Technical talent evaluation method and system based on technical relevancy

Abstract

The invention discloses a technical talent evaluation method and system based on technical relevancy, and the method comprises the steps: extracting keywords from technical description provided by an enterprise through a natural language processing technology; patent text matching is carried out in a patent database according to the extracted keywords, and patents highly related to enterprise requirements are screened out; based on a patent matching result, evaluating the technical contribution degree of the inventor in matching patents; by analyzing the migration condition of the inventor in the related technical field, the technical activity and the field correlation are determined; in combination with occupational stability analysis of the inventor, whether the inventor meets the technical talent demand of the enterprise is evaluated; and generating a matching degree score of the technical talents according to the evaluation results.

Description

Background Art

[0002] The patent system protects the creative work of inventors by granting patent rights and prevents others from imitating and infringing. The patent system encourages innovation and invention, and promotes scientific and technological progress by exchanging exclusive market advantages for the disclosure and sharing of technical information. Patented technology can be sold or transferred as a commodity, forming intangible assets with economic value. At the same time, patents have good publicity effects and can improve product quality and corporate reputation.

[0003] By searching and analyzing the number and quality of patents, we can reveal hot spots and trends in technological innovation and help accurately match the talents we need. However, patent searches are currently conducted based on keywords, but the information obtained is relatively scattered, limited and inconvenient to summarize, resulting in a low accuracy rate. The discovery and evaluation of technical talents is complex and difficult. Summary of the invention

[0004] In view of the deficiencies in the prior art, the present invention provides a technical talent evaluation method and system based on technology relevance, which aims to match and evaluate relevant technical talents through the patent database according to the technical description of the R&D project provided by the enterprise, so as to help the enterprise accurately discover talents that are highly matched with its technical needs. The method combines natural language processing, patent analysis, dynamic technology evaluation and machine learning optimization to provide a multi-dimensional technical talent evaluation model.

[0005] In order to achieve the above object, the present invention adopts the following technical solutions: A technical talent evaluation method based on technical relevance, the method comprising the following steps: Extract keywords from the technical descriptions provided by enterprises through natural language processing technology; Match patent texts in the patent database based on the extracted keywords to screen out patents that are highly relevant to enterprise needs; Based on the patent matching results, evaluate the inventor's technical contribution in the matching patents; By analyzing the migration of inventors in related technical fields, their technological activity and field relevance can be determined; Combined with the career stability analysis of the inventor, assess whether he meets the technical talent needs of the enterprise; Based on the above evaluation results, a matching score for technical talents is generated.

[0006] Preferably, the keyword extraction step uses a TF-IDF, BERT or LDA model to extract and vectorize keywords, specifically including the following process: The TF-IDF algorithm is used to calculate the weight of keywords in enterprise technology descriptions and patent texts. The formula is:

[0007] Among them, TF(t,d) represents the frequency of occurrence of term t in document d, DF(t) represents the number of documents containing term t, and N is the total number of documents.

[0008] The BERT model is used to semantically vectorize the text and extract context-related embedding vectors through a deep learning model.

[0009] Preferably, the patent text matching step calculates the similarity between the enterprise technology description and the patent text through a vector space model, using the following similarity calculation formula:

[0010] Among them, Ai is the TF-IDF value of the i-th keyword in the enterprise technology description, Bi is the TF-IDF value of the i-th keyword in the patent text, and Sim(A,B) is the cosine similarity between the two.

[0011] Preferably, the inventor contribution evaluation is based on the inventor's ranking in the patent and the patent publication time, and is calculated using a dynamic contribution formula:

[0012] Among them, Position represents the ranking of inventors in the patent. The higher the ranking, the greater the contribution. λ is the time decay coefficient. Tcurrent is the current time, and Tpatent is the patent publication time.

[0013] Preferably, the technology field migration analysis is evaluated by a time-weighted technology relevance formula, specifically:

[0014] Among them, Simpast is the similarity between the inventor and the current technology demand of the enterprise in a certain period of time in the past, and W is the time weight. The closer to the current time, the higher the weight.

