An Expert Matching Method and System Based on Knowledge Expansion and Fusion from a Large Language Model

Abstract

This invention discloses an expert matching method and system based on knowledge expansion and fusion using a large language model. First, basic expert information is collected and enhanced using a pre-trained large language model to obtain expanded expert information. Then, the basic and expanded information are integrated, analyzed, and fused using Roberta and a knowledge graph pre-trained model to form an expert fusion representation. Based on enterprise technical needs, the details of the needs are expanded using the large language model, and combined with the representation analysis of technical needs and enterprise information, a technical needs fusion representation is generated. Finally, experts are screened for keyword relevance to narrow the matching range. The fusion representations of experts and needs are compared, similarity is calculated and ranked, and relevant matching experts are recommended. This invention effectively performs skill matching analysis on external expert information and improves expert matching accuracy and recommendation effectiveness based on semantic analysis of technical needs and expert matching.

Description

Technical Field

[0001] This invention relates to the field of large language model analysis, and more specifically, to an expert matching method and system based on the knowledge expansion and fusion of large language models. Background Technology

[0002] Technology transfer platforms are designed to promote talent-centric technology transfer, connecting enterprises' technological needs with university research talent to effectively match technological achievements with industrial demands. However, current platforms suffer from challenges due to the large number of industries involved and the diverse technical fields of skilled personnel, making precise matching between talent capabilities and enterprise needs difficult and resulting in low efficiency. Furthermore, since most platform data comes from registered users' input and uploads, keyword matching methods based on the talent's professional fields often fail to achieve satisfactory results due to the broad scope of the fields. Therefore, leveraging external knowledge bases and databases to further improve the matching effectiveness of experts is a crucial issue that needs to be addressed.

[0003] In recent years, with the rapid development and widespread application of large language models, their rich background knowledge and massive data training models have provided new possibilities for improving the external knowledge base for expert matching. Therefore, this solution proposes an expert matching method and system based on the knowledge expansion and fusion of large language models. By promoting the matching of enterprise technology needs and experts, it can improve the level of technological innovation in industries and the efficiency of technology transfer of scientific and technological achievements of universities and research institutes. Summary of the Invention

[0004] This invention overcomes the shortcomings of existing technologies and proposes an expert matching method and system based on knowledge expansion and fusion of large language models.

[0005] The first aspect of this invention provides an expert matching method based on knowledge expansion and fusion of a large language model, comprising:

[0006] Obtain basic information about experts in the target domain, generate first query information based on a pre-trained large language model using the basic information of experts, perform information query in the pre-trained large language model, and obtain extended information about experts based on the query results;

[0007] The basic information and extended information of experts are integrated into intermediate information. The intermediate information is represented and analyzed by the Roberta pre-trained representation model and the knowledge graph pre-trained representation model respectively. The two representations are then spliced ​​and fused to obtain the expert fused representation information.

[0008] Based on the target company's technical needs and basic information, a second query is generated. The company's needs are then expanded using a large language model and the second query. The technical needs, basic information, and expanded information are then represented, analyzed, and fused together to obtain the fused representation information of the technical needs.

[0009] Based on keywords related to technical needs, a relevance analysis and filtering were performed on all experts, and an expert list was obtained based on the filtering results.

[0010] Matching analysis is performed based on expert fusion representation information and technology demand fusion representation information. Matching results are calculated based on the similarity of the representations, ranked based on the similarity, and expert recommendations are made based on the ranking results.

[0011] In this solution, the steps of obtaining basic expert information in the target domain, generating first query information based on a pre-trained large language model using the expert information, performing information querying within the pre-trained large language model, and obtaining extended expert information based on the query results are as follows:

[0012] Retrieve basic information about experts in the target field from the expert database;

[0013] In the expert's basic information, determine whether there is any extended information. If not, call the preset prompt template and generate the first query information by combining it with the expert's basic information.

[0014] The first query information is sent to a pre-trained large language model for information retrieval, and the query results are converted into text information and expert extended information is obtained.

[0015] In this solution, the basic expert information and extended expert information are integrated into intermediate information. The intermediate information is then represented and analyzed using both the Roberta pre-trained representation model and the knowledge graph pre-trained representation model. The two representations are then concatenated and fused to obtain the expert fused representation information, including:

[0016] The basic information and extended information of experts are integrated into textual information to form intermediate information;

[0017] Construct a pre-trained representation model based on Roberta;

[0018] The intermediate information is de-noised, and the encoding is segmented using Roberta's bytes.

[0019] The intermediate information is encoded and imported into a pre-trained representation model. The representation data is obtained through forward propagation and selection of a preset output layer.

[0020] By summing all the output representation data, a first text-based representation is obtained;

[0021] The basic and extended information of experts are imported into the semantic analysis model to perform semantic analysis in three dimensions: entity, relation, and attribute, and triple data is generated.

[0022] An expert knowledge graph was constructed using triplet data.

[0023] In the expert knowledge graph, the TransE algorithm is used to vectorize entity and relation data, resulting in graph vector representation data.

[0024] In the graph vector representation data, a similarity distance value is set, and knowledge reasoning is performed on the vectors corresponding to entity data and relation data. The reasoning process is as follows: based on the cosine similarity method, the distance between the selected vector and other domain vectors is calculated. If the distance is lower than the similarity distance value, it is determined that the selected vector has related knowledge. The association information between the corresponding domain vector and the selected vector is recorded. The association analysis is performed on all vectors in the graph vector representation data, and the obtained association information is used as knowledge graph reasoning data.

