A qualification certificate intelligent screening method for communication project bidding in the construction industry

Abstract

The application relates to the technical field of building communication engineering, in particular to a qualification certificate intelligent screening method for bidding of a communication project in the building industry, which comprises the following steps: obtaining project information text of a target communication project in the building industry; extracting ability information of each qualification certificate object conforming to a required qualification certificate type from a qualification certificate database; qualification certificate objects of each qualification certificate type correspond to employees of a post role; constructing a three-dimensional matching graph according to post roles of a proposed team of the target communication project and the ability information of the qualification certificate objects; based on the three-dimensional matching graph, a pre-constructed multi-channel neural semantic matching network is adopted to perform semantic matching scoring on the qualification certificate objects and the project information text of the target communication project, so that a semantic matching score of each qualification certificate object is obtained; and the optimal team combination and a final qualification certificate set are determined according to the post roles of the proposed team of the target communication project and the semantic matching score of each qualification certificate object.

Description

Technical Field

[0001] This invention relates to the field of building communication engineering technology, and in particular to an intelligent screening method for qualification certificates used in bidding for communication projects in the construction industry. Background Technology

[0002] In the bidding process of the building communications industry, contractors typically need to assemble a qualified team based on project requirements and submit relevant qualification certificates to pass the bidding qualification review and scoring stages. During this process, factors such as the scale, structure, construction period, geographical environment, and technical limitations of different projects impose varying requirements on the qualifications, experience, and regional adaptability of the personnel.

[0003] Existing qualification certificate screening methods primarily rely on human experience and manual searches. Evaluation personnel or project managers must repeatedly compare job requirements with the certificate holder's capabilities across multiple qualification certificate databases, including their major, professional title, validity period, and past project experience. This process is not only inefficient but also prone to inaccurate matching due to subjective judgment or incomplete information, impacting bid quality and success rate.

[0004] In recent years, although some enterprises have begun to introduce simple information systems to manage qualification certificate information, most systems are still limited to static filtering conditions, such as keyword filtering based on qualification certificate type and certificate status. They lack semantic understanding and intelligent matching capabilities between job responsibilities and the qualifications of certificate holders, making it difficult to meet the comprehensive assessment needs of job fit and team collaboration in complex construction industry communication projects. Furthermore, current technologies generally fail to effectively utilize environmental information from past projects (such as geographical region, climate conditions, and construction type) to help determine a certificate holder's suitability for the current project, and also lack analysis of historical collaboration relationships and time-based coordination among team members. Summary of the Invention

[0005] In view of the above-mentioned shortcomings and deficiencies of the prior art, the present invention provides a method for intelligent screening of registration certificates for bidding in the construction industry.

[0006] To achieve the above objectives, the main technical solutions adopted by the present invention include:

[0007] This invention provides a method for intelligent screening of qualification certificates for bidding on telecommunications projects in the construction industry, comprising the following steps:

[0008] S1. Obtain the project information text of the target communication project in the construction industry;

[0009] The project information text includes: a description of the communication project, total cost, construction period requirements, geographical and climatic characteristics, construction technology limitations, required qualification certificates, and the job roles of the proposed team.

[0010] S2. Extract the capability information of each qualification object that matches the required qualification type from the qualification database;

[0011] Each type of qualification certificate corresponds to a specific job role for the practitioner.

[0012] The competency information for each certified individual includes: professional category of the certification, professional title level, current certification status, conflict information of ongoing communication engineering projects, certification validity period, historical communication project experience, types of construction techniques mastered, special operation permit qualifications, experience in the service area, and environmental information for each past communication project participated in.

[0013] S3. Construct a three-dimensional matching map based on the job roles and qualification information of the proposed team members for the target communication project.

[0014] S4. Based on the three-dimensional matching graph, a pre-constructed multi-channel neural semantic matching network is used to perform semantic matching scoring on the project information text of the qualification certificate object and the target communication project, and obtain the semantic matching score of each qualification certificate object.

[0015] S5. Based on the job roles of the proposed team for the target communication project and the semantic matching score of each certification object, determine the optimal team combination and the final set of certifications.

[0016] Preferably, S3 specifically includes:

[0017] S31. Based on the job descriptions published in the National Vocational Qualification Standards Database, extract the job task keywords and ability requirements for each role in the proposed team for the target communication project, and construct the job description vector corresponding to each role in the proposed team for the target communication project.

[0018] The National Vocational Qualification Standards Database is a collection of publicly available vocational qualification standards information resources released by the Ministry of Human Resources and Social Security or industry regulatory authorities. This database includes texts of job descriptions for all positions in the field of telecommunications construction.

[0019] S32. Extract the capability information of each qualification object that matches the required qualification type from the qualification database, and construct a qualification capability vector;

[0020] The qualification certificate capability vector is a multi-dimensional vector representation generated by structuring the capability information of the qualification certificate object;

[0021] S33. Natural language processing technology is used to semantically parse the job description vector and qualification certificate ability vector, and map them into a unified vector space to form a three-dimensional matching map with a three-dimensional structure.

[0022] The three-dimensional structure includes: job title dimension, job responsibility description vector dimension, and qualification certificate ability vector dimension.

[0023] Preferably, S31 specifically includes:

[0024] S311. Obtain the list of proposed job roles in the target communication project;

[0025] S312. Extract the text of the job description corresponding to each job role from the national vocational qualification standard database;

[0026] S313. Perform natural language preprocessing on the text of the job description to extract the job task keywords and ability requirements corresponding to the job role.

[0027] S314. Convert the extracted keywords and ability requirements into vector representations of a unified dimension, and construct a job responsibility description vector for each role in the proposed team for the target communication project.

[0028] Preferably, S33 specifically includes:

[0029] S331. Based on the preset unified semantic vector space model, the job description vector and the qualification certificate ability vector are mapped to semantic vector representations respectively.

[0030] S332. Based on three dimensions—job title, job description, and qualification / ability—construct a three-dimensional structural matching graph to represent the semantic matching relationship between each job responsibility and qualification / ability.

[0031] Preferably, the pre-constructed multi-channel neural semantic matching network includes:

[0032] The task intent channel is used to process the job description vector using an attention mechanism network and output a task matching score.

[0033] The competency feature channel is used to perform a non-linear transformation on the qualification competency vector using a deep feedforward neural network and output a competency matching score.

[0034] The environmental experience channel is used to input the environmental information of the target project and the historical project environmental information of the certification object into a graph neural network, extract environmental features and output an environmental matching score.

[0035] The project environmental information comes from the fields of geographical and climatic characteristics and construction technology limitations in the project information text;

[0036] The historical project environmental information of the qualification certificate holder comes from the environmental information of its past participation in communication projects, including geographical area, construction conditions, and construction type;

[0037] The three channels output task matching score, ability matching score and environment matching score respectively. They are input into the weighted fusion module to calculate the semantic matching score and output the semantic matching score between the job role and the qualification certificate object.

