Resume generation method and device, electronic equipment, storage medium and program product

Abstract

This application relates to the field of resume generation technology, and in particular to a resume generation method, apparatus, electronic device, storage medium, and program product. The method includes: acquiring at least one of a user's voice introduction, work uniform photo, and behavioral interaction data from a recruitment platform; extracting the user's professional characteristics from at least one of the voice introduction, work uniform photo, and behavioral interaction data; filling the professional characteristics into corresponding professional tags; and generating the user's resume based on the filled-in professional tags. This solves the problems of related technologies requiring users to manually fill in resumes, resulting in high operational barriers and limited applicability to certain groups.

Description

Technical Field

[0001] This application relates to the field of resume generation technology, and in particular to a resume generation method, apparatus, electronic device, storage medium, and program product. Background Technology

[0002] Resumes are an important medium for collecting and matching talent information. The existing resume creation mode generally adopts the form of manually filling in structured forms and is applied in various talent information sorting scenarios. The implementation of this mode is highly dependent on the user's text expression ability and form operation ability, and there are natural differences in adaptability to people with different ability characteristics.

[0003] In related technologies, resume generation mainly relies on users manually filling out structured forms. This method significantly raises the barrier to resume creation. For people with weak writing skills and unfamiliar with electronic form operations, it is difficult to independently complete a standardized and complete resume. As a result, many such people are unable to overcome this barrier, leading to missing or incomplete resume information and an inability to effectively retain talent information. Summary of the Invention

[0004] This application provides a resume generation method, apparatus, electronic device, storage medium, and program product to solve the problems in related technologies, such as the need for users to manually fill in resumes, high operational threshold, and limited applicable group.

[0005] The first aspect of this application provides a resume generation method, including the following steps: obtaining at least one of a user's voice introduction, work uniform photo, and behavioral interaction data from a recruitment platform; extracting the user's professional characteristics from at least one of the voice introduction, work uniform photo, and behavioral interaction data; filling the professional characteristics into corresponding professional tags; and generating the user's resume based on the filled professional tags.

[0006] Preferably, the user's occupational characteristics are extracted from at least one of the voice introduction, work uniform photos, and behavioral interaction data, including: extracting the user's first occupational characteristics from the voice introduction; extracting the user's second occupational characteristics from the work uniform photos; extracting the user's third occupational characteristics from the behavioral interaction data; and fusing the first, second, and third occupational characteristics to obtain the final occupational characteristics.

[0007] Preferably, in one embodiment of this application, extracting a user's first occupational feature from the voice introduction includes: inputting the voice introduction into a target speech recognition model, the target speech recognition model outputting text; and extracting the user's first occupational feature from the text using rule matching and keyword matching.

[0008] Preferably, in one embodiment of this application, extracting a user's second occupational feature from a work uniform photo includes: inputting the work uniform photo into a target multi-task detection network, the target multi-task detection network outputting the user's occupational detection result, wherein the target multi-task detection network uses a Swing Transformer as its backbone network, and the occupational detection result includes at least one of job probability distribution, tool recognition result, environmental category, and safety compliance; inputting the occupational detection result into a visual-text joint understanding model, the visual-text joint understanding model matching the occupational detection result with occupational features in a preset occupational feature description library, and outputting the user's second occupational feature.

[0009] Preferably, in one embodiment of this application, extracting a user's third occupational characteristic from behavioral interaction data includes: identifying the user's behavior type and corresponding behavior content in the behavioral interaction data, wherein the behavior type includes at least one of submission, click, browsing, and searching; determining multiple occupational information tags based on the behavior type and behavior content; using a target occupational knowledge graph to perform reasoning mapping on the occupational information tags to complete the occupational information tags; obtaining the confidence level of each occupational information tag, and calculating the confidence level of the corresponding occupational information tag based on the confidence level of the occupational information tag and the weight of the corresponding behavior type; and determining the user's third occupational characteristic based on the confidence level of each occupational information tag.

[0010] Preferably, after generating the user's resume based on the completed occupational tags, the method further includes: identifying recruitment information for multiple job postings on the recruitment platform; using a rule engine and natural language processing to parse the recruitment information in the job postings into corresponding occupational characteristics; calculating the matching degree between the user's occupational characteristics and the job postings' occupational characteristics; and if the matching degree is greater than a preset matching degree, pushing the job postings to the user.

[0011] A second aspect of this application provides a resume generation device, comprising: an acquisition module for acquiring at least one of a user's voice introduction, work uniform photo, and behavioral interaction data from a recruitment platform; an extraction module for extracting the user's professional characteristics from at least one of the voice introduction, work uniform photo, and behavioral interaction data; and a generation module for filling the professional characteristics into corresponding professional tags and generating the user's resume based on the filled professional tags.

[0012] Preferably, the extraction module is further used to: extract the user's first occupational feature from the voice introduction; extract the user's second occupational feature from the work uniform photo; extract the user's third occupational feature from the behavioral interaction data; and fuse the first occupational feature, the second occupational feature, and the third occupational feature to obtain the final occupational feature.

[0013] Preferably, the extraction module is further configured to: input the voice introduction into the target speech recognition model, the target speech recognition model outputs text; and extract the user's primary occupational feature from the text using rule matching and keyword matching.