[0015] Preferably, the career stability analysis is based on the inventor's job-hopping frequency and is calculated using the following career stability scoring formula:

[0016] Among them, β is the job-hopping frequency sensitivity coefficient. Hopping Frequency represents the inventor's job-hopping frequency. The more job-hopping times, the lower the Job Stability Score value.

[0017] Preferably, the comprehensive scoring model is calculated using the following formula: Overall Score=f(α·Technical Expertise,β·Contribution(t),γ·Tech_Migration(t),·δKeyword_Match(t),ϵ·Job Stability) Among them, Overall Score is the final comprehensive score; f() is a comprehensive scoring function that can be tuned through a machine learning model; α·Technical Expertise is the weighted value of the technical expertise score, and α is the weight of the technical expertise score dimension; β·Contribution(t) is the weighted value of the dynamic contribution, and β is the weight of the contribution dimension; γ·Tech_Migration(t) is the weighted value of the technology migration factor, and γ is the weight of the technology migration factor dimension; δ·Keyword_Match(t) is the weighted value of keyword matching, and δ is the weight of the keyword matching dimension; ϵ·Job Stability is the weighted value of job stability, and ϵ is the weight of the job stability dimension.

[0018] Preferably, the comprehensive scoring model is optimized by a machine learning model, trained using random forest, XGBoost or support vector machine, and the weights of each dimension are adjusted to improve the matching degree and evaluation accuracy.

[0019] A technical talent evaluation system based on technical relevance, characterized in that the system comprises: Enterprise demand input module, used to receive the enterprise's technical description and key technical requirements; Data preprocessing module, used to extract patent data from the patent database and perform keyword extraction and patent matching; Technical talent evaluation module, which is used to evaluate inventors’ technical contributions, technical field activity, and career stability; The comprehensive scoring module is used to generate matching scores for technical talents.

[0020] Preferably, the system is personalized and tuned through machine learning models based on the special needs of the enterprise to provide customized talent recommendation solutions.

[0021] Compared with the prior art, the present invention has the following beneficial effects: The present application provides a technical talent evaluation method and system based on technology relevance. Based on multi-dimensional evaluation, the system can comprehensively analyze the inventor's technical capabilities and stability to ensure the accuracy of the match.

[0022] This application provides machine learning optimization, which can be continuously optimized based on enterprise feedback and success cases to continuously improve the matching and evaluation results.

[0023] This application only requires the enterprise to enter a project description, and the system can automatically recommend technical talents that are highly matched with the project requirements, greatly improving the discovery efficiency. DETAILED DESCRIPTION

[0024] The following description is used to disclose the present invention so that those skilled in the art can implement the present invention. The preferred embodiments described below are only examples, and those skilled in the art can think of other obvious variations. The basic principles of the present invention defined in the following description can be used for other embodiments, variations, improvements, equivalents, and other technical solutions that do not deviate from the spirit and scope of the present invention.

[0025] The present application provides a technical talent evaluation method based on technical relevance, the method comprising the following steps: Step 1: Extract keywords from the technical description provided by the enterprise through natural language processing technology; Step 2: Match patent texts in the patent database based on the extracted keywords to screen out patents that are highly relevant to the company's needs; Step 3: Based on the patent matching results, evaluate the inventor’s technical contribution in the matching patents; Step 4: Determine the inventor’s technological activity and field relevance by analyzing the inventor’s migration in related technical fields; Step 5: Combined with the career stability analysis of the inventor, assess whether he meets the technical talent needs of the enterprise; Step 6: Based on the above evaluation results, generate a matching score for technical talents.

[0026] In step one, the enterprise technology description is first made according to the project requirements provided by the enterprise, including keywords, technical field description, target technical capabilities and other information, and then keywords are extracted from the technical description provided by the enterprise through natural language processing technology.