[0038] Preferably, the weighted fusion module uses formula (1) to calculate the matching score and obtain the semantic matching score between the job role and the qualification certificate object; the formula (1) is:

[0039] Score(i,j)=α·Ftask(i,j)+β·Fability(i,j)+γ·Fenv(i,j);

[0040] Score(i,j) represents the semantic matching score between the i-th job role and the j-th qualification certificate object;

[0041] Ftask(i,j) represents the task matching score between the i-th job role and the j-th qualification object output by the task intent channel;

[0042] Fability(i,j) represents the ability matching score between the i-th job role and the j-th qualification certificate object output by the ability feature channel;

[0043] Fenv(i,j) represents the environmental experience matching score between the i-th job role and the j-th qualification object, output by the environmental experience channel;

[0044] α is the first weighting coefficient; β is the second weighting coefficient; γ is the third weighting coefficient; where 1 = α + β + γ.

[0045] Preferably, the task matching score is the semantic similarity value between the job description vector and the qualification certificate ability vector, which is measured by cosine similarity.

[0046] The competency matching score is a weighted score based on the specific requirements of the job role for the type of qualification certificate, professional title level, and mastery of construction technology, which assesses the degree of matching between the qualification certificate holder's competency vector and the corresponding fields.

[0047] The environment matching score is calculated by embedding the historical communication project environment information of the qualification certificate object and the environmental information of the target communication project into a graph structure using a graph neural network, and then calculating the similarity value of the embedding vectors between the two.

[0048] Preferably, S5 specifically includes:

[0049] S51. For each job role in the proposed team, select M qualification certificate objects in descending order of semantic matching score with the job role, combine them into a candidate set of personnel for the job role, and use the semantic matching score between the M qualification certificate objects and the job role as the job suitability score respectively.

[0050] For each job role candidate set, divide all objects in that job role candidate set into groups;

[0051] The number of people in each target group is the number of qualified candidates required for the specific role in the proposed team.

[0052] Randomly select one group from the candidate set of personnel for each position and combine them to obtain a complete team candidate combination set;

[0053] S52. Obtain the comprehensive score for each complete team candidate combination set using a preset comprehensive evaluation function;

[0054] The comprehensive evaluation score is formed by weighting the following indicators: individual job suitability score of the candidate team, team schedule coordination index, multi-job conflict avoidance coefficient, and team historical cooperation tacit understanding.

[0055] S53. Based on the comprehensive score of each complete team candidate combination set, determine the optimal team combination and the final set of qualifications;

[0056] The optimal team combination is the set of complete team candidate combinations with the highest comprehensive score; the final certificate set includes the certificate corresponding to each certificate object in the optimal team combination.

[0057] Preferably, the comprehensive evaluation function is:

[0058]

[0059] MatchScore k The job suitability score for the kth qualified candidate in the complete team candidate combination set;

[0060] XT represents the team duration coordination index within the complete set of candidate team combinations;

[0061] GB is the multi-position conflict avoidance coefficient in the complete team candidate combination set;

[0062] LS represents the historical teamwork and synergy among the complete set of candidate team combinations.

[0063] δ is the preset first coefficient; ε is the preset second coefficient; θ is the preset third coefficient; ω is the preset fourth coefficient.

[0064] Preferably, the method for calculating the team schedule coordination index includes:

[0065] For each certified object in the complete team candidate combination set, obtain the estimated end time and reserved buffer period of its current ongoing communication project to determine its available entry time window; map the available entry time window of each certified object to a Gaussian time distribution function; calculate the overlap integral of the entry time distribution functions among all certified objects to obtain the degree of collaboration overlap of each certified object pair; take the average of the degree of collaboration overlap of all certified object pairs as the team's schedule collaboration index;

[0066] The method for calculating the conflict avoidance coefficient includes:

[0067] Check whether each certified individual in the complete team candidate combination set has already held multiple roles in other ongoing communication projects. If so, mark it as a conflict. Check whether each certified individual in the complete team candidate combination set holds multiple positions in the target communication project, and whether these positions are mutually exclusive. If so, mark it as a conflict. Divide the actual number of conflicts by the theoretical maximum number of conflicts and take the inverse proportional form to obtain the conflict avoidance coefficient, with a value range of 0 to 1.

[0068] The calculation method for the degree of cooperation includes:

[0069] Extract the joint participation records of any two certified objects in historical communication projects from the complete team candidate combination set; count the number of historical collaborations and the number of successful collaborations between each certified object pair; calculate the collaboration success rate of each certified object pair, and calculate the weighted average based on the number of historical collaborations to obtain the collaboration synergy of the entire candidate team.

[0070] The beneficial effects of this invention are:

[0071] This invention provides an intelligent screening method for qualification certificates in bidding for communication projects in the construction industry. By employing a technical approach that includes structured extraction of project information text, construction of capability vectors for qualification certificate objects, three-dimensional matching graph modeling, and a multi-channel neural semantic matching network scoring mechanism, this method, compared to existing technologies, can accurately assess the matching relationship between job responsibilities and qualification certificate objects from multiple dimensions based on the specific needs of the communication project. It automatically generates highly matched personnel combination schemes, achieving the beneficial effects of improving personnel selection efficiency, increasing team matching accuracy, and enhancing the intelligent level of personnel allocation.

[0072] Furthermore, by constructing job description vectors using job descriptions from the National Vocational Qualification Standards Database and combining them with natural language processing technology to semantically analyze job tasks and competency information, compared to traditional methods based on keyword matching or static filtering rules, it can more accurately understand the semantic relationship between the actual job requirements and the competency of the qualification certificate holder. This achieves the beneficial effects of enhancing the intelligent analysis capability of job-person matching and reducing human judgment errors.

[0073] Furthermore, by constructing a multi-channel neural semantic matching network and embedding task intent features, ability features, and environmental experience features into different channels for semantic modeling, compared with existing single-indicator evaluation methods, it can evaluate the job competence, process experience matching, and environmental adaptability of qualification certificate candidates from multiple dimensions, achieving the beneficial effect of improving the multi-source credibility of scoring results and the business adaptability of matching results.

[0074] Furthermore, in the process of optimizing team composition, by introducing comprehensive evaluation indicators such as job suitability score, schedule coordination index, multi-job conflict avoidance coefficient and team historical cooperation tacit understanding, compared with the traditional manual combination or random recommendation method, it can systematically evaluate the overall stability and construction coordination of the team, thus achieving the beneficial effect of improving the rationality of team composition. Attached Figure Description

[0075] Figure 1 This is a flowchart of a qualification certificate intelligent screening method for bidding on communication projects in the construction industry, according to the present invention. Detailed Implementation

[0076] To better explain and facilitate understanding of the present invention, the present invention will be described in detail below with reference to the accompanying drawings and specific embodiments.

[0077] To better understand the above technical solutions, exemplary embodiments of the present invention will be described in more detail below with reference to the accompanying drawings. Although exemplary embodiments of the present invention are shown in the drawings, it should be understood that the present invention can be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that the present invention can be understood more clearly and thoroughly, and that the scope of the present invention can be fully conveyed to those skilled in the art.

[0078] Example 1

[0079] See Figure 1 This embodiment provides a method for intelligent screening of qualification certificates for bidding on telecommunications projects in the construction industry, including the following steps:

[0080] S1. Obtain the project information text of the target communication project in the construction industry;

[0081] The project information text includes: a description of the communication project, total cost, construction period requirements, geographical and climatic characteristics, construction technology limitations, required qualification certificates, and the job roles of the proposed team.