[0014] Preferably, the extraction module is further configured to: input the workwear photo into a target multi-task detection network, the target multi-task detection network outputs the user's occupation detection result, wherein the target multi-task detection network uses a Swing Transformer as its backbone network, and the occupation detection result includes at least one of the following: job probability distribution, tool recognition result, environmental category, and safety compliance; input the occupation detection result into a visual-text joint understanding model, the visual-text joint understanding model matches the occupation detection result with occupation features in a preset occupation feature description library, and outputs the user's second occupation feature.

[0015] Preferably, the extraction module is further configured to: identify the user's behavior type and corresponding behavior content in the behavioral interaction data, wherein the behavior type includes at least one of submission, click, browsing, and search; determine multiple occupational information tags based on the behavior type and behavior content; use a target occupational knowledge graph to perform reasoning mapping on the occupational information tags to complete the occupational information tags; obtain the confidence score of each occupational information tag, and calculate the confidence score of the corresponding occupational information tag based on the confidence score of the occupational information tag and the weight of the corresponding behavior type; and determine the user's third occupational characteristic based on the confidence score of each occupational information tag.

[0016] Preferably, it also includes: a push module, used to identify recruitment information of multiple job postings on the recruitment platform after generating the user's resume based on the filled-in occupational tags; use a rule engine and natural language processing to parse the recruitment information in the job postings into corresponding occupational characteristics; calculate the matching degree between the user's occupational characteristics and the job postings' occupational characteristics; if the matching degree is greater than a preset matching degree, then push the job postings to the user.

[0017] A third aspect of this application provides an electronic device, including: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to perform the resume generation method as described in the above embodiments.

[0018] A fourth aspect of this application provides a computer-readable storage medium having a computer program or instructions stored thereon, which are executed by a processor to perform the resume generation method as described in the above embodiments.

[0019] A fifth aspect of this application provides a computer program product, including a computer program or instructions, which, when executed, implement the resume generation method as described in the above embodiments.

[0020] Therefore, this application has at least the following beneficial effects: This application embodiment can acquire at least one of the following: a user's voice introduction, a photo of them in work clothes, and behavioral interaction data from a recruitment platform. It then extracts the user's professional characteristics from these sources, fills them into corresponding professional tags, and generates the user's resume based on the completed professional tags. This eliminates the need for manual resume creation by the user, achieving automatic resume generation and lowering the barrier to entry for resume creation, thus adapting to users with different levels of education and knowledge. Therefore, it solves the technical problems of related technologies, such as the need for users to manually fill in resumes, high operational barriers, and limited applicability to certain groups.

[0021] Additional aspects and advantages of this application will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of this application. Attached Figure Description

[0022] The above and / or additional aspects and advantages of this application will become apparent and readily understood from the following description of the embodiments taken in conjunction with the accompanying drawings, wherein: Figure 1 This is a flowchart of a resume generation method provided according to an embodiment of this application; Figure 2 This is a flowchart illustrating the specific process of resume generation according to the embodiments of this application; Figure 3 This is a flowchart illustrating the operation of the resume generation system provided according to an embodiment of this application. Figure 4 This is an example diagram of a resume generation apparatus provided according to an embodiment of this application; Figure 5 This is a schematic diagram of the structure of an electronic device provided according to an embodiment of this application. Detailed Implementation

[0023] The embodiments of this application are described in detail below. Examples of these embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain this application, and should not be construed as limiting this application.

[0024] Before describing the solution of this application, let me first introduce the relevant technologies of this application to help you understand the solution of this application.

[0025] Currently, recruitment platforms targeting blue-collar workers generally use the traditional resume format, requiring users to manually fill in structured fields such as name, job title, years of work experience, and skill descriptions. However, many frontline workers, due to limitations in education level, communication skills, or operational habits, find it difficult to complete this type of filling efficiently, resulting in incomplete or missing resumes. Some platforms have attempted to introduce voice input functions (such as "voice resume" tools), allowing users to dictate their experiences and have them transcribed into text. However, these methods only offer simple voice-to-text conversion and do not integrate with other user behavior data on the platform (such as browsing, clicking, and application records), nor can they automatically extract structured professional elements.

[0026] Meanwhile, although most recruitment software allows users to upload work photos (such as photos of themselves in work clothes), these images are only used for personal homepage display and are not systematically analyzed or used for resume building. Image recognition technology in related fields has not yet achieved the transformation from "visual content" to "professional semantics" in recruitment scenarios; that is, it cannot automatically infer key information such as a user's job type, tools they use, and work environment from a photo.

[0027] Even if users complete the form, their resumes are often vague and unstructured, lacking detailed skill information and work environment details. Matching engines in related technologies rely solely on keyword comparison (e.g., a job title containing "electrician" is considered a match), failing to understand skill semantics or environmental suitability. This results in seemingly relevant but actually mismatched recommendations. This mismatch between coarse-grained resume information and refined job requirements is the root cause of low matching accuracy.

[0028] Furthermore, a significant amount of implicit user behavior on the platform (such as repeatedly browsing welder positions and clicking on jobs tagged with "covered") is not incorporated into resume building or matching logic. This is because the existing system processes resume generation separately from user behavior, resulting in the waste of valuable behavioral signals. This "data silo" phenomenon further weakens the system's ability to judge users' true intentions and abilities.

[0029] Therefore, the relevant technologies have significant shortcomings in terms of input methods, data fusion, automated resume generation, and matching depth. They have failed to effectively address the core pain points of blue-collar users who cannot write, are unwilling to write, or cannot write accurate resumes, and they are also unable to support the demand for high-precision job matching.

[0030] Therefore, this application provides a resume generation method to solve at least one of the above-mentioned technical problems.

