the label similarity with the intelligent label features, and construct a set of resumes to be recommended according to the resume files with label similarity greater than the set similarity threshold. At this time, the extracted position labels can be matched with the resume labels in the database. Each resume in the database has also been parsed by the large model to extract the corresponding labels. To improve the accuracy of the matching, multiple similarity calculation methods are used to ensure the diversity and reliability of the recommendation results.
[0093] In a feasible similarity calculation method, a matching algorithm based on the Boolean model can be used. In this method, the position labels and resume labels are regarded as Boolean values (present or absent). By calculating the Boolean matching degree between the position labels and resume labels, their similarity is evaluated. Specifically, the ratio of the intersection to the union of the position labels and resume labels can be calculated as the similarity score. In another feasible implementation, a matching algorithm based on the Edit Distance can be used. The Edit Distance refers to the minimum number of edit operations required to convert one string to another, including insertion, deletion, and replacement operations. By calculating the Edit Distance between the position labels and resume labels, their similarity is evaluated. The smaller the Edit Distance, the higher the similarity. Finally, there is also a matching algorithm based on the Jaccard similarity coefficient. The Jaccard similarity coefficient refers to the ratio of the intersection to the union of two sets. By calculating the Jaccard similarity coefficient of the position labels and resume labels, their matching degree is evaluated. The higher the Jaccard similarity coefficient, the higher the similarity.
[0094] In addition, during the similarity calculation process, a weight factor can also be introduced to weight the importance of different labels. For example, for some key skill labels, higher weights can be assigned to ensure that these labels occupy a larger proportion in the similarity calculation.
[0095] To further optimize the recommendation effect, a feedback mechanism can also be introduced. The feedback information of users on the recommendation results during the use process will be recorded and used to adjust the label weights of each one, thereby continuously improving the recommendation accuracy.
[0096] After completing the similarity calculation, sort the resumes according to the similarity scores and generate a set of recommendation results. The resumes in the set of recommendation results are arranged from high to low according to the similarity scores to ensure that the resumes that best meet the job requirements are displayed to users first.
[0097] For the vector feature of the text content, extract the demand vector feature corresponding to the resume recommendation requirement, calculate the feature distance between the demand vector feature and the text content vector feature, and determine the set of resumes to be recommended according to the feature distance. Specifically, a pre-trained text embedding model can be applied to extract the vector representation of the text content related to the position to be recommended, and this text content can be job responsibilities, job requirements, high-quality excellent resume templates, etc.
[0098] Perform vector retrieval on the vector database, calculate the dot product distance to describe the similarity of two vector representations, and the dot product distance can also be replaced by other methods such as Euclidean distance and cosine distance to calculate the similarity. Take the top k resumes ranked from high to low in similarity as the recall result of this strategy.
[0099] When scoring the quality of the target resume in the set of resumes to be recommended, the resume of the candidate can be input into a large model for parsing. The large model will perform structured processing on each part of the resume and extract key information such as the candidate's educational background, work experience, and project experience. Next, score these information item by item according to the subjective scoring criteria preset by the recruiter. First, evaluate the candidate's highest education level and the type of graduating institution based on the educational experience. Give corresponding scores according to different educational levels and institution types. Then, based on the candidate's work experience, evaluate the company background where the candidate has worked. Give corresponding scores according to the company's popularity and scale. Finally, evaluate the number of projects participated by the candidate and the project achievements based on the project experience. Give corresponding scores according to the number of projects and the achievements obtained in the projects.
[0100] Specifically, an exemplary prompt text design is as follows:
[0101] "Please score the candidate according to the candidate's resume.
[0102] Instruction:
[0103] 1. Extract candidate information (educational background, work experience, and project experience);
[0104] 2. Score the candidates according to the provided {scoring criteria};
[0105] Context:
[0106] The following is the resume of the candidate:
[0107] {Resume};
[0108] Please score the candidates according to the {scoring criteria}, and the scoring criteria are as follows:
[0109] {Scoring criteria};
[0110] Output:
[0111] When replying to me, it must be output in the following JSON format;
[0112] {Output format};
[0113] ...”
[0114] After obtaining the resume file in the embodiment of the present application, extracting the feature information of the resume file from different perspectives, and adopting corresponding recall strategies for each type of feature information can make the established recommendation results more diverse, avoid the over-homogenization of the recommendation results, and at the same time, through the recall complementarity of multiple recall strategies, it is possible to avoid the poor applicability of a single recall strategy to certain types of positions, have high flexibility, be able to meet the resume recommendation needs of different types of positions, and have a high resume recommendation accuracy.