[0082] For example, the communication project description in the project information text of this embodiment is as follows: For example, the description of a specific intelligent building control system communication project; total cost: approximately 200 million yuan; construction period: 18 months; geographical and climatic characteristics: hot and humid climate in the southern coastal area; construction technology limitations: for example, communication distance limitations (limited bus length, requiring repeaters if exceeding the limit); wiring environment limitations (difficult wiring in high humidity and high temperature environments); required qualification certificate types: Level 1 Construction Engineer Certificate (Mechanical and Electrical Engineering), Registered Electrical Engineer Certificate, Low Voltage Engineer Certificate, Safety Officer Filing Certificate, Quality Inspector Filing Certificate, etc.; the proposed team's job roles consist of 5 categories of roles: Level 1 Construction Engineer, Registered Electrical Engineer, Low Voltage Engineer, Safety Officer, and Quality Inspector.

[0083] S2. Extract the capability information of each qualification object that matches the required qualification type from the qualification database;

[0084] Each type of qualification certificate corresponds to a specific job role for the practitioner.

[0085] The competency information for each certificate holder includes: the professional category of the certificate (e.g., telecommunications engineering), the professional title level (e.g., intermediate, advanced, or first-level, second-level), the current status of the certificate (whether it is valid), conflict information of ongoing telecommunications projects (whether there are time conflicts), the validity period of the certificate (remaining validity period), historical telecommunications project experience (the type and scale of telecommunications projects in which the certificate holder participated), the types of construction techniques mastered, special operation permits, experience in the service area (whether the certificate holder has similar project experience in the southern region), and environmental information for each telecommunications project in which the certificate holder participated (e.g., geographical location, construction conditions, climate characteristics, etc.).

[0086] In this embodiment, the certificate holder is the person who holds the certificate.

[0087] In this embodiment, the qualification database pre-stores the competency information of multiple qualification holders for each qualification type in the construction industry's communications field. For example, assuming there are five qualification types in the construction industry's communications field, the qualification database stores competency information for multiple selectable qualification holders for each qualification type. For instance, for the qualification type of Registered Electrical Engineer, the database stores competency information for seven individuals holding Registered Electrical Engineer qualification certificates.

[0088] S3. Based on the job roles and qualification information of the proposed team members for the target communication project, construct a three-dimensional matching map; in this embodiment, S3 specifically includes:

[0089] S31. Based on the publicly available job descriptions in the National Vocational Qualification Standards Database (or a pre-defined database), extract the job task keywords and competency requirements for each role in the proposed team for the target communication project, and construct a job description vector corresponding to each role in the proposed team for the target communication project. The National Vocational Qualification Standards Database is a collection of publicly available vocational qualification standards information resources released by the Ministry of Human Resources and Social Security or industry regulatory authorities. This National Vocational Qualification Standards Database (or a pre-defined database) includes the text of job descriptions for all roles in the field of communication construction. Specifically, S31 includes: S311. Obtaining a list of proposed roles in the target communication project; assuming these include: Level 1 Construction Engineer, Registered Electrical Engineer, Low-Voltage Engineer, Safety Officer, Quality Inspector, etc. S312. Extracting the text of the job description corresponding to each role from the National Vocational Qualification Standards Database; for example, a Level 1 Construction Engineer (Mechanical and Electrical Engineering) is positioned as a project manager or project leader, mainly responsible for the overall organization and management of the project. Main responsibilities: Fully responsible for the construction organization and implementation of the project, ensuring that indicators such as schedule, quality, safety, and cost are met. Organize and review project construction organization design plans. Responsible for communication and resource coordination with owners, supervisors, subcontractors, and other parties. Coordinate work among multiple disciplines including civil engineering, electrical engineering, communications, low-voltage systems, and equipment installation. Ensure construction complies with specifications and standards, and implement safety production responsibilities. Liaise with owners and relevant units, and participate in project acceptance and handover upon completion. Registered Electrical Engineer (Power Supply / Distribution / Transmission / Transformation) serves as an expert in the design and review of electrical systems, participating in system planning, design calculations, and technical oversight. Main responsibilities: Responsible for the design of building power supply and distribution systems, equipment power systems, emergency power supplies, lighting, grounding, etc. Review construction drawings and design drawings to confirm compliance with national standards and actual usage requirements. Perform power load calculations, cable selection, and distribution cabinet selection. Responsible for designing electrical interface linkage schemes with intelligent building systems, communication systems, fire protection systems, etc. Provide technical consultation and support for project construction, commissioning, and acceptance phases. If the project includes substations or UPS / EPS computer rooms, their review and confirmation are mandatory. S313. Perform natural language preprocessing on the job description text to extract the job task keywords and competency requirements corresponding to the job role; S314. Convert the extracted keywords and competency requirements into a unified vector representation to construct a job responsibility description vector for each job role in the proposed team for the target communication project. (Use natural language preprocessing techniques (such as word segmentation, keyword extraction, syntactic analysis, etc.) to extract job task keywords (such as "construction coordination" and "on-site guidance") and competency requirements. Then, convert these keywords and competency requirements into a unified vector representation (such as BERT embedding vectors) to form a job responsibility description vector for each job role).

[0090] S32. Extract the capability information of each qualification object that matches the required qualification type from the qualification database, and construct a qualification capability vector; the qualification capability vector is a multi-dimensional vector representation generated by structuring the capability information of the qualification object;

[0091] For example, Electrical Engineer A, a senior engineer, is familiar with 10kV power distribution systems and emergency power supply design, has 7 years of experience in public building projects, has participated in the design of electrical systems for multiple large hospitals and airports, is skilled in UPS and EPS system configuration optimization, and possesses excellent cross-disciplinary coordination abilities. Low-voltage engineer B holds an Information Systems Project Management Certificate, is proficient in integrated cabling, security monitoring, intelligent lighting, and building automation system integration, has 5 years of experience in intelligent building and rail transit projects, and has led the implementation of smart park systems in multiple hot and humid climate zones. These information fields, after structured processing, are embedded into a unified semantic vector, forming a capability vector representation of the corresponding qualification object.

[0092] S33. Natural language processing technology is used to semantically parse the job description vector and qualification certificate ability vector, mapping them to a unified vector space to form a three-dimensional matching graph with a three-dimensional structure. The three-dimensional structure includes: a job title dimension, a job description vector dimension, and a qualification certificate ability vector dimension. Specifically, S33 includes: S331. Based on a pre-defined unified semantic vector space model, the job description vector and qualification certificate ability vector are mapped to semantic vector representations.

[0093] S332. Based on three dimensions—job title, job description, and qualification / ability—construct a three-dimensional structural matching graph to represent the semantic matching relationship between each job responsibility and qualification / ability.

[0094] Taking the job title "Electrical Engineer" as an example, the job responsibility vector is semantically mapped and organized with the ability vectors of multiple candidate qualification objects such as Electrical Engineer A and Electrical Engineer B. Based on the following three dimensions: job title name dimension (e.g., "Electrical Engineer"), job responsibility description vector dimension, and qualification ability vector dimension, a structured three-dimensional matching graph is constructed for subsequent semantic matching score calculation.