[0031] Specifically, Figure 1 This is a flowchart illustrating a resume generation method provided in an embodiment of this application.

[0032] like Figure 1As shown, this resume generation method includes the following steps: In step S101, at least one of the user's voice introduction, work uniform photo, and recruitment platform behavior interaction data is obtained.

[0033] Among them, the voice introduction is a personal professional introduction entered by the user in the form of voice, such as work experience and skill characteristics; the work uniform recruitment is a work-related photo uploaded by the user that clearly identifies the professional characteristics; and the behavioral interaction data is the browsing history, application history, search history and other data generated by the user on the recruitment platform.

[0034] In step S102, the user's occupational characteristics are extracted from at least one of the voice introduction, work uniform photos, and behavioral interaction data.

[0035] Among them, occupational characteristics are information used to characterize a user's occupational abilities and work background, including job type, skills, work experience, work environment preferences, etc.

[0036] It is understood that embodiments of this application may extract a user’s professional characteristics from at least one of voice introduction, work uniform photos, and behavioral interaction data, so as to automatically generate the user’s resume in the future.

[0037] In one embodiment of this application, extracting a user's occupational characteristics from at least one of voice introduction, work uniform photo, and behavioral interaction data includes: extracting a first occupational characteristic from the voice introduction; extracting a second occupational characteristic from the work uniform photo; extracting a third occupational characteristic from the behavioral interaction data; and fusing the first, second, and third occupational characteristics to obtain the final occupational characteristics.

[0038] The first occupational feature is occupational information extracted from the user's voice introduction, including years of work experience and skill descriptions; the second occupational feature is occupational information parsed from the user's work clothes photos, such as the type of work corresponding to the work clothes, handheld tools, and working environment; the third occupational feature is occupational information mined from the user's platform behavior interaction data, such as job intentions and preferred work environments.

[0039] It is understood that the embodiments of this application can extract the user's first occupational characteristics from the voice introduction, the user's second occupational characteristics from the work clothes photo, and the user's third occupational characteristics from the behavioral interaction data. The first occupational characteristics, the second occupational characteristics, and the third occupational characteristics are then integrated to obtain the user's final occupational characteristics. This achieves the collection and processing of multi-dimensional information, avoids the problem of information bias or incompleteness from a single source, and improves the accuracy and completeness of occupational characteristics.

[0040] In one embodiment of the present application, extracting the user's first occupational feature from the voice profile includes: inputting the voice profile into a target voice recognition model, and the target voice recognition model outputs text; extracting the user's first occupational feature from the text by using rule matching and keyword matching.

[0041] Among them, the target voice recognition model can be a dialect-adaptive ASR (Automatic Speech Recognition) model; rule matching is a way of information extraction based on preset occupational-related grammar rules; keyword matching can be performed through an occupational dictionary.

[0042] It can be understood that the embodiment of the present application can convert the voice profile into text by using the target voice recognition model, and then extract the first occupational feature from the text by using rule matching and keyword matching to accurately extract the first occupational feature of the user.

[0043] Specifically, since the user may use dialect or colloquial expressions, resulting in low accuracy of general voice recognition, the embodiment of the present application can adopt a dialect-adaptive ASR model.

[0044] For example: Original input: An 8-second user voice, the content is: "I have worked as a steelworker in Zhengzhou for three years. I can bind, read drawings, and also operate a tower crane." Technical features include: Audio format: 16kHz sampling rate, 16bit PCM; Language feature: Containing the Central Plains Mandarin accent (typical pronunciation in Henan region, such as using "俺" instead of "我"). The accuracy of this model's speech recognition is increased to 95%. This model is implemented based on the following process: 1. Data preparation.

[0045] Collect a labeled dialect speech corpus covering the main blue-collar gathering areas (such as Sichuan-Chongqing, Guangdong-Guangxi, Jiangsu-Zhejiang, etc.), which mainly includes occupational-related terms (such as electrical box, argon arc welding, scaffolding).

[0046] 2. Model structure.

[0047] Based on a general end-to-end ASR model (such as Conformer or Whisper-small), introduce a dialect embedding vector as a conditional input to enable the model to dynamically adapt to different accents.

[0048] 3. Fine-tuning strategy.

[0049] Perform domain adaptation fine-tuning on the target dialect subset to improve the recognition accuracy of professional terms.

[0050] 4. Post-processing correction.

[0051] By combining the blue-collar occupational dictionary for vocabulary constraint decoding, we can avoid misidentifying "argon arc welding" as "argon arc welding" or other non-professional expressions.

[0052] 5. Deployment method.

[0053] The model is deployed on edge servers or in the cloud. After the user uploads their voice, it returns structured text in real time with a recognition accuracy of 95% (verified by internal test sets).

[0054] Furthermore, it should be noted that if the user inputs negative semantics in the speech, negative semantic detection and context-aware decoding can be used to avoid misunderstanding of negative semantics. Specifically, negative detection: in the ASR post-processing stage, a dependency parser is used to identify negative structures (such as not, never done, can't), and a negative trigger word list is constructed, such as not, none, no, etc.; context filtering: if the keyword welder appears in a negative sentence (such as I am not a welder), it is marked as no, and the skill is only retained in an affirmative context (such as I have done welding).

