Intelligent recommendation system for employment service knowledge graph

Abstract

The invention discloses an employment service knowledge graph intelligent recommendation system, and relates to the technical field of employment service and artificial intelligence, and the system comprises a multi-source data fusion module which is used for integrating a resume library, a recruitment website, an industry report and social media data; the dynamic knowledge graph construction module comprises an ontology modeling sub-module and a real-time updating sub-module; the user portrait engine module supports multi-dimensional feature extraction and dynamic weight adjustment; the intelligent recommendation engine module adopts a mixed recommendation algorithm to combine knowledge graph reasoning and collaborative filtering; and the visual interaction module is used for providing a force-oriented graph to display the position relation network. The method has the advantages that real-time updating of graph node attributes is achieved through the incremental learning algorithm, and the problem of data lag of a traditional system is solved. For example, when an enterprise newly adds a'meta universe architecturist 'post, the system automatically identifies skills (Unity3D, 3D modeling) required by the post and updates the atlas.

Description

Technical Field

[0001] This invention relates to the fields of employment services and artificial intelligence technology, and in particular to an intelligent recommendation system based on an employment service knowledge graph. Background Technology

[0002] Employment services are a series of services provided by specific institutions to meet the needs of job seekers and employers. These services mainly include four categories: job placement, employment training, unemployment insurance, and labor and employment service companies. Job placement encompasses job registration, employment guidance, and information dissemination. Unemployment insurance, enforced by law, guarantees basic living standards, promotes re-employment, and prevents unemployment; it includes policy implementation, job recommendations, and employment assistance, and promotes new service models such as digital empowerment.

[0003] Traditional systems rely on a single data source, leading to biased recommendation results. Most systems use offline model training, which cannot respond to market changes in real time. New users or emerging professional fields lack historical behavioral data, resulting in a significant drop in recommendation accuracy. Recommendation systems based on keyword matching cannot handle synonyms or polysemous words. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention provides an intelligent recommendation system based on an employment service knowledge graph, which solves the problems mentioned in the background section.

[0005] To achieve the above objectives, the present invention adopts the following technical solution:

[0006] An intelligent recommendation system based on a knowledge graph for employment services includes:

[0007] A multi-source data fusion module is used to integrate resume databases, recruitment websites, industry reports, and social media data;

[0008] The dynamic knowledge graph construction module includes an ontology modeling submodule and a real-time update submodule.

[0009] The user profile engine module supports multi-dimensional feature extraction and dynamic weight adjustment.

[0010] The intelligent recommendation engine module employs a hybrid recommendation algorithm that combines knowledge graph reasoning with collaborative filtering.

[0011] The visual interactive module provides a force-guided graph to display the job relationship network.

[0012] Preferably, the multi-source data fusion module integrates heterogeneous data from resume databases, recruitment websites, industry reports, and social media, and performs data cleaning, deduplication, and standardization. The dynamic knowledge graph construction module includes an ontology modeling submodule and a real-time update submodule, responsible for building and maintaining a temporally sequential domain knowledge graph. The user profile engine module supports the extraction of multi-dimensional features from user data and has the ability to dynamically adjust feature weights. The intelligent recommendation engine module adopts a hybrid recommendation algorithm, combining knowledge graph reasoning and collaborative filtering technology to generate personalized job recommendations. The visualization and interaction module provides a force-directed graph-based visualization interface to display the relationship network between entities such as jobs and skills, and supports interactive exploration.

[0013] Furthermore, the real-time update submodule of the dynamic knowledge graph construction module adopts an incremental learning algorithm to capture new job data and update the graph node attributes every 24 hours.

[0014] Preferably, this submodule automatically collects new job information from recruitment websites, company postings, and other channels through scheduled tasks. It uses incremental learning technology to process only new or changed data, avoiding full reconstruction and ensuring the timeliness of the map and system efficiency.

[0015] Furthermore, the multi-dimensional feature extraction submodule of the user profile engine module extracts hard and soft skills from the resume text and generates a user profile by combining behavioral data.

[0016] Preferably, hard skills include quantifiable abilities such as programming languages ​​and tool usage; soft skills include textual descriptions such as communication skills and teamwork; and behavioral data covers interaction logs such as job browsing and application records, which together constitute a three-dimensional user profile.

[0017] Furthermore, the hybrid recommendation algorithm of the intelligent recommendation engine module generates a recommendation list by weighted fusion of knowledge graph reasoning results and collaborative filtering results.

[0018] Preferably, knowledge graph reasoning discovers potential job associations based on entity relationship paths; collaborative filtering calculates recommendation weights based on the similarity of user group behavior; the two are fused through an adaptive weighting strategy to balance interpretability and recommendation novelty.

[0019] Furthermore, the visualization and interaction module allows users to explore the skill evolution path by clicking on nodes, showcasing the development of technology.