[0115] In addition, when scoring the quality of the target resume, the subjective experience of the recruiter can be effectively utilized. The recruiter can transform their recruitment requirements into scoring criteria and inject them into the system. The automated resume scoring module can quickly score and re-rank the recall results, further improving the recruitment efficiency. By transforming the recruitment requirements into scoring criteria, the resume scoring module can ensure that each resume is evaluated according to a unified standard, avoiding the subjective biases and inconsistencies that may occur in the manual screening process. The recruiter can flexibly adjust the scoring criteria and weights according to the specific requirements of different positions, ensuring that the scoring module can adapt to various recruitment scenarios and requirements, making the present invention have better flexibility and meeting the customized requirements of different positions.
[0116] In an alternative embodiment, after obtaining the resume file, the resume can be structured. Specifically, resume parameters can be obtained, parameter values in the resume file can be extracted based on the resume parameters, and a mapping relationship with the resume parameters can be established. At the same time, resume layout data can be set, and finally, structured resume data of the resume file can be generated according to the mapping relationship and the resume layout data.
[0117] See Figure 2 , Figure 2A schematic structural diagram of a resume recommendation system provided by an embodiment of the present application is shown below. A resume recommendation system provided by an embodiment of the present application will be introduced hereinafter. The resume recommendation system described below can be correspondingly referred to the resume recommendation method described above.
[0118] A resume acquisition module, configured to acquire resume files;
[0119] A feature extraction module, configured to extract entity features, intelligent label features, and text content vector features of the resume file;
[0120] A resume recall module, configured to call respective corresponding resume recall strategies according to the entity features, the intelligent label features, and the text content vector features, and obtain a set of resumes to be recommended;
[0121] A resume recommendation module, configured to perform resume quality scoring on the target resumes in the set of resumes to be recommended, and screen out the resumes with resume quality scores greater than a set threshold as recommended resumes.
[0122] Based on the above embodiments, as a preferred embodiment, after acquiring the resume file, it further includes:
[0123] A preprocessing module, configured to acquire resume parameters; extract parameter values in the resume file based on the resume parameters, and establish a mapping relationship with the resume parameters; set resume layout data; generate structured resume data of the resume file according to the mapping relationship and the resume layout data.
[0124] Based on the above embodiments, as a preferred embodiment, the feature extraction module is a module for performing the following steps:
[0125] Obtain resume recommendation requirements;
[0126] For the entity features, call a precise recall strategy to screen out target resumes that meet the entity features corresponding to the resume recommendation requirements as the set of resumes to be recommended;
[0127] For the intelligent label features, determine the job labels corresponding to the resume recommendation requirements, calculate the label similarity with the intelligent label features, and construct a set of resumes to be recommended according to the resume files with label similarity greater than a set similarity threshold;
[0128] For the text content vector features, extract the demand vector features corresponding to the resume recommendation requirements, calculate the feature distance between the demand vector features and the text content vector features, and determine the set of resumes to be recommended according to the feature distance.
[0129] Based on the above embodiments, as a preferred embodiment, it further includes:
[0130] The weight setting module is used to determine the weight corresponding to each resume parameter in the resume recommendation requirement;
[0131] Resume files that meet resume parameters with higher weights are preferentially screened as the resumes to be recommended in the set of resumes to be recommended.
[0132] This system can be divided into two scenarios: offline and online. In the offline scenario, the system mainly performs offline preprocessing tasks, while in the online scenario, the system mainly triggers real-time resume recommendation tasks triggered by users.
[0133] The specific operations in the offline scenario are as follows:
[0134] The system continuously runs an offline preprocessing task to synchronously read the resume texts included in the enterprise talent pool. The system calls the feature preprocessing module to extract the entity features, smart label features and text content vector features required for subsequent recommendations from each resume text, and stores the above three types of features in the corresponding database respectively.
[0135] The specific operations in the online scenario are as follows:
[0136] Users can configure the web front-end provided by this system to recommend relevant job responsibilities, job requirements, excellent resume templates and other information, and then initiate a resume recommendation request. Based on the job responsibilities, job requirements, excellent resume templates and other content carried in the request, this system first calls the preprocessing module to preprocess the request, and can parse this request into the entities, tags, text embedding vectors and other data required for this recommendation. This part of data is the metadata of this recommendation request.
[0137] Then, the system takes the recommended metadata parsed by the preprocessing module and calls the resume recall module to distribute different types of metadata to the corresponding recall strategy components. The recall strategy component retrieves and compares the similarity between the request metadata and the resume metadata in the database according to the corresponding metadata, sorts them from high to low according to the similarity, and recalls the top k resumes. In addition, the multi-strategy recall module will merge the recall results of multiple components to form a collection of resumes to be recommended.
[0138] Finally, the system takes the resume collection to be recommended obtained by the resume recall module and calls the resume scoring module. The resume recommendation module calls the big model one by one in the resume collection to be recommended. The big model comprehensively scores the resumes based on the scoring criteria pre-specified by the recruiter, from the aspects of resume content completeness, educational background matching, work experience matching, etc. The top n resumes are sorted from high to low according to the scores and recalled.