[0027] The keyword extraction step uses the TF-IDF, BERT or LDA model to extract and vectorize keywords, which specifically includes the following processes: The TF-IDF algorithm is used to calculate the weight of keywords in enterprise technology descriptions and patent texts. The formula is:

[0028] Where TF(t,d) represents the frequency of occurrence of term t in document d, DF(t) represents the number of documents containing term t, and N is the total number of documents; The BERT model is used to semantically vectorize the text and extract context-related embedding vectors through a deep learning model.

[0029] In step 2, the patent database includes: inventor information, technical field classification (IPC / CPC), patent text and application time, etc.

[0030] Technical field classification: Use the classification labels (IPC / CPC) in the patent database to mark the inventor's patent technical field and construct technical capability labels.

[0031] Inventor information: extract the inventor, applicant (company), patent publication time, inventor ranking, etc. from the patent information.

[0032] Patent texts are matched in the patent database based on the extracted keywords to screen out patents that are highly relevant to corporate needs.

[0033] Among them, keyword matching: the similarity between the technical description provided by the enterprise and the patent text is calculated through the vector space model (VSM). The similarity formula is:

[0034]

[0035] Among them, Sim(A,B) represents the similarity between the enterprise technology description and the patent text; A is the keyword vector extracted from the enterprise technology description; B is the keyword vector extracted from the patent text; Ai is the TF-IDF value of the i-th keyword in the enterprise technology description; Bi is the TF-IDF value of the i-th keyword in the patent text; ∑Ai² is the sum of squares of keyword vectors in enterprise technology descriptions; ∑Bi² is the sum of squares of keyword vectors in patent texts.

[0036] The higher the similarity of a patent, the more its corresponding inventor is considered a matching candidate.

[0037] In step 3, the inventors are ranked according to their contribution in each patent and their technical contributions are evaluated using the dynamic contribution formula:

[0038] Among them, Contribution(t) represents the contribution of the inventor in the patent.

[0039] Position refers to the ranking of inventors in the patent. The higher the ranking, the greater the contribution.

[0040] λ is the time attenuation coefficient, which controls the influence of time on the contribution. The larger the value, the faster the attenuation.

[0041] Tcurrent is the current time.

[0042] Tpatent is the patent publication time.

[0043] It is the time decay factor, the longer the time, the lower the contribution.

[0044] This formula combines the inventor's contribution with time decay. The more recent the contribution, the greater the impact on the matching degree. Based on the patent matching results, the inventor's technical contribution in the matching patents is evaluated.

[0045] In step 4, by analyzing the patent technology fields of the inventors in different time periods, we evaluate whether they are still active in the technology fields that the enterprise needs. This analysis uses time-weighted technology relevance:

[0046]

[0047] Tech_Migration(t) indicates the migration of inventors in different technological fields at different times.

[0048] Simpast is the similarity between the inventor and the current technology needs of the enterprise at a certain time in the past.

[0049] Wt is the time weight, and usually the closer to the current time the higher the weight.

[0050] ∑Wt is the sum of the time weights of all time periods.

[0051] This formula is used to assess whether the inventor is still active in the technology field required by the business.

[0052] If the inventor made a high contribution in the early technology field but has recently moved to other fields, the matching degree will decrease accordingly. Therefore, by analyzing the migration of inventors in related technology fields, their technology activity and field relevance can be determined.

[0053] In step 5, the frequency of job hopping of inventors is calculated by the company changes in patent applications. Frequent job hopping may indicate unstable technology focus. Its stability is evaluated by the following formula:

[0054]

[0055] Job Stability Score indicates the career stability score of the inventor.

[0056] β is the job-hopping frequency sensitivity coefficient, which controls the impact of job-hopping frequency on the score. The larger the value, the more significant the impact of job-hopping.

[0057] Hopping Frequency is the inventor's job-hopping frequency. The more times the inventor changes jobs, the higher the value.

[0058] This formula reflects the stability of an inventor’s career trajectory; frequent job hopping will reduce his or her stability score.

[0059] A higher job-hopping frequency will reduce the career stability score and affect the final technical talent evaluation. Combined with the career stability analysis of the inventor, assess whether he meets the technical talent needs of the enterprise.