[0020] Preferably, users can select specific skill nodes through the interactive interface to view their associated job trends, related skill evolution paths, and changes in market demand, thus assisting in career planning decisions.

[0021] Furthermore, the ontology modeling submodule of the dynamic knowledge graph construction module defines five types of entities—job, skill, enterprise, industry, and user—and twelve types of relationships.

[0022] The preferred 12 relationships include "job-required skills", "user-possessed skills", "company's industry", and "job-affiliated company", forming a multi-dimensional semantic network covering "person-job-company-industry".

[0023] Furthermore, the dynamic weight adjustment submodule of the user profiling engine module adopts a reinforcement learning model to optimize feature weights in real time based on user feedback.

[0024] Preferably, the system uses reinforcement learning to dynamically adjust feature weights based on user feedback behaviors such as clicking, saving, and submitting recommendations, thereby achieving adaptive updates to user profiles and preferences.

[0025] Furthermore, the intelligent recommendation engine module employs a "skill-job" reverse reasoning strategy for cold-start users, recommending frequently associated jobs by analyzing the user's resume skills.

[0026] Preferably, for newly registered users or users with sparse behavioral data, the system extracts a skill set from their resume, queries the knowledge graph for jobs that are frequently associated with that skill set in the past, and generates an initial recommendation list.

[0027] Compared with existing technologies, the advantages of this invention are:

[0028] 1. Real-time updates of graph node attributes are achieved through incremental learning algorithms, solving the data lag problem of traditional systems. For example, when a company adds a "Metaverse Architect" position, the system automatically identifies the required skills (Unity3D, 3D modeling) for the position and updates the graph, defining 5 types of entities (position, skill, company, industry, user) and 12 types of relationships ("position-required skills", "company-industry"). New position data is crawled every 24 hours, and incremental learning algorithms are used to update graph node attributes.

[0029] 2: Integrate textual data (resumes), behavioral data (browsing history), and social data (LinkedIn endorsements) to build a 360-degree user profile. Extract hard skills "Python" and soft skills "teamwork" from resume text, and combine behavioral data such as job browsing time to generate user profiles. Use a reinforcement learning model to optimize feature weights in real time based on user feedback of clicking / ignoring recommended jobs.

[0030] 3. The GraphSAGE algorithm is used to embed representation learning of the knowledge graph, capturing the complex relationships between nodes. It combines knowledge graph reasoning (job recommendation based on skill inheritance path) and collaborative filtering (recommendation based on user similarity) to generate the final recommendation list through weighted fusion. For new users, a "skill-job" inverse reasoning strategy is adopted, and a force-directed graph is provided to display the job relationship network, allowing users to explore the skill evolution path by clicking on nodes. Attached Figure Description

[0031] Fig. 1 This is a system architecture diagram of the present invention;

[0032] Fig. 2 This is a flowchart of the dynamic knowledge graph update process of this invention. Detailed Implementation

[0033] Example 1

[0034] Reference Figs. 1-2 A knowledge graph-based intelligent recommendation system for employment services;

[0035] Dynamic knowledge graph updates;

[0036] Data scraping;

[0037] Technical details:

[0038] Configure a scheduled task (CronJob) using the Scrapy framework to trigger a full crawl of the "New Jobs" page on Lagou.com every 24 hours. The crawled fields include job name, job description, salary range, skill requirements, company name, industry classification, etc. To avoid anti-crawling mechanisms, IP polling and User-Agent randomization strategies are adopted.

[0039] Formula definition: Let the amount of data to be crawled be... Success rate of crawling The effective data volume .

[0040] Node matching;

[0041] A pre-trained BERT-Based Chinese model was used to perform entity recognition on job descriptions, extracting entities such as job title, skills, and company. Cosine similarity was used to calculate the semantic similarity between the extracted entities and existing nodes in the knowledge graph, with a matching threshold set to [value missing]. .

[0042] Formula definition: Similarity calculation formula:

[0043]

[0044] in , For entities , BERT embedding vector.

[0045] Attribute update;

[0046] Technical details:

[0047] Match successful: Update node attributes (e.g., salary range is the weighted average of the scraped data and the original data, with a weight ratio of 1). In the experiment ).

[0048] Unmatched: Create a new node and establish relationships through graph reasoning (e.g., connect the new job title "Metaverse Architect" with the skill nodes "Unity3D" and "3D Modeling").

[0049] Formula Definition: Salary Update Formula:

[0050]

[0051] Effect verification;

[0052] Experimental design: Test on a dataset of 100,000 entries and compare the accuracy with manually labeled results.

[0053] Quantification formula:

[0054] accuracy , where TP = number of correctly updated nodes and FP = number of incorrectly updated nodes.