[0139] The present application also provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed, the steps provided in the above embodiments can be implemented. The storage medium may include: various media that can store program codes, such as USB flash drives, mobile hard disks, read-only memories (ROMs), random access memories (RAMs), magnetic disks, or optical discs.
[0140] The present application also provides an electronic device, which may include a memory and a processor. A computer program is stored in the memory. When the processor calls the computer program in the memory, the steps provided in the above embodiments can be implemented. Of course, the electronic device may further include various network interfaces, power supplies, and other components.
[0141] The various embodiments in the specification are described in a progressive manner. The key point of each embodiment is to illustrate the differences from other embodiments. For the same or similar parts among the various embodiments, reference can be made to each other. For the system provided in the embodiment, since it corresponds to the method provided in the embodiment, the description is relatively simple, and reference can be made to the description in the method part for related parts.
[0142] Specific examples are used in this article to elaborate on the principles and implementation manners of the present application. The descriptions of the above embodiments are only used to help understand the method and its core idea of the present application. It should be noted that for those of ordinary skill in the art in the technical field, without departing from the principle of the present application, several improvements and modifications can be made to the present application, and these improvements and modifications also fall within the protection scope of the present application.
[0143] It should also be noted that in this specification, relational terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the term "comprising", "including" or any other variant thereof is intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements not only includes those elements, but also includes other elements not expressly listed, or further includes elements inherent to such process, method, article or device. Without further limitation, an element defined by the statement "including a..." does not exclude the existence of additional identical elements in the process, method, article or device including the said element.
Claims
1. A resume recommendation method, characterized in that: include: Get resume files; Extracting entity features, intelligent tag features, and text content vector features of the resume file; According to the entity features, the smart tag features and the text content vector features, the corresponding resume recall strategies are called to obtain a set of resumes to be recommended; The target resumes in the set of resumes to be recommended are scored for resume quality, and resumes with resume quality scores greater than a set threshold are selected as recommended resumes.
2. The resume recommendation method according to claim 1, characterized in that: After obtaining the resume file, it also includes: Get resume parameters; Extracting parameter values in the resume file based on the resume parameters, and establishing a mapping relationship with the resume parameters; Set resume layout data; The structured resume data of the resume file is generated according to the mapping relationship and the resume typesetting data.
3. The resume recommendation method according to claim 1, characterized in that: The entity feature extraction process includes: Extracting personal information from the resume file; Screening the work experience information related to the position information in the resume file; An entity feature is generated based on the personal information and the work experience information.
4. The resume recommendation method according to claim 1, characterized in that: The process of extracting the smart tag features includes: Calling a language model to extract skill information from the resume file; Performing context classification according to the skill information and setting classification labels; assigning a proficiency score to the assigned tags; The skill information, the classification label and the quantitative score are used as the smart label features.
5. The resume recommendation method according to claim 2, characterized in that: The process of extracting the text content vector feature includes: Extract the resume file, and map the parameter values corresponding to the resume parameters into corresponding vector representations; Determine the resume number of the resume file; The resume number and all corresponding vector representations are used as the text content vector features.
6. The resume recommendation method according to claim 1, characterized in that: According to the entity feature, the smart tag feature and the text content vector feature, the corresponding resume recall strategies are called to obtain a set of resumes to be recommended, including: Obtain resume recommendation requirements; For the entity features, a precise recall strategy is called to select target resumes that meet the entity features corresponding to the resume recommendation requirements as a set of resumes to be recommended; For the smart tag feature, determine the job tag corresponding to the resume recommendation requirement, calculate the tag similarity with the smart tag feature, and build a set of resumes to be recommended based on resume files whose tag similarity is greater than a set similarity threshold; For the text content vector features, the demand vector features corresponding to the resume recommendation requirements are extracted, the feature distance between the demand vector features and the text content vector features is calculated, and the resume set to be recommended is determined according to the feature distance.
7. The resume recommendation method according to claim 6, characterized in that: After obtaining resume recommendation requirements, it also includes: Determine the weights corresponding to each resume parameter in the resume recommendation requirements; Resume files that meet resume parameters with higher weights are preferentially screened as the resumes to be recommended in the resume set to be recommended.
8. A resume recommendation system, characterized in that: include: Resume acquisition module, used to obtain resume files; A feature extraction module, used to extract entity features, smart tag features and text content vector features of the resume file; A resume recall module, used to call the corresponding resume recall strategies according to the entity features, the smart tag features and the text content vector features, to obtain a set of resumes to be recommended; The resume recommendation module is used to score the target resumes in the set of resumes to be recommended, and select resumes with resume quality scores greater than a set threshold as recommended resumes.
9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.
10. An electronic device, characterized in that: The method comprises a memory and a processor, wherein a computer program is stored in the memory, and when the processor calls the computer program in the memory, the steps of the method according to any one of claims 1 to 7 are implemented.