[0055] In one embodiment of this application, extracting a user's second occupational feature from a work uniform photo includes: inputting the work uniform photo into a target multi-task detection network, the target multi-task detection network outputting the user's occupational detection result, wherein the target multi-task detection network uses a Swing Transformer as its backbone network, and the occupational detection result includes at least one of job probability distribution, tool recognition result, environment category, and safety compliance; inputting the occupational detection result into a visual-text joint understanding model, the visual-text joint understanding model matching the occupational detection result with occupational features in a preset occupational feature description library, and outputting the user's second occupational feature.

[0056] The target multi-task detection network uses the Swing Transformer as its backbone network and integrates deep neural networks for multiple tasks such as job classification, tool detection, scene segmentation, and safety compliance judgment. The occupation detection result is the visual analysis result directly related to the occupation output by the detection network, including job probability distribution, tool recognition result, and environment category. The visual-text joint understanding model is used to semantically match the visual detection result with the occupation feature description in text form. The preset occupation feature description library is a set of text descriptions containing the tool, environment, and behavioral features corresponding to each job.

[0057] It is understood that, in the embodiments of this application, work clothes photos can be input into a target multi-task detection network to obtain occupation detection results, and then matched with a preset occupation feature description library through a visual semantic joint understanding model to output a second occupation feature, thereby achieving accurate conversion from visual information to occupation semantics. By combining target detection and semantic understanding, the accuracy of occupation feature extraction in complex scenarios can be improved.

[0058] Specifically, the embodiments of this application mainly employ a visual-text joint understanding model + multi-task detection network to achieve automatic generation of second occupational features, and the process is as follows: 1. Photo preprocessing.

[0059] Input the user-uploaded work uniform photo (supports landscape and portrait modes, low-light scenes), and perform image enhancement (gamma correction, color balance) and cropping (focusing on the subject).

[0060] 2. Multi-task deep networks.

[0061] Backbone network: Swing Transformer extracts global features; Branch tasks: Job classification head: outputs probability distribution (e.g., welder = 0.92); Tool detection head: YOLOv8 (YouOnly Look Once version 8) detects key objects such as welding torches, wrenches, and safety helmets; Environment segmentation head: Mask2Former (Masked-attention Mask Transformer 2) identifies backgrounds such as workshops, construction sites, and high-altitude areas; Safety compliance judgment: whether protective equipment is worn.

[0062] 3. Visual-linguistic alignment.

[0063] The detection results are input into a CLIP-like model (a CLIP-like model (Contrastive Language-Image Pre-training, a vision-language joint understanding model based on CLIP principles)) and semantically matched with a preset occupation description (such as "holding a TIG welding torch and wearing heat-resistant clothing"); the second occupation feature is output (such as argon arc welder (TIG)).

[0064] Construct a pre-defined rule base for occupational characteristics, with industry experts defining "visual-skill" mapping rules, such as: { "tool": "TIG welding torch", "material": "stainless steel", "action": "welding", "inferred_skill": "argon arc welding"}.

[0065] Furthermore, the embodiments of this application can support combinatorial reasoning: if “welding gun + gloves + fireproof clothing” is detected, it is inferred that “high temperature operation capability” is possessed. At the same time, each inference result is accompanied by a confidence level, which can reach 95%. If multiple rules point to the same skill, they are weighted and merged. Finally, only when the confidence level is greater than a certain threshold is the job tag field corresponding to the resume written.

[0066] It should be noted that when using the target multi-task detection network in this embodiment, it needs to be trained to ensure the accuracy of the multi-task detection network in recognizing tooling photos. Specifically, the tooling recognition problem is solved through the following multi-stage visual understanding, the process of which is as follows: 1. Building a dedicated dataset: More than 100,000 images of blue-collar work scenes were manually labeled, covering more than 50 types of jobs (welders, electricians, plumbers, etc.). Each image is labeled with: tool type, tools, environment, and safety signs.

[0067] 2. Multi-task deep networks.

[0068] Backbone network: Global features are extracted using the Swing Transformer; Branch tasks: Job classification head (Fine-grained Classification); Tool detection head (YOLOv8-based Object Detection); Scene segmentation head (Mask2Former) to identify environments such as workshops, construction sites, and high-altitude areas; Safety compliance judgment (such as whether safety helmets and goggles are worn).

[0069] Color and lighting robustness design: Introduce color-invariant features (such as Lab color space + texture descriptor); use CycleGAN (Cycle Generative Adversarial Network) for lighting / color bias enhancement training to improve generalization ability.

[0070] Tag fusion mechanism: Cross-validate visual recognition results with user voice / behavioral data (e.g., photo showing welding torch + voice mentioning welding → high confidence in skill confirmation).

[0071] The multi-task detection network of this application embodiment can achieve a job identification accuracy of 92% and a skill tag recall rate of 88% on a self-built test set.

[0072] In one embodiment of this application, extracting a user's third occupational characteristic from behavioral interaction data includes: identifying the user's behavior type and corresponding behavior content in the behavioral interaction data, wherein the behavior type includes at least one of submission, click, browsing, and searching; determining multiple occupational information tags based on the behavior type and behavior content; using a target occupational knowledge graph to perform reasoning mapping on the occupational information tags to complete the occupational information tags; obtaining the confidence score of each occupational information tag, and calculating the confidence score of the corresponding occupational information tag based on the confidence score of the occupational information tag and the weight of the corresponding behavior type; and determining the user's third occupational characteristic based on the confidence score of each occupational information tag.