[0055] Example 2

[0056] Reference Figs. 1-2 A knowledge graph-based intelligent recommendation system for employment services;

[0057] Recommended for users starting from cold;

[0058] Skill extraction;

[0059] Technical details: Hard skills (such as "deep learning" and "Python") are extracted from resume text using Jieba word segmentation and TF-IDF algorithm, and invalid words are filtered out by combining a pre-trained skill dictionary. Soft skills (such as "teamwork") are extracted using sentiment analysis model (such as SnowNLP).

[0060] Formula Definition: Skill Weight Calculation Formula:

[0061]

[0062] in For emotional enhancement coefficient, .

[0063] Reverse reasoning;

[0064] Technical details:

[0065] Query the knowledge graph for positions that are frequently associated with the extracted skills; the strength of the association is determined by the edge weights in the graph. The edge weights are calculated by normalizing the job-skill co-occurrence frequency.

[0066] Formula definition: Job recommendation score:

[0067]

[0068] in For user skill sets, For skills With position Edge weights.

[0069] Recommended generation;

[0070] Technical details:

[0071] The job scores are then re-ranked based on users' job search intentions (expected salary, work location). A multi-objective optimization model is used with the following weights: skill matching degree (60%), salary satisfaction (30%), and geographical location preference (10%).

[0072] Formula definition: Overall score:

[0073]

[0074] Effect verification;

[0075] Experimental Design:

[0076] Tested on 500 new users, comparing the click-through rate (CTR) of recommendations with traditional collaborative filtering (CF) algorithms.

[0077] Quantification formula:

[0078] Comparing CTR with CF.

[0079]

[0080] Example 3

[0081] Reference Figs. 1-2 A knowledge graph-based intelligent recommendation system for employment services;

[0082] Visual interactive exploration;

[0083] Node click;

[0084] Technical details: After a user clicks on the "Computer Vision" skill node, the system dynamically displays related nodes (such as "Image Processing" and "Object Detection") through the D3.js force-guided graph. The node spacing is calculated by the force simulation algorithm.

[0085]

[0086] in The spring constant is... The distance between nodes. This is the ideal distance.

[0087] Path display;

[0088] Technical details:

[0089] The shortest path from "computer vision" to other skills is calculated using Dijkstra's algorithm, and key technological breakthroughs are identified.

[0090] Formula definition: Path weight:

[0091]

[0092] in The edge weight (technical influence score).

[0093] Job recommendations;

[0094] Technical details:

[0095] Based on the user's exploration path (e.g., "computer vision → object detection"), the system recommends job positions associated with the skills at the end of the path (e.g., "autonomous driving perception algorithm engineer"). The recommendation logic is the same as in Example 2.

[0096] Quantitative Results: Average user exploration time decreased from 5 minutes to 3 minutes, as shown in the formula. Job understanding depth (based on questionnaire rating) increased by 50%.

[0097] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the claimed invention.

Claims

1. An intelligent recommendation system based on a knowledge graph for employment services, characterized in that, include: A multi-source data fusion module is used to integrate resume databases, recruitment websites, industry reports, and social media data; The dynamic knowledge graph construction module includes an ontology modeling submodule and a real-time update submodule. The user profile engine module supports multi-dimensional feature extraction and dynamic weight adjustment. The intelligent recommendation engine module employs a hybrid recommendation algorithm that combines knowledge graph reasoning with collaborative filtering. The visual interactive module provides a force-guided graph to display the job relationship network.

2. The intelligent recommendation system for employment services based on a knowledge graph according to claim 1, characterized in that, The real-time update submodule of the dynamic knowledge graph construction module adopts an incremental learning algorithm to capture new job data and update the graph node attributes every 24 hours.

3. The intelligent recommendation system for employment services based on a knowledge graph according to claim 1, characterized in that, The multi-dimensional feature extraction submodule of the user profile engine module extracts hard and soft skills from resume text and combines them with behavioral data to generate user profiles.

4. The intelligent recommendation system for employment services based on a knowledge graph according to claim 1, characterized in that, The intelligent recommendation engine module's hybrid recommendation algorithm generates a recommendation list by weightedly fusing knowledge graph reasoning results with collaborative filtering results.

5. The intelligent recommendation system for employment services based on a knowledge graph according to claim 1, characterized in that, The visualization and interaction module allows users to explore the skill evolution path by clicking on nodes, showcasing the development of technology.

6. The intelligent recommendation system for employment services based on a knowledge graph according to claim 1, characterized in that, The ontology modeling submodule of the dynamic knowledge graph construction module defines five types of entities—job, skill, enterprise, industry, and user—and twelve types of relationships.

7. The intelligent recommendation system for employment services based on a knowledge graph according to claim 1, characterized in that, The dynamic weight adjustment submodule of the user profiling engine module adopts a reinforcement learning model to optimize feature weights in real time based on user feedback.

8. The intelligent recommendation system for employment services based on a knowledge graph according to claim 1, characterized in that, The intelligent recommendation engine module employs a "skill-job" reverse reasoning strategy for users starting out, recommending frequently associated jobs by analyzing the skills in the user's resume.

Images