[0095] In this embodiment, during the construction of the three-dimensional matching graph between job roles and qualification certificate objects, the method of automatically extracting job descriptions from the national vocational qualification standard database and semantically vectorizing job tasks and ability requirements using natural language processing technology provides a more accurate representation of the job's responsibilities and ability characteristics compared to traditional methods that rely on manual reading of job requirements or static field filtering. Furthermore, by simultaneously mapping the job description vector and the qualification certificate ability vector to a unified semantic vector space, and constructing a semantic matching graph according to three dimensions—job role name, job responsibilities, and qualification certificate ability—it comprehensively expresses the multi-faceted and multi-layered semantic relationship between jobs and individuals, achieving the beneficial effects of improving the accuracy of job suitability judgment and enhancing the intelligence of matching. In addition, the three-dimensional matching graph serves as the basic input structure for subsequent neural network model scoring, helping to improve the quality of contextual modeling in semantic matching, reducing the risk of shallow keyword mismatches, and thus achieving high-quality and efficient intelligent filtering and recommendation.

[0096] S4. Based on the three-dimensional matching graph, a pre-constructed multi-channel neural semantic matching network is used to perform semantic matching scoring on the project information text of the qualification certificate object and the target communication project, and obtain the semantic matching score of each qualification certificate object.

[0097] In this embodiment, the pre-constructed multi-channel neural semantic matching network includes:

[0098] The task intent channel is used to process the job description vector using an attention mechanism network and output a task matching score.

[0099] For example, the job description vector contains semantic information. An attention mechanism based on a Transformer structure (such as Self-Attention) is used to encode this job description vector, extracting the semantic components representing the core work tasks to generate a task preference feature vector. The qualification certificate candidate's ability vector contains similar semantic representations. The attention-weighted similarity between the two is calculated, outputting the qualification certificate candidate's task matching score for the job description, for example, 0.89.

[0100] The competency feature channel is used to perform a non-linear transformation on the qualification competency vector using a deep feedforward neural network and output a competency matching score.

[0101] For example, the ability vector of qualification certificate holder A is input into a multilayer feedforward neural network (such as an MLP). The network automatically performs non-linear transformations on dimensions such as "professional title level = senior" and "qualification certificate status = valid" to extract ability feature representations. The target ability requirements for the job role include tags such as "having experience in communication projects". The network aligns its mapping results with the job target requirements at the vector level and outputs an ability matching score, for example, 0.93.

[0102] The environmental experience channel is used to input the environmental information of the target project and the historical project environmental information of the certification object into a graph neural network, extract environmental features and output an environmental matching score.

[0103] For example, extracting project environmental information from project text includes: geographical and climatic characteristics: hot and humid coastal areas in the south; fields related to construction technology limitations, etc.

[0104] The historical project environment information of qualification certificate holder A includes: having participated in communication projects in 3 southern cities; and being familiar with damp working conditions.

[0105] The “target project environment” and “historical project environment” are represented as environmental feature nodes, and modeled with a graph structure. After embedding with a graph neural network (such as GraphSAGE), the graph vector similarity between the two is calculated, and the environment matching score is output, for example, 0.85.

[0106] The project environmental information comes from the fields of geographical and climatic characteristics and construction technology limitations in the project information text;

[0107] The historical project environment information of the qualification certificate holder comes from the environmental information of the projects they have participated in in the past, including geographical area, communication type, etc.

[0108] The three channels output task matching score, ability matching score and environment matching score respectively. They are input into the weighted fusion module to calculate the semantic matching score and output the semantic matching score between the job role and the qualification certificate object.

[0109] The weighted fusion module uses formula (1) to calculate the matching score and obtain the semantic matching score between the job role and the qualification certificate object; the formula (1) is:

[0110] Score(i,j)=α·Ftask(i,j)+β·Fability(i,j)+γ·Fenv(i,j);

[0111] Score(i,j) represents the semantic matching score between the i-th job role and the j-th qualification object; Ftask(i,j) represents the task matching score between the i-th job role and the j-th qualification object, output by the task intent channel; Fability(i,j) represents the ability matching score between the i-th job role and the j-th qualification object, output by the ability feature channel; Fenv(i,j) represents the environmental experience matching score between the i-th job role and the j-th qualification object, output by the environmental experience channel; α is the first weight coefficient; β is the second weight coefficient; γ is the third weight coefficient; where 1 = α + β + γ.

[0112] The three channels output the following scores respectively: Task matching score: 0.89; Ability matching score: 0.93; Environment matching score: 0.85. The input to the weighted fusion module calculates the comprehensive semantic matching score according to preset weights (e.g., α = 0.4, β = 0.4, γ = 0.2): 0.4 × 0.89 + 0.4 × 0.93 + 0.2 × 0.85 = 0.896. That is, the comprehensive semantic matching score between this qualification certificate holder and the "Electrical Engineer" position is 0.896.

[0113] The task matching score is the semantic similarity between the job description vector and the qualification certificate ability vector, measured by cosine similarity. The ability matching score is a weighted score of the matching degree of the qualification certificate ability vector of the qualification certificate object on the corresponding fields according to the specific requirements of the job role for qualification certificate category, professional title level, and mastery of construction technology. The environment matching score is the similarity value between the embedded vectors of the qualification certificate object and the target communication project by embedding the historical project environment information of the qualification certificate object and the environmental information of the target communication project through graph neural network.

[0114] In this embodiment, a comprehensive semantic matching score is formed by weighted fusion of multi-channel outputs, making the matching results more in line with the comprehensive consideration of engineering practice, thereby improving the intelligence level and decision-making efficiency of the team's screening.

[0115] S5. Based on the job roles of the proposed team for the target communication project and the semantic matching score of each certification object, determine the optimal team combination and the final set of certifications.

[0116] In this embodiment, S5 specifically includes:

[0117] S51. For each job role in the proposed team, select M qualification certificate objects in descending order of semantic matching score with the job role, combine them into a candidate set of personnel for the job role, and use the semantic matching score between the M qualification certificate objects and the job role as the job suitability score respectively.

[0118] For each job role candidate set, divide all objects in that job role candidate set into groups;

[0119] The number of people in each target group is the number of qualified candidates required for the specific role in the proposed team.

[0120] Randomly select one group from the candidate set of personnel for each position and combine them to obtain a complete team candidate combination set;

[0121] In other words, within the candidate set for each job role, the candidates are divided into multiple groups based on the number of people required for the position; one combination is selected from each group corresponding to each job role to form a complete team candidate set. Assume the target project requires 5 job roles: Level 1 Construction Engineer (1 person), Registered Electrical Engineer (1 person), Low-Voltage Engineer (1 person), Safety Officer (2 people), and Quality Inspector (1 person). The qualification certificate intelligent screening method, based on the matching score results of step S4, extracts the top M (e.g., M=5) qualification certificate objects with the highest semantic matching scores for each job role as candidates, forming the following job candidate set:

[0122] The top 5 qualified certificate holders (numbered) by matching score for Level 1 Construction Engineer are: A1, A2, A3, A4, A5; the top 5 qualified certificate holders (numbered) by matching score for Registered Electrical Engineer are: B1, B2, B3, B4, B5; the top 5 qualified certificate holders (numbered) by matching score for Low-Voltage Electrical Engineer are: C1, C2, C3, C4, C5; the top 5 qualified certificate holders (numbered) by matching score for Safety Officer (requires 2 people) are: D1, D2, D3, D4, D5; the top 5 qualified certificate holders (numbered) by matching score for Quality Inspector are: E1, E2, E3, E4, E5.