[0073] Among them, behavior type refers to the type of interactive action of users on the recruitment platform, including application, click, browsing, search, etc. Different types reflect the strength of users' intentions; occupational information tags are standardized relevant identifiers extracted from the behavior content; target occupational knowledge graph is an occupational ontology library containing relationships such as job level, skill dependence, and environmental association; confidence score is a quantitative indicator that measures the degree of matching between occupational information tags and users' actual situation.

[0074] It is understood that the embodiments of this application can first identify the behavior type and content and extract occupational information tags, complete the tags through occupational knowledge graphs, avoid misjudgments caused by simple keyword matching, and then calculate the final confidence score by combining the behavior type weight and the confidence score of the tag itself. Based on the confidence score, the third occupational feature is determined, fully explore the value of the implicit behavioral data of the recruitment platform, and accurately capture the user's real job search intentions and professional abilities.

[0075] Specifically, this application embodiment mainly extracts the user's third occupation feature through multimodal behavior sequence modeling + domain knowledge graph mapping method, and the specific process is as follows: 1. Log collection and structuring.

[0076] Collect user behavior logs on the platform, including job clicks, search terms, resume submissions, and dwell time; aggregate each record within a time window (e.g., the last 30 days) to form a behavior sequence.

[0077] 2. Keyword extraction and semantic enhancement.

[0078] The BERT-CRF (Bidirectional Encoder Representations from Transformers, Conditional Random Field) model is used to identify entities such as skills, job type, and years of experience from job titles / descriptions; TF-IDF (Term Frequency-Inverse Document Frequency) + NER (Named Entity Recognition) are combined to extract user search keywords (such as welder recruitment, high-altitude work permit); a behavior-tag mapping table is constructed to transform the original behavior into standardized occupational tags (such as clicking on a welder job → intent: welder).

[0079] 3. Domain knowledge graph construction (i.e., target occupation knowledge graph).

[0080] Establish a professional ontology database for the user's occupation, including: job level (electrician → high-voltage electrician); skill dependence (argon arc welding → welding basics); and environmental correlation (high altitude → safety rope use).

[0081] Behavioral labels are mapped to structured feature vectors through graph reasoning (e.g., [welder: 0.8, high-altitude work: 0.6]).

[0082] Different weights are assigned based on the type of behavior: submission > click > search. Weighted fusion and confidence calculation are then performed to output the final third occupational feature.

[0083] It should be noted that, in the embodiments of this application, differential privacy technology is used to desensitize the original data when utilizing user behavior interaction data, while retaining key behavioral features for resume generation.

[0084] The various occupational characteristics and determination methods included in the resumes of this application embodiment are shown in Table 1. Table 1 contains only some examples.

[0085]

[0086] In addition, it should be noted that the embodiments of this application can use voice introduction, behavioral interaction data and tooling photos in pairs or all three at the same time, but the effect of using all three at the same time is better than using two.

[0087] In step S103, occupational characteristics are filled into the corresponding occupational tags, and the user's resume is generated based on the filled occupational tags.

[0088] Among them, occupational tags are occupational characteristic identifiers, such as job type, years of work experience, age, etc.

[0089] It is understood that the embodiments of this application can obtain at least one of the user's voice introduction, work uniform photo, and recruitment platform behavior interaction data, extract the user's occupational characteristics from at least one of the voice introduction, work uniform photo, and recruitment platform behavior interaction data, fill the occupational characteristics into the corresponding occupational tags, and generate the user's resume based on the filled occupational tags. This eliminates the need for the user to manually fill in and create the resume, realizes automatic resume generation, lowers the resume operation threshold, and adapts to users with different knowledge and cultural levels.

[0090] In one embodiment of this application, after generating a user's resume based on the completed occupational tags, the method further includes: identifying recruitment information for multiple job postings on a recruitment platform; using a rule engine and natural language processing to parse the recruitment information in the job postings into corresponding occupational features; calculating the matching degree between the user's occupational features and the job postings' occupational features; and if the matching degree is greater than a preset matching degree, pushing the job postings to the user.

[0091] Among them, the combination of rule engine and natural language processing is used to parse enterprise recruitment information. The rule engine extracts structured fields based on preset rules, and natural language processing is used for semantic understanding and entity recognition. The preset matching degree can be set according to specific circumstances, without specific limitations, such as setting it to 8% or 85%.

[0092] It is understood that the embodiments of this application can parse recruitment information to generate job occupational characteristics, calculate the matching degree between the user and the job occupational characteristics, and only push jobs with a matching degree exceeding a preset threshold to the user, thereby improving the accuracy of job recommendations.

[0093] The formula for calculating the matching degree in this application embodiment can be: Matching degree = Experience weight × Experience matching degree + Skill weight × Skill matching degree + Environment weight × Environment matching degree.

[0094] The default weights are 30% experience, 40% skills, and 30% environment, and dynamic calibration based on historical data is supported.

[0095] For example, the job posting requires 5+ years of experience, skills in TIG welding, and environmental workshop work; the user's resume states: 10 years of experience, skills in TIG welding, pipe welding, and environmental workshop work.

[0096] Experience matching = min(10 / 5, 1.0) = 1.0 → 30% × 1.0 = 30%; Skill matching = intersection / requirement = 1 / 1 = 1.0 → 40% × 1.0 = 40%; Environment matching = 1.0 → 30% × 1.0 = 30%; The overall match rate is 100%, and after normalization, it is displayed as 95% (to avoid being misleading by exceeding 100%).

[0097] Specifically, the resume generation method of this application includes the following steps: Step 1: Receive user input data.