[0060] In step six, multiple dimensions are introduced in the comprehensive evaluation process, such as technology field matching, inventor contribution, technology field migration trend, career trajectory stability, etc. The comprehensive scoring formula is as follows:

[0061] Overall Score=f(α·Technical Expertise,β·Contribution(t),γ·Tech_Migration(t),·δKeyword_Match(t),ϵ·Job Stability) Overall Score is the final comprehensive score.

[0062] f() is a comprehensive scoring function that can be tuned through a machine learning model.

[0063] α·Technical Expertise is the weighted value of the technical expertise score, and α is the weight of the technical expertise score dimension; β·Contribution(t) is the weighted value of the dynamic contribution, and β is the weight of the contribution dimension; γ·Tech_Migration(t) is the weighted value of the technology migration factor, and γ is the weight of the technology migration factor dimension; δ·Keyword_Match(t) is the weighted value of keyword matching, and δ is the weight of the keyword matching dimension; ϵ·Job Stability is the weighted value of job stability, and ϵ is the weight of the job stability dimension.

[0064] This formula uses a combination of multiple dimensions to generate scores and ultimately arrives at the degree of match between the inventor and the company’s technological needs.

[0065] Further optimized using machine learning models, the system can provide personalized recommendations based on the specific needs of the enterprise. For example:

[0066] Enterprises focus on inventors who make long-term technological contributions, and the model can be tuned through historical data to increase the importance of long-term stability.

[0067] The present invention provides a technical talent evaluation system based on technical relevance, the system comprising: Enterprise demand input module; Data preprocessing module; Technical talent evaluation module; Comprehensive scoring module.

[0068] In this system, enterprises need to input modules: Function: Users input the technical description of the company's R&D project through this module, which mainly includes project background, target technology field, key technical requirements and other information.

[0069] Input: Technical keywords: for example, "artificial intelligence algorithm", "chip design", etc.

[0070] Classification of relevant technical fields: for example, “IPC / CPC technical classification number”.

[0071] Key technology description: detailed technical background and requirements description.

[0072] Time requirement: requirements on the time of activity of relevant technology, such as “inventor within the past 5 years”.

[0073] In this system, the data preprocessing module: Function: Obtain relevant information of technical talents from the patent database, and clean, process and extract features of patent data.

[0074] Main tasks: Keyword extraction: Use TF-IDF, BERT and other technologies to extract key technical terms from enterprise demand descriptions and patent texts and generate vocabulary vectors.

[0075] Patent data analysis: Extract and match relevant technical data based on patent text, IPC / CPC classification number, and inventor information.

[0076] Time feature processing: extract time information from the patent release time to determine the activity and recency of the technology.

[0077] This system also includes a technical talent matching module: Function: According to enterprise needs, filter out matching inventors and related information from the patent database.

[0078] Main steps: Keyword matching: Use keyword vectors to calculate the similarity between the enterprise technology description and the patent text (see keyword matching formula) to screen out patents and inventors with higher similarity.

[0079] Technical field matching: Based on IPC / CPC classification numbers, match the number and activity of patents of inventors in the same technical field.

[0080] Dynamic contribution assessment: Calculate the inventor’s contribution to the patent, and use the dynamic contribution formula to adjust the time factor so that the most recent contribution is weighted more heavily than the earlier contributions.

[0081] Technology field migration analysis: The technology migration factor formula is used to evaluate whether the inventor remains active in the technology field required by the enterprise. If the inventor was active in the early stage but has moved to other fields, the matching degree will be reduced.

[0082] In this system, the technical talent evaluation module: Function: Comprehensively evaluate the matched inventors and generate scores and reports.

[0083] Evaluation dimensions: Technical expertise score: based on the degree of match between the technical keywords in the inventor’s patent and the needs of the enterprise (keyword matching formula).

[0084] Contribution score: Based on the inventor’s contribution to each patent, weighted calculation is performed in combination with the time factor (dynamic contribution formula).