[0123] Then, the individuals in these candidate sets are combined:

[0124] For the safety officer position (requires 2 people), enumerate all unique pairs from the 5 people to form 10 groups; for each other position (requires 1 person), each person is considered as a group, for a total of 5 groups;

[0125] Randomly select one member from the target group for each position (or use exhaustive search or heuristic algorithms), and merge the selected groups for all positions to form a complete team combination.

[0126] For example, it includes the following members:

[0127] First-class construction engineer: A3; Registered electrical engineer: B1; Low-voltage electrical engineer: C4; Safety officer: D2, D5; Quality inspector: E1; This combination is one of the items in a "complete team candidate combination set".

[0128] S52. Obtain the comprehensive score for each complete team candidate combination set using a preset comprehensive evaluation function;

[0129] The comprehensive evaluation score is formed by weighting the following indicators: individual job suitability score of the candidate team, team schedule coordination index, multi-job conflict avoidance coefficient, and team historical cooperation tacit understanding.

[0130] The comprehensive evaluation function is:

[0131]

[0132] MatchScore k This is the job suitability score for the k-th certified candidate in the complete team candidate set. For example, suppose the complete team candidate set is: Level 1 Construction Engineer: A3; Registered Electrical Engineer: B1; Low Voltage Engineer: C4; Safety Officer: D2, D5; Quality Inspector: E1; and the job suitability scores of all members in the combination are averaged. For example: A3 = 0.91, B1 = 0.87, C4 = 0.90, D2 = 0.82, D5 = 0.84, E1 = 0.86.

[0133] Average job fit score = (0.91 + 0.87 + 0.90 + 0.82 + 0.84 + 0.86) / 6 = 0.866;

[0134] XT represents the team duration coordination index within the complete set of candidate team combinations;

[0135] GB is the multi-position conflict avoidance coefficient in the complete team candidate combination set;

[0136] LS represents the historical teamwork and synergy among the complete set of candidate team combinations.

[0137] δ is a preset first coefficient (in this embodiment, δ = 0.4); ε is a preset second coefficient (in this embodiment, ε = 0.2); θ is a preset third coefficient (in this embodiment, θ = 0.2); ω is a preset fourth coefficient (in this embodiment, ω = 0.2).

[0138] The calculation method for the team schedule coordination index includes:

[0139] For each certified object in the complete team candidate combination set, obtain the estimated end time and reserved buffer period of its current ongoing communication project to determine its available entry time window; map the available entry time window of each certified object to a Gaussian time distribution function; calculate the overlap integral of the entry time distribution functions among all certified objects to obtain the degree of collaboration overlap of each certified object pair; take the average of the degree of collaboration overlap of all certified object pairs as the team's schedule collaboration index;

[0140] For example, a candidate team consists of 5 qualified individuals. The estimated completion time and reserved buffer period of each member's current ongoing communication projects are extracted from the database to obtain their available entry time windows. These time windows are then converted to a Gaussian distribution, and the degree of overlap between the time windows of all members is calculated. A higher degree of overlap indicates that the members are more likely to enter the site simultaneously. Finally, the average overlap between all member pairs is taken as the project schedule coordination index for this team; a higher value indicates a more coordinated team entry.

[0141] The method for calculating the conflict avoidance coefficient includes:

[0142] Check whether each certified individual in the complete team candidate set has already assumed multiple roles in other ongoing communication projects; if so, mark it as a conflict. Also check whether each certified individual in the complete team candidate set holds multiple mutually exclusive roles in the target communication project; if so, mark it as a conflict. Calculate the conflict avoidance coefficient by dividing the actual number of conflicts by the theoretical maximum number of conflicts, taking the inverse proportionality. The value range is 0 to 1. Taking a candidate team as an example, check whether any members have already assumed multiple roles in other ongoing construction projects, or have been assigned multiple mutually exclusive roles in the target communication project (e.g., simultaneously serving as safety officer and construction worker). If a conflict exists, mark it as one item, and calculate the conflict avoidance coefficient based on the actual number of conflicts and the theoretical maximum number of conflicts. A higher coefficient indicates a more reasonable allocation of responsibilities among team members and fewer conflicts.

[0143] The calculation method for the degree of cooperation includes:

[0144] Extract the joint participation records of any two certified objects in historical communication projects from the complete team candidate combination set; count the number of historical collaborations and the number of successful collaborations between each certified object pair; calculate the collaboration success rate of each certified object pair, and calculate the weighted average based on the number of historical collaborations to obtain the collaboration synergy of the entire candidate team.

[0145] For example, the system can analyze whether team members have previously participated in the same project and whether they successfully completed it. For instance, if two members have collaborated multiple times with a high success rate, it indicates good teamwork. The system then calculates and weights the collaboration records of all members to arrive at the overall teamwork score; a higher score indicates a better foundation for team collaboration.

[0146] S53. Based on the comprehensive score of each complete team candidate combination set, determine the optimal team combination and the final set of qualifications;

[0147] The optimal team combination is the set of complete team candidate combinations with the highest comprehensive score; the final certificate set includes the certificate corresponding to each certificate object in the optimal team combination.

[0148] In this embodiment, during the team combination optimization process, by introducing comprehensive evaluation indicators such as job suitability score, project schedule coordination index, multi-job conflict avoidance coefficient, and team historical cooperation tacit understanding, compared with the traditional manual combination or random recommendation method, it can systematically evaluate the overall stability and construction coordination of the team, and achieve the beneficial effects of improving the rationality of team combination and reducing construction risks and management costs.

[0149] Example 2

[0150] This embodiment provides a method for intelligent screening of qualification certificates for bidding in telecommunications projects in the construction industry, including the following steps:

[0151] Step 1: Obtain the project information text of the target communication project in the construction industry; the project information text includes: a description of the communication project, total cost, construction period requirements, geographical and climatic characteristics, construction technology limitations, required qualification types, and the job roles of the proposed team.

[0152] Step 2: Extract the capability information for each qualification object that matches the required qualification type from the qualification database;

[0153] The competency information for each certified individual includes: professional category of the certification, professional title level, current certification status, conflict information of ongoing communication engineering projects, certification validity period, historical communication project experience, types of construction techniques mastered, special operation permit qualifications, experience in the service area, and environmental information for each past communication project participated in.

[0154] Step 3: Construct a 3D matching map, which specifically includes:

[0155] The job description texts for different job roles are extracted from the National Vocational Qualification Standards Database, and keywords are extracted. Then, natural language processing technology (such as BERT embedding) is used to vectorize the job tasks and ability requirements to obtain job description vectors.