[0098] It receives four types of source data: voice input; platform browsing history; platform job application history; and workplace photos of the user in work clothes.

[0099] Simultaneous data preprocessing includes speech-to-text conversion, behavioral data structuring, and photo feature extraction.

[0100] Step 2: Resume generated automatically.

[0101] Inputs: voice text, platform behavior characteristics, and photo occupational tags.

[0102] Processing logic (pseudocode illustration): If "welder" is used in the voice text: Job Title = "welder"; If "argon arc welding" appears in the photo tag ["skills"]: skill += ["argon arc welding"]; If the browsing history contains "welder position": work environment = "workshop"; Output: Structured resume (including fields such as job type, years of experience, skills list, and work environment).

[0103] Specifically, such as Figure 2 As shown, this application determines a user's professional characteristics by integrating voice, work uniform photos, and behavioral data, and automatically generates a resume based on the user's professional characteristics.

[0104] Step 3: Structure recruitment information.

[0105] Input: The original recruitment information published by the company.

[0106] Processing: The data is parsed into structured job requirement objects using NLP (Automatic Speech Recognition) and a rule engine.

[0107] Step 4: Multi-dimensional matching.

[0108] Input: Structured resume + structured job posting information.

[0109] Matching degree calculation formula: Matching degree = 0.3 × (resume experience / required experience) + 0.4 × skill overlap rate + 0.3 × environment matching (0 or 1).

[0110] The results were normalized to ensure they were within the 0–100% range.

[0111] Step 5: Work notification.

[0112] Jobs with a match rate of ≥80% will be filtered, sorted in descending order of match rate, and notified via push notification through the APP (e.g., "[95% match] Welder position at XX factory | Salary 8K-12K").

[0113] In addition, it should be noted that the embodiments of this application require training with a photo tooling feature library, but this is not necessary for every run; it is only required for initial construction and periodic updates. Differential privacy data anonymization is required as a necessary step to ensure the compliant use of user behavior data. Dynamic calibration of matching weights is required, but it is not necessary in real time; it is recommended to optimize the weight parameters monthly based on historical matching results.

[0114] This application also provides a resume generation system, the workflow of which is as follows: Figure 3 As shown, the system includes: 1. Voice input module: Receives user voice input (such as "10 years of welding experience") and outputs structured text.

[0115] 2. Platform Data Module: Collects users' browsing history and job application history on the platform, and outputs behavioral feature vectors (such as ["welder position", "pipeline welding"]).

[0116] 3. Photo Analysis Module: Receives "work photos in work clothes" uploaded by users and outputs occupational tags (such as {"job type": "welder", "skill": "argon arc welding", "environment": "workshop") through AI recognition.

[0117] 4. Resume generation module: Integrates the above three data sources to automatically generate a structured resume (JSON format).

[0118] 5. Job Posting Module: Receives job postings from companies and parses them into structured job requirements (e.g., {"Experience":"5+ years", "Skills":["Argon Arc Welding"], "Environment":"Workshop"}).

[0119] 6. Matching Engine and Job Push Module: Calculates the multi-dimensional matching degree between resumes and job postings, and pushes job notifications to users only when the matching degree is ≥80%.

[0120] The resume generation method of this application embodiment is described below through a specific example, in the scenario of user A applying for a welder position.

[0121] 1. Input data.

[0122] Voiceover: "5 years of welding experience, specializing in argon arc welding and pipe welding"; Browsing history: ["Welder position", "Pipeline welding position"]; Job application record: ["2023-11-01: Welder position"]; Photo: Wearing blue overalls, operating argon arc welding in the workshop.

[0123] 2. Processing procedure.

[0124] Successfully extracted experience and skills from speech-to-text conversion; Photo analysis output: Job type = welder, skill = argon arc welding / pipe welding, environment = workshop; Generate resume: {Job: "Welder", Experience: "5 years", Skills: ["Argon Arc Welding", "Pipe Welding"], Environment: "Workshop"}; Job posting: {Requirements: {Experience: "5+ years", Skills: ["Argon arc welding"], Environment: "Workshop"}}; Matching calculation: Experience = 5 / 5 = 1.0 (30%), Skill intersection rate = 100% (40%), Environment matching = 100% (30%) → Total matching degree 100% → Normalized display 95%.

[0125] 3. Output push.

[0126] [95% match] Welder position at XX factory | Salary 8K-12K | Workshop environment.

[0127] In summary, the resume generation method of this application addresses the common challenge faced by blue-collar users who are not proficient in typing or filling out complex forms. For the first time, it uses a "work photo in work clothes" as the core input, combined with simple voice descriptions and platform interaction history (such as browsing, application, and onboarding records) to automatically generate a structured resume. Furthermore, an AI model analyzes visual cues in the photo, such as work clothes, tools, and work environment, to accurately infer specific job type, practical skills, work scenarios, and safety awareness—key occupational context information that is difficult to express in words but crucial for job matching. This eliminates the need for manual input by the user, significantly lowering the barrier to entry. Simultaneously, it greatly enhances the recruitment system's ability to identify and match non-standardized labor, driving the upgrade of blue-collar recruitment from reliance on text-based input to a multimodal intelligent perception paradigm. Specifically, it can achieve the following technical effects: 1. Lower the barrier to resume creation by replacing manual filling with natural interaction methods such as voice, photos, and behavioral data, thus solving the problems of users being unwilling or unable to fill out resumes. 2. Leveraging the model's multimodal understanding capabilities (especially visual analysis of photos of people in work clothes), it automatically extracts structured occupational features such as job type, tools, and environment, supplementing key dimensions missing from traditional resumes; 3. By integrating explicit input (voice) and implicit behavior (browsing, clicking) to build dynamic user profiles, and using multi-dimensional semantic matching (rather than keywords) to achieve high-precision job recommendations, the efficiency of matching people to jobs and the conversion rate of user applications can be improved.