[0085] Technology field migration score: assesses whether there is obvious migration or transfer of the inventor’s technological activities in the required technology field (technology migration factor formula).

[0086] Career stability score: Calculate the stability of an inventor’s career trajectory by looking at job-hopping frequency and changes in patent applicants (career stability score formula).

[0087] Comprehensive scoring: The above dimensions are weighted and integrated using a comprehensive scoring formula to generate the inventor’s final comprehensive matching score. Companies can screen talents based on the score.

[0088] Based on the special needs of enterprises, this system also provides customized talent recommendation solutions through personalized tuning of machine learning models. The machine learning optimization model: Function: Continuously optimize the weights and evaluation models of each dimension to improve the accuracy and adaptability of the system.

[0089] Optimize the process: Data collection: The system extracts inventor matching results from the company's historical recruitment data and success cases and annotates them as training data.

[0090] Feature extraction: Train machine learning models based on features in historical data (such as technology fields, inventor contributions, career trajectories, etc.).

[0091] Model training: Use supervised learning models (such as random forest, XGBoost, SVM, etc.) to optimize the various dimensions of the evaluation model. The model will automatically adjust the weights of each dimension to find the optimal scoring method.

[0092] Personalized tuning: Enterprises can perform personalized tuning based on their special needs. For example, for some projects, technical stability is more important, and the system can increase the weight of this dimension through tuning.

[0093] Model update: The system will regularly retrain and update the model based on new data to ensure the accuracy of the evaluation results.

[0094] Functional flow of this system Enterprises input technical requirements: Enterprises provide project technical descriptions and key requirements through the input module.

[0095] Patent data preprocessing: The system obtains patent data of relevant inventors from the patent database and performs cleaning and feature extraction.

[0096] Inventor matching: The system matches inventors with a high degree of similarity to the company's technical needs based on technical keywords, field classifications and patent information.

[0097] Technical talent evaluation: The system conducts a comprehensive evaluation of inventors from multiple dimensions such as technical expertise, contribution, technology migration trend, and career stability, and generates a score.

[0098] Machine learning optimization: The system continuously optimizes the scoring model through historical data and successful cases, and adjusts the weights of evaluation dimensions.

[0099] Output report: The system provides enterprises with inventors’ technology matching reports to help them select the best technical talents.

[0100] Output Technical Talent Matching Report: includes multi-dimensional analysis of each inventor’s technical expertise, contribution score, career trajectory, technology migration trend, etc. Enterprises can screen talents based on the comprehensive score.

[0101] Talent recommendation list: candidate technical talents arranged in order of score, with their technical fields, matching degree and other information marked.

[0102] Advantages of this system: Accurate matching: Based on multi-dimensional evaluation, the system can comprehensively analyze the inventor's technical capabilities and stability to ensure the accuracy of the match.

[0103] Machine learning optimization: The system can be continuously optimized based on enterprise feedback and success cases to continuously improve the matching and evaluation effects.

[0104] Automated recommendation: Enterprises only need to enter the project description, and the system can automatically recommend technical talents that are highly matched with project requirements, greatly improving discovery efficiency.

[0105] Those skilled in the art should understand that the embodiments of the present invention shown in the above description are only examples and do not limit the present invention. The purpose of the present invention has been fully and effectively achieved. The functions and structural principles of the present invention have been demonstrated and explained in the embodiments, and the embodiments of the present invention may be deformed or modified in any way without departing from the principles.

Claims

1. A technical talent evaluation method based on technical relevance, characterized by: The method comprises the following steps: Extract keywords from the technical descriptions provided by enterprises through natural language processing technology; Match patent texts in the patent database based on the extracted keywords to screen out patents that are highly relevant to enterprise needs; Based on the patent matching results, evaluate the inventor's technical contribution in the matching patents; By analyzing the migration of inventors in related technical fields, their technological activity and field relevance can be determined; Combined with the career stability analysis of the inventor, assess whether he meets the technical talent needs of the enterprise; Based on the above evaluation results, a matching score for technical talents is generated.