[0156] Based on the fields of the competency information of the qualification certificate object, a structured processing is performed to construct the qualification certificate competency vector.

[0157] Construct a three-dimensional matching map, including: job title dimension: such as "Registered Electrical Engineer"; job description vector dimension: representing the ability requirements of the role through semantic embedding; and qualification certificate ability vector dimension: representing the comprehensive ability of the qualification certificate holder.

[0158] In a three-dimensional structure, the job description vector of the "Registered Electrical Engineer" position is semantically matched with the ability vectors of multiple candidate qualification objects to construct a three-dimensional matching graph.

[0159] Step 4: Based on the three-dimensional matching graph, a pre-constructed multi-channel neural semantic matching network is used to match and score the project information text of the qualification certificate object and the target communication project, so as to obtain the semantic matching score of each qualification certificate object.

[0160] In this embodiment, the preset multi-channel neural network includes the following three channels:

[0161] Task Intent Channel: Extract the job description vector of "Registered Electrical Engineer" using an attention mechanism and output the task matching score;

[0162] Specifically, the task intent channel introduces an attention mechanism to extract key information from the job description vector using weighted methods. This attention mechanism, based on a defined task intent query vector, calculates the similarity between the query vector and each word vector to obtain the importance weight of each word. This weighted sum is then applied to the original word vectors to obtain a task intent vector representing the core intent of the job responsibilities. This task intent vector effectively expresses the key functional requirements and task orientation of the target job.

[0163] After obtaining the task intent vector, the channel calculates its similarity with the vectors of candidate matching objects (such as qualification certificate ability vectors), and outputs a task matching score using matching functions such as cosine similarity, vector dot product, or multilayer perceptron. This score measures the degree of semantic matching between the candidate object and the job task requirements, providing input for subsequent multi-channel fusion.

[0164] In this way, the task intent channel can extract semantic features related to the core tasks of a job from the unstructured job description vector, and achieve efficient task intent modeling and matching evaluation, thereby improving the accuracy and intelligence of the overall multi-channel matching network in talent recommendation or job matching scenarios.

[0165] Competency Feature Channel: Extract key competency dimensions of the qualification certificate holder (e.g., "professional title = senior"); combine them with the competency requirements set for the job, and perform in-depth matching, with a score such as 0.91.

[0166] Environmental experience channel: This channel is used to input the environmental information of the target project and the historical environmental information of the qualification certificate object into a graph neural network, extract environmental features, and output an environmental matching score.

[0167] The scores from the three channels will be weighted together to generate a final semantic matching score for each qualification holder for each job role in the target project.

[0168] Step 5: Based on the job roles of the proposed team for the target communication project and the semantic matching score of each certification object, determine the optimal team combination and the final set of certifications, specifically including:

[0169] Step 5-1: For each job role in the proposed team, select M qualification certificate objects in descending order of semantic matching score with the job role, combine them into a candidate set of personnel for the job role, and use the semantic matching score between the M qualification certificate objects and the job role as the job suitability score.

[0170] For each job role candidate set, divide all objects in that job role candidate set into groups;

[0171] The number of people in each target group is the number of qualified candidates required for the specific role in the proposed team.

[0172] Randomly select one group from the candidate set of personnel for each position and combine them to obtain a complete team candidate combination set;

[0173] For example, suppose the required job roles are: 1 Level 1 Construction Engineer, 1 Registered Electrical Engineer, 1 Low-Voltage Engineer, 2 Safety Officers, and 1 Quality Inspector.

[0174] Then, the semantic matching score between each qualification certificate object and each job position is obtained.

[0175] For each job role, select the top M=5 qualification certificate objects with the highest semantic matching degree (job candidate set).

[0176] The candidate set for Level 1 Construction Engineers is as follows: A1 (semantic matching score 0.94), A2 (semantic matching score 0.92), A3 (semantic matching score 0.90), A4 (semantic matching score 0.88), and A5 (semantic matching score 0.85).

[0177] Candidate set for Registered Electrical Engineers: B1 (semantic matching score 0.93), B2 (semantic matching score 0.91), B3 (semantic matching score 0.89), B4 (semantic matching score 0.87), B5 (semantic matching score 0.85).

[0178] Candidate set of low-voltage engineers: C1 (semantic matching score 0.96), C2 (semantic matching score 0.95), C3 (semantic matching score 0.93), C4 (semantic matching score 0.90), C5 (semantic matching score 0.88).

[0179] Safety officer candidate set (requires 2 people): D1 (semantic matching score 0.91), D2 (semantic matching score 0.90), D3 (semantic matching score 0.88), D4 (semantic matching score 0.87), D5 (semantic matching score 0.86).

[0180] Quality inspector candidate set: E1 (semantic matching score 0.89), E2 (semantic matching score 0.87), E3 (semantic matching score 0.86), E4 (semantic matching score 0.85), E5 (semantic matching score 0.84).

[0181] Generate target groups based on the number of people required: First-level construction engineer (1 person): Group = [A1], [A2], [A3], [A4], [A5] (5 groups in total); Registered electrical engineer (1 person): Group = [B1], [B2], [B3], [B4], [B5]; Low-voltage engineer (1 person): Group = [C1], [C2], [C3], [C4], [C5]; Safety officer (2 people): Select 2 people from 5 people, and form a total of C(5,2) = 10 groups, for example: [D1+D2], [D1+D3], [D1+D4], [D1+D5], [D2+D3], [D2+D4], [D2+D5], [D3+D4], [D3+D5], [D4+D5]. Quality Inspector (1 person): Group = [E1], [E2], [E3], [E4], [E5].

[0182] Randomly select a group from each job role's target group to form a complete team candidate combination. For example, the following combination could be selected: Level 1 Construction Engineer A3, Registered Electrical Engineer B1, Low-Voltage Engineer C2, Safety Officer D1+D4, Quality Inspector E2. This combination is one of the candidates in the "Complete Team Candidate Combination Set".

[0183] Step 5-2: Obtain the comprehensive score for each complete team candidate combination set using the preset comprehensive evaluation function;

[0184] The comprehensive evaluation score is formed by weighting the following indicators: individual job suitability score of the candidate team, team schedule coordination index, multi-job conflict avoidance coefficient, and team historical cooperation tacit understanding.

[0185] The comprehensive evaluation function is:

[0186]

[0187] MatchScore k The job suitability score for the kth qualified candidate in the complete team candidate combination set;

[0188] XT represents the team duration coordination index within the complete set of candidate team combinations;

[0189] GB is the multi-position conflict avoidance coefficient in the complete team candidate combination set;

[0190] LS represents the historical teamwork and synergy among the complete set of candidate team combinations.

[0191] δ is a preset first coefficient; ε is a preset second coefficient; θ is a preset third coefficient; and ω is a preset fourth coefficient. In this embodiment, the preset first coefficient δ is 0.1, the preset second coefficient ε is 0.2, the preset third coefficient θ is 0.3, and the preset fourth coefficient ω is 0.4.