[0128] According to the resume generation method proposed in the embodiments of this application, at least one of the user's voice introduction, work uniform photo, and recruitment platform behavior interaction data can be obtained. The user's professional characteristics are extracted from at least one of the voice introduction, work uniform photo, and recruitment platform behavior interaction data. The professional characteristics are filled into the corresponding professional tags. The user's resume is generated based on the filled professional tags. The user does not need to fill in and create the resume manually, realizing automatic resume generation, reducing the resume operation threshold, and adapting to users with different knowledge and cultural levels.

[0129] Next, the resume generation apparatus according to the embodiments of this application is described with reference to the accompanying drawings.

[0130] Figure 4 This is a block diagram of a resume generation device according to an embodiment of this application.

[0131] like Figure 4 As shown, the resume generation device 10 includes: an acquisition module 100, an extraction module 200, and a generation module 300.

[0132] The acquisition module 100 is used to acquire at least one of the user's voice introduction, work uniform photo, and behavioral interaction data from the recruitment platform; the extraction module 200 is used to extract the user's professional characteristics from at least one of the voice introduction, work uniform photo, and behavioral interaction data; and the generation module 300 is used to fill the professional characteristics into the corresponding professional tags and generate the user's resume based on the filled professional tags.

[0133] In one embodiment of this application, the extraction module 200 is further configured to: extract the user's first occupational feature from the voice introduction; extract the user's second occupational feature from the work uniform photo; extract the user's third occupational feature from the behavioral interaction data; and fuse the first occupational feature, the second occupational feature, and the third occupational feature to obtain the final occupational feature.

[0134] In one embodiment of this application, the extraction module 200 is further configured to: input a speech introduction into a target speech recognition model, the target speech recognition model outputs text; and extract the user's first occupational feature from the text using rule matching and keyword matching.

[0135] In one embodiment of this application, the extraction module 200 is further configured to: input the workwear photo into a target multi-task detection network, the target multi-task detection network outputs the user's occupation detection result, wherein the target multi-task detection network uses a Swing Transformer as its backbone network, and the occupation detection result includes at least one of the following: job probability distribution, tool recognition result, environmental category, and safety compliance; input the occupation detection result into a visual-text joint understanding model, the visual-text joint understanding model matches the occupation detection result with occupation features in a preset occupation feature description library, and outputs the user's second occupation feature.

[0136] In one embodiment of this application, the extraction module 200 is further configured to: identify the user's behavior type and corresponding behavior content in the behavioral interaction data, wherein the behavior type includes at least one of submission, click, browsing, and search; determine multiple occupational information tags based on the behavior type and behavior content; perform reasoning mapping on the occupational information tags using a target occupational knowledge graph to complete the occupational information tags; obtain the confidence score of each occupational information tag, and calculate the confidence score of the corresponding occupational information tag based on the confidence score of the occupational information tag and the weight of the corresponding behavior type; and determine the user's third occupational characteristic based on the confidence score of each occupational information tag.

[0137] In one embodiment of this application, the resume generation device 10 of this application embodiment further includes: a push module.

[0138] The push module is used to identify recruitment information from multiple job postings on the recruitment platform after generating a user's resume based on the completed occupational tags; it uses a rule engine and natural language processing to parse the recruitment information in the job postings into corresponding occupational characteristics; it calculates the matching degree between the user's occupational characteristics and the job postings' occupational characteristics; if the matching degree is greater than the preset matching degree, the job postings are pushed to the user.

[0139] It should be noted that the foregoing explanation of the resume generation method embodiment also applies to the resume generation device of this embodiment, and will not be repeated here.

[0140] According to the resume generation device proposed in the embodiments of this application, at least one of the user's voice introduction, work uniform photo, and recruitment platform behavior interaction data can be obtained. The user's occupational characteristics can be extracted from at least one of the voice introduction, work uniform photo, and recruitment platform behavior interaction data. The occupational characteristics are filled into the corresponding occupational tags, and the user's resume is generated based on the filled occupational tags. The user does not need to fill in the resume manually, thus realizing automatic resume generation, reducing the resume operation threshold, and adapting to users with different knowledge and cultural levels.

[0141] Figure 5 is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. The electronic device may include: The memory 501, the processor 502, and the computer program stored on the memory 501 and capable of running on the processor 502.

[0142] When processor 502 executes the program, it implements the resume generation method provided in the above embodiments.

[0143] Furthermore, electronic devices also include: Communication interface 503 is used for communication between memory 501 and processor 502.

[0144] The memory 501 is used to store computer programs that can run on the processor 502.

[0145] Memory 501 may include high-speed RAM memory, and may also include non-volatile memory, such as at least one disk storage device.

[0146] If the memory 501, processor 502, and communication interface 503 are implemented independently, then the communication interface 503, memory 501, and processor 502 can be interconnected via a bus to complete communication between them. The bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, or an Extended Industry Standard Architecture (EISA) bus, etc. The bus can be divided into address bus, data bus, control bus, etc. For ease of representation, Figure 5 The bus is represented by a single thick line, but this does not mean that there is only one bus or one type of bus.