2. A technical talent evaluation method based on technical relevance according to claim 1, characterized in that: The keyword extraction step uses the TF-IDF, BERT or LDA model to extract and vectorize keywords, specifically including the following processes: The TF-IDF algorithm is used to calculate the weight of keywords in enterprise technology descriptions and patent texts. The formula is: ； Where TF(t,d) represents the frequency of occurrence of term t in document d, DF(t) represents the number of documents containing term t, and N is the total number of documents; The BERT model is used to semantically vectorize the text and extract context-related embedding vectors through a deep learning model.

3. The technical talent evaluation method based on technical relevance according to claim 1 is characterized by: The patent text matching step calculates the similarity between the enterprise technology description and the patent text through a vector space model, using the following similarity calculation formula: ； Among them, Ai is the TF-IDF value of the i-th keyword in the enterprise technology description, Bi is the TF-IDF value of the i-th keyword in the patent text, and Sim(A,B) is the cosine similarity between the two.

4. The technical talent evaluation method based on technical relevance according to claim 1 is characterized by: The inventor contribution evaluation is based on the inventor's ranking in the patent and the patent publication time, and is calculated using the dynamic contribution formula: ； Among them, Position represents the ranking of inventors in the patent. The higher the ranking, the greater the contribution. λ is the time decay coefficient. Tcurrent is the current time, and Tpatent is the patent publication time.

5. The technical talent evaluation method based on technical relevance according to claim 1 is characterized by: The technology area migration analysis is evaluated by the time-weighted technology relevance formula, specifically: ； Among them, Simpast is the similarity between the inventor and the current technology demand of the enterprise in a certain period of time in the past, and W is the time weight. The closer to the current time, the higher the weight.

6. The technical talent evaluation method based on technical relevance according to claim 1 is characterized by: The career stability analysis is based on the inventor's job-hopping frequency and is calculated using the following career stability scoring formula: ； Among them, β is the job-hopping frequency sensitivity coefficient. Hopping Frequency represents the inventor's job-hopping frequency. The more job-hopping times, the lower the Job Stability Score value.

7. The technical talent evaluation method based on technical relevance according to claim 1 is characterized by: The comprehensive scoring model is calculated using the following formula: Overall Score=f(α·Technical Expertise,β·Contribution(t),γ·Tech_Migration(t),·δKeyword_Match(t),ϵ·Job Stability) Among them, Overall Score is the final comprehensive score; f() is a comprehensive scoring function that can be tuned through a machine learning model; α·Technical Expertise is the weighted value of the technical expertise score, and α is the weight of the technical expertise score dimension; β·Contribution(t) is the weighted value of the dynamic contribution, and β is the weight of the contribution dimension; γ·Tech_Migration(t) is the weighted value of the technology migration factor, and γ is the weight of the technology migration factor dimension; δ·Keyword_Match(t) is the weighted value of keyword matching, and δ is the weight of the keyword matching dimension; ϵ·Job Stability is the weighted value of job stability, and ϵ is the weight of the job stability dimension.

8. The technical talent evaluation method based on technical relevance according to claim 1 is characterized by: The comprehensive scoring model is optimized through a machine learning model, trained using random forest, XGBoost or support vector machine, and the weights of each dimension are adjusted to improve the matching degree and evaluation accuracy.

9. A system based on the technical talent evaluation method according to any one of claims 1 to 8, characterized in that: The system includes: Enterprise demand input module, used to receive the enterprise's technical description and key technical requirements; Data preprocessing module, used to extract patent data from the patent database and perform keyword extraction and patent matching; Technical talent evaluation module, which is used to evaluate inventors’ technical contributions, technical field activity, and career stability; The comprehensive scoring module is used to generate matching scores for technical talents.

10. The system according to claim 9, characterized in that: Based on the special needs of the enterprise, the system performs personalized tuning through machine learning models to provide customized talent recommendation solutions.

Images