[0192] The calculation method for the team schedule coordination index includes:

[0193] For each certified object in the complete team candidate combination set, obtain the estimated end time and reserved buffer period of its current construction project to determine its available entry time window; map the available entry time window of each certified object to a Gaussian time distribution function; calculate the overlap integral of the entry time distribution functions among all certified objects to obtain the degree of collaboration between each certified object pair; take the average of the degree of collaboration between all certified object pairs as the team's schedule collaboration index;

[0194] For example, to obtain the available entry time window for each qualification holder, for example:

[0195] A3: 2025 / 06 / 01-2025 / 09 / 01 (Center Date = July 1st);

[0196] B1: 2025 / 06 / 15-2025 / 08 / 15 (Center Date = July 15);

[0197] C2: 2025 / 05 / 20-2025 / 08 / 20 (Center Date = July 1st);

[0198] D1: 2025 / 06 / 01-2025 / 07 / 30 (Center Date = June 30);

[0199] D4: 2025 / 06 / 05-2025 / 08 / 05 (Center Date = July 5th);

[0200] E2: 2025 / 06 / 20-2025 / 08 / 10 (Center Date = July 15);

[0201] Convert each time window to a Gaussian distribution function, assuming a uniform variance of 10 days.

[0202] Calculate the overlap integral of the time distribution of each pair of members to obtain the degree of collaboration overlap, for example:

[0203] A3 and B1: 0.85;

[0204] A3 and C2: 0.95;

[0205] A3 and D1: 0.90;

[0206] A3 and D4: 0.92;

[0207] A3 and E2: 0.88;

[0208] ...(A total of C(6,2) = 15 combinations);

[0209] Take the average of all overlap rates, let's assume: average overlap rate = 0.89.

[0210] The method for calculating the conflict avoidance coefficient includes:

[0211] Check whether each certified individual in the complete team candidate combination set has already held multiple roles in other ongoing projects. If so, mark it as a conflict. Check whether each certified individual in the complete team candidate combination set holds multiple job roles in this project, and whether these job roles are mutually exclusive. If so, mark it as a conflict. Divide the actual number of conflicts by the theoretical maximum number of conflicts and take the inverse proportional form to obtain the conflict avoidance coefficient, with a value range of 0 to 1.

[0212] For example, check if a candidate holds multiple roles in other projects:

[0213] Suppose that D1 holds two positions in two other projects simultaneously (referred to as one conflict);

[0214] The rest of the staff did not hold multiple positions.

[0215] Check if the same person holds multiple positions in the current group (not found);

[0216] Total number of conflicts = 1; Theoretical maximum number of conflicts = 6 (each person can only have one conflict at most);

[0217] Conflict avoidance coefficient = 1 - 1 / 6 = 0.833.

[0218] The calculation method for the degree of cooperation includes:

[0219] Extract the joint participation records of any two certified objects in historical communication projects from the complete team candidate combination set; count the number of historical collaborations and the number of successful collaborations between each certified object pair; calculate the collaboration success rate of each certified object pair, and calculate the weighted average based on the number of historical collaborations to obtain the collaboration synergy of the entire candidate team.

[0220] For example, to query the historical collaboration records between individuals:

[0221] A3 and B1: Collaborated 3 times, 2 times successfully → success rate 66.7%;

[0222] A3 and C2: 2 attempts, 2 successes → 100%;

[0223] A3 and D1: No cooperation record;

[0224] B1 and D1: 1 collaboration attempt, failure rate → 0%;

[0225] C2 and D4: 1 collaboration, success rate → 100%;

[0226] D1 and D4: Cooperate 2 times, succeed 1 time → 50%;

[0227] E2 collaborates with other members 1-2 times, with an average success rate of approximately 70%.

[0228] The team's level of teamwork is obtained by weighting the success rate of cooperation among all members:

[0229] (66.7+100+0+0+100+50+70+…) / co-logarithm≈0.71.

[0230] Step 5-3: Based on the comprehensive score of each complete team candidate combination set, determine the optimal team combination and the final set of qualifications;

[0231] The optimal team combination is the set of complete team candidate combinations with the highest comprehensive score; the final certificate set includes the certificate corresponding to each certificate object in the optimal team combination.

[0232] The method in this embodiment does not only select the individual with the strongest ability, but also selects a combination with high overall job suitability score, strong teamwork and good coordination of entry time. It introduces a multi-job conflict avoidance coefficient to prevent one person from appearing in multiple project teams at the same time. The team's historical cooperation tacitness index takes into account actual cooperation experience to improve overall cooperation efficiency.

[0233] In the description of this invention, it should be understood that 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 indicated technical features. Therefore, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0234] In this invention, unless otherwise explicitly specified and limited, the terms "installation," "connection," "linking," and "fixing," etc., should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral part; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; they can refer to the internal communication of two components or the interaction between two components. Those skilled in the art can understand the specific meaning of the above terms in this invention according to the specific circumstances.

[0235] In this invention, unless otherwise explicitly specified and limited, "above" or "below" the second feature can mean that the first and second features are in direct contact, or that they are in indirect contact through an intermediate medium. Furthermore, "above," "over," or "on top" the second feature can mean that the first feature is directly above or diagonally above the second feature, or simply indicates that the first feature is at a higher horizontal level than the second feature. "Below," "below," or "beneath" the second feature can mean that the first feature is directly below or diagonally below the second feature, or simply indicates that the first feature is at a lower horizontal level than the second feature.

[0236] In the description of this specification, the terms "one embodiment," "some embodiments," "embodiment," "example," "specific example," or "some examples," etc., refer to specific features, structures, materials, or characteristics described in connection with that embodiment or example, which are included in at least one embodiment or example of the present invention. 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.

[0237] Although embodiments of the present invention have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those skilled in the art can make modifications, alterations, substitutions and variations to the above embodiments within the scope of the present invention.

Claims

1. A method for intelligent screening of qualification certificates for bidding on telecommunications projects in the construction industry, characterized in that, Includes the following steps: S1. Obtain the project information text of the target communication project in the construction industry; The project information text includes: a description of the communication project, total cost, construction period requirements, geographical and climatic characteristics, construction technology limitations, required qualification certificates, and the job roles of the proposed team. S2. Extract the capability information of each qualification object that matches the required qualification type from the qualification database; Each type of qualification certificate corresponds to a specific job role for the practitioner. The competency information for each certified individual includes: professional category of the certification, professional title level, current certification status, conflict information of ongoing communication engineering projects, certification validity period, historical communication project experience, types of construction techniques mastered, special operation permit qualifications, experience in the service area, and environmental information for each past communication project participated in. S3. Construct a three-dimensional matching map based on the job roles and qualification information of the proposed team members for the target communication project. S4. Based on the three-dimensional matching graph, a pre-constructed multi-channel neural semantic matching network is used to perform semantic matching scoring on the project information text of the qualification certificate object and the target communication project, and obtain the semantic matching score of each qualification certificate object. S5. Based on the job roles of the proposed team for the target communication project and the semantic matching score of each certification object, determine the optimal team combination and the final set of certifications. The pre-built multi-channel neural semantic matching network includes: The task intent channel is used to process the job description vector using an attention mechanism network and output a task matching score. The competency feature channel is used to perform a non-linear transformation on the qualification competency vector using a deep feedforward neural network and output a competency matching score. The environmental experience channel is used to input the environmental information of the target project and the historical project environmental information of the certification object into a graph neural network, extract environmental features and output an environmental matching score. The project environmental information comes from the fields of geographical and climatic characteristics and construction technology limitations in the project information text; The historical project environmental information of the qualification certificate holder comes from the environmental information of its past participation in communication projects, including geographical area, construction conditions, and construction type; The three channels output task matching score, ability matching score and environment matching score respectively. They are input into the weighted fusion module to calculate the semantic matching score and output the semantic matching score between the job role and the qualification certificate object.