[0147] Optionally, in a specific implementation, if the memory 501, processor 502, and communication interface 503 are integrated on a single chip, then the memory 501, processor 502, and communication interface 503 can communicate with each other through an internal interface.

[0148] Processor 502 may be a central processing unit (CPU), an application specific integrated circuit (ASIC), or one or more integrated circuits configured to implement the embodiments of this application.

[0149] This application also provides a computer-readable storage medium storing a computer program or instructions thereon, which, when executed by a processor, implements the resume generation method described above.

[0150] This application also provides a computer program product, including a computer program or instructions, which, when executed, implement the resume generation method described above.

[0151] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of this application. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0152] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this application, "N" means at least two, such as two, three, etc., unless otherwise explicitly specified.

[0153] Any process or method described in the flowchart or otherwise herein can be understood as representing a module, segment, or portion of code comprising one or N executable instructions for implementing custom logic functions or processes, and the scope of the preferred embodiments of this application includes additional implementations in which functions may be performed not in the order shown or discussed, including substantially simultaneously or in reverse order depending on the functions involved, as should be understood by those skilled in the art to which embodiments of this application pertain.

[0154] It should be understood that the various parts of this application can be implemented using hardware, software, firmware, or a combination thereof. In the above embodiments, the N steps or methods can be implemented using software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, it can be implemented using any one or more of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (FPGAs), field-programmable gate arrays (FPGAs), etc.

[0155] Those skilled in the art will understand that all or part of the steps of the methods in the above embodiments can be implemented by a program instructing related hardware. The program can be stored in a computer-readable storage medium, and when executed, the program includes one or a combination of the steps of the method embodiments.

Claims

1. A resume generation method, characterized in that, Includes the following steps: Obtain at least one of the following: user voice profile, work uniform photo, and behavioral interaction data from the recruitment platform; Extract the user's occupational characteristics from at least one of the voice introduction, the work uniform photo, and the behavioral interaction data; The occupational characteristics are filled into the corresponding occupational tags, and the user's resume is generated based on the filled occupational tags.

2. The resume generation method according to claim 1, characterized in that, Extracting the user's occupational characteristics from at least one of the voice introduction, the work uniform photo, and the behavioral interaction data includes: Extract the user's primary occupational characteristic from the voice profile; Extract the user's secondary occupational characteristics from the workwear photo; Extract the user's third occupational characteristics from the behavioral interaction data; The first occupational characteristic, the second occupational characteristic, and the third occupational characteristic are combined to obtain the final occupational characteristic.

3. The resume generation method according to claim 2, characterized in that, The step of extracting the user's first occupational characteristic from the voice profile includes: The speech summary is input into the target speech recognition model, and the target speech recognition model outputs text. The user's primary occupational characteristics are extracted from the text using rule matching and keyword matching.

4. The resume generation method according to claim 2, characterized in that, The extraction of the user's second occupational characteristic from the workwear photo includes: The workwear photo is input into the target multi-task detection network, and the target multi-task detection network outputs the user's occupation detection result. The target multi-task detection network uses Swing Transformer as the backbone network, and the occupation detection result includes at least one of the following: job probability distribution, tool recognition result, environment category, and safety compliance. The occupation detection results are input into the visual-text joint understanding model, which matches the occupation detection results with occupation features in a preset occupation feature description library and outputs the user's second occupation feature.

5. The resume generation method according to claim 2, characterized in that, The extraction of the user's third occupational characteristic from the behavioral interaction data includes: Identify the user's behavior type and corresponding behavior content in the behavioral interaction data, wherein the behavior type includes at least one of delivery, click, browsing, and search; Multiple occupational information tags are determined based on the behavior type and the behavior content; The occupational information tags are inferred and mapped using a target occupational knowledge graph to complete the occupational information tags; Obtain the confidence score of each occupational information tag, and calculate the confidence score of the corresponding occupational information tag based on the confidence score of the occupational information tag and the weight of the corresponding behavior type; The user's third occupational characteristic is determined based on the confidence level of each occupational information tag.

6. The resume generation method according to claim 1, characterized in that, After generating the user's resume based on the completed career tags, the process also includes: Identify recruitment information for multiple job postings on the recruitment platform; The recruitment information in the job postings is parsed into corresponding occupational characteristics using a rule engine and natural language processing. Calculate the degree of matching between the user's occupational characteristics and the occupational characteristics of the job posting; If the matching degree is greater than the preset matching degree, the job posting will be pushed to the user.

7. A resume generation device, characterized in that, include: The acquisition module is used to acquire at least one of the user's voice introduction, work uniform photo, and behavioral interaction data from the recruitment platform. The extraction module is used to extract the user's occupational characteristics from at least one of the voice introduction, the work uniform photo, and the behavioral interaction data; The generation module is used to fill the occupational characteristics into the corresponding occupational tags and generate the user's resume based on the filled occupational tags.

8. An electronic device, characterized in that, include: A memory, a processor, and a computer program stored in the memory and executable on the processor, the processor executing the program to implement the resume generation method as described in any one of claims 1-6.

9. A computer-readable storage medium having a computer program or instructions stored thereon, characterized in that, The computer program or instructions are executed by a processor to implement the resume generation method as described in any one of claims 1-6.

10. A computer program product, comprising a computer program or instructions, characterized in that, When the computer program or instructions are executed, they implement the resume generation method as described in any one of claims 1-6.

Images