2. The intelligent qualification screening method for bidding on telecommunications projects in the construction industry according to claim 1, characterized in that, S3 specifically includes: S31. Based on the job descriptions published in the National Vocational Qualification Standards Database, extract the job task keywords and ability requirements for each role in the proposed team for the target communication project, and construct the job description vector corresponding to each role in the proposed team for the target communication project. The National Vocational Qualification Standards Database is a collection of publicly available vocational qualification standards information resources released by the Ministry of Human Resources and Social Security or industry regulatory authorities. This database includes texts of job descriptions for all positions in the field of telecommunications construction. S32. Extract the capability information of each qualification object that matches the required qualification type from the qualification database, and construct a qualification capability vector; The qualification certificate capability vector is a multi-dimensional vector representation generated by structuring the capability information of the qualification certificate object; S33. Natural language processing technology is used to semantically parse the job description vector and qualification certificate ability vector, and map them into a unified vector space to form a three-dimensional matching map with a three-dimensional structure. The three-dimensional structure includes: job title dimension, job responsibility description vector dimension, and qualification certificate ability vector dimension.

3. The intelligent qualification screening method for bidding on communication projects in the construction industry according to claim 2, characterized in that, S31 specifically includes: S311. Obtain the list of proposed job roles in the target communication project; S312. Extract the text of the job description corresponding to each job role from the national vocational qualification standard database; S313. Perform natural language preprocessing on the text of the job description to extract the job task keywords and ability requirements corresponding to the job role. S314. Convert the extracted keywords and ability requirements into vector representations of a unified dimension, and construct a job responsibility description vector for each role in the proposed team for the target communication project.

4. The intelligent qualification screening method for bidding on communication projects in the construction industry according to claim 3, characterized in that, S33 specifically includes: S331. Based on the preset unified semantic vector space model, the job description vector and the qualification certificate ability vector are mapped to semantic vector representations respectively. S332. Based on three dimensions—job title, job description, and qualification / ability—construct a three-dimensional structural matching graph to represent the semantic matching relationship between each job responsibility and qualification / ability.

5. The intelligent qualification screening method for bidding on communication projects in the construction industry according to claim 4, characterized in that, The weighted fusion module uses formula (1) to calculate the matching score and obtain the semantic matching score between the job role and the qualification certificate object; the formula (1) is: ； This represents the semantic matching score between the i-th job role and the j-th qualification certificate object; This represents the task matching score between the i-th job role and the j-th qualification certificate object, as output by the task intent channel. This represents the ability matching score between the i-th job role and the j-th qualification certificate object, output by the ability feature channel. This represents the environmental experience matching score between the i-th job role and the j-th qualification object, output by the environmental experience channel. The first weighting coefficient; This is the second weighting coefficient; This is the third weighting coefficient; where, .

6. The intelligent qualification screening method for bidding on communication projects in the construction industry according to claim 5, characterized in that, The task matching score is the semantic similarity value between the job description vector and the qualification certificate ability vector, which is measured by cosine similarity. The competency matching score is a weighted score based on the specific requirements of the job role for the type of qualification certificate, professional title level, and mastery of construction technology, which assesses the degree of matching between the qualification certificate holder's competency vector and the corresponding fields. The environment matching score is calculated by embedding the historical communication project environment information of the qualification certificate object and the environmental information of the target communication project into a graph structure using a graph neural network, and then calculating the similarity value of the embedding vectors between the two.

7. The intelligent qualification screening method for bidding on communication projects in the construction industry according to claim 6, characterized in that, S5 specifically includes: S51. For each job role in the proposed team, select M qualification certificate objects in descending order of semantic matching score with the job role, combine them into a candidate set of personnel for the job role, and use the semantic matching score between the M qualification certificate objects and the job role as the job suitability score respectively. For each job role candidate set, divide all objects in that job role candidate set into groups; The number of people in each target group is the number of qualified candidates required for the specific role in the proposed team. Randomly select one group from the candidate set of personnel for each position and combine them to obtain a complete team candidate combination set; S52. Obtain the comprehensive score for each complete team candidate combination set using a preset comprehensive evaluation function; The comprehensive evaluation score is formed by weighting the following indicators: individual job suitability score of the candidate team, team schedule coordination index, multi-job conflict avoidance coefficient, and team historical cooperation tacit understanding. S53. Based on the comprehensive score of each complete team candidate combination set, determine the optimal team combination and the final set of qualifications; The optimal team combination is the set of complete team candidate combinations with the highest comprehensive score; the final certificate set includes the certificate corresponding to each certificate object in the optimal team combination.

8. The intelligent qualification screening method for bidding on telecommunications projects in the construction industry according to claim 7, characterized in that, The comprehensive evaluation function is: ； The job suitability score for the kth qualified candidate in the complete team candidate combination set; The team schedule coordination index is the total team project timeline within the complete set of candidate team combinations. The multi-position conflict avoidance coefficient in the complete team candidate combination set; The historical teamwork and synergy among the complete set of candidate team combinations; The first preset coefficient; This is a preset second coefficient; This is a preset third coefficient; This is the preset fourth coefficient.

9. The intelligent qualification screening method for bidding on telecommunications projects in the construction industry according to claim 8, characterized in that, The calculation method for the team schedule coordination index includes: For each certified object in the complete team candidate combination set, obtain the estimated end time and reserved buffer period of its current ongoing communication project to determine its available entry time window; map the available entry time window of each certified object to a Gaussian time distribution function; calculate the overlap integral of the entry time distribution functions among all certified objects to obtain the degree of collaboration overlap of each certified object pair; take the average of the degree of collaboration overlap of all certified object pairs as the team's schedule collaboration index; The method for calculating the conflict avoidance coefficient includes: Check whether each certified individual in the complete team candidate combination set has already held multiple roles in other ongoing communication projects. If so, mark it as a conflict. Check whether each certified individual in the complete team candidate combination set holds multiple positions in the target communication project, and whether these positions are mutually exclusive. If so, mark it as a conflict. Divide the actual number of conflicts by the theoretical maximum number of conflicts and take the inverse proportional form to obtain the conflict avoidance coefficient, with a value range of 0 to 1. The calculation method for the degree of cooperation includes: Extract the joint participation records of any two certified objects in historical communication projects from the complete team candidate combination set; count the number of historical collaborations and the number of successful collaborations between each certified object pair; calculate the collaboration success rate of each certified object pair, and calculate the weighted average based on the number of historical collaborations to obtain the collaboration synergy of the entire candidate team.

Images