[0025] Based on knowledge graph reasoning data, knowledge data is supplemented to the expert knowledge graph. At the same time, the triple data in the expert knowledge graph and the knowledge graph structure are updated.

[0026] By using a knowledge graph pre-trained representation model, text representation analysis based on entities, relationships, and attributes is performed on the expert knowledge graph to generate a second representation based on the knowledge graph.

[0027] In this solution, the process of integrating basic expert information and extended expert information into intermediate information, performing representation analysis on the intermediate information using both the Roberta pre-trained representation model and the knowledge graph pre-trained representation model, and then concatenating and fusing the two representations to obtain the expert fused representation information, further includes:

[0028] Obtain the first and second representations;

[0029] Based on the splicing and fusion method, the first representation and the second representation are spliced ​​together into a single representation data, and expert fusion representation information is obtained.

[0030] In this solution, the second query information is generated based on the target company's technical needs and basic information. Then, a large language model is used to perform an extended query on the company's needs, yielding extended information. Finally, the technical needs, basic information, and extended information are represented, analyzed, and fused together to obtain integrated technical needs representation information. Specifically:

[0031] Obtain the target company's technology needs and basic company information;

[0032] A preset prompt template is invoked to generate second query information based on the target company's technical needs and basic company information;

[0033] The second query information is sent to a pre-trained large language model for information retrieval, and the query results are converted into text information to obtain the demand expansion information.

[0034] Semantic representation analysis is performed based on technical requirements, basic enterprise information, and extended requirement information. The resulting representations are then fused and combined to obtain fused representation information of technical requirements.

[0035] In this solution, based on keywords related to technical requirements, a relevance analysis and filtering are performed on all experts, and an expert list is obtained based on the filtering results, specifically as follows:

[0036] The technical requirements are converted into text and segmented into words, and relevant technical keywords are selected through word frequency calculation.

[0037] In the expert database, domain keywords for each expert are extracted. Based on the correlation analysis between domain keywords and technical keywords, experts with a correlation of 0 are filtered out, and an expert list is obtained based on the filtering results.

[0038] In this solution, the matching analysis based on expert fusion representation information and technical demand fusion representation information is performed. The matching results are calculated based on the similarity of the representations, ranked according to the similarity, and expert recommendations are made based on the ranking results. Specifically:

[0039] Obtain the corresponding expert fusion representation information from the expert list and mark it as the current expert fusion representation information;

[0040] The current expert fusion representation information and the technology demand fusion representation information are matched and analyzed based on similarity.

[0041] Similarity is calculated based on a preset distance calculation method, and the similarity is used as the matching degree to sort the expert list and obtain the sorting result;

[0042] Based on the sorting results, a preset number of expert information items are sent from the expert list to preset user terminals for expert recommendation.

[0043] A second aspect of this invention also provides an expert matching system based on the knowledge expansion and fusion of a large language model. This system includes: an expert information expansion module, a technical requirement expansion module, a prompt template management module, an expert information knowledge graph representation module, an expert information text representation module, an expert information fusion representation module, a technical requirement fusion representation module, a similarity calculation module, a matching filtering module, and a matching result output module. The system operates by implementing the following steps:

[0044] Obtain basic information about experts in the target domain, generate first query information based on a pre-trained large language model using the basic information of experts, perform information query in the pre-trained large language model, and obtain extended information about experts based on the query results;

[0045] The basic information and extended information of experts are integrated into intermediate information. The intermediate information is represented and analyzed by the Roberta pre-trained representation model and the knowledge graph pre-trained representation model respectively. The two representations are then spliced ​​and fused to obtain the expert fused representation information.

[0046] Based on the target company's technical needs and basic information, a second query is generated. The company's needs are then expanded using a large language model and the second query. The technical needs, basic information, and expanded information are then represented, analyzed, and fused together to obtain the fused representation information of the technical needs.

[0047] Based on keywords related to technical needs, a relevance analysis and filtering were performed on all experts, and an expert list was obtained based on the filtering results.

[0048] Matching analysis is performed based on expert fusion representation information and technology demand fusion representation information. Matching results are calculated based on the similarity of the representations, ranked based on the similarity, and expert recommendations are made based on the ranking results.

[0049] In this solution, the steps of obtaining basic expert information in the target domain, generating first query information based on a pre-trained large language model using the expert information, performing information querying within the pre-trained large language model, and obtaining extended expert information based on the query results are as follows:

[0050] Retrieve basic information about experts in the target field from the expert database;

[0051] In the expert's basic information, determine whether there is any extended information. If not, call the preset prompt template and generate the first query information by combining it with the expert's basic information.

[0052] The first query information is sent to a pre-trained large language model for information retrieval, and the query results are converted into text information and expert extended information is obtained.

[0053] A third aspect of the present invention also provides a computer-readable storage medium comprising an expert matching program based on the knowledge extension and fusion of a large language model, wherein when the expert matching program based on the knowledge extension and fusion of a large language model is executed by a processor, it implements the steps of the expert matching method based on the knowledge extension and fusion of a large language model as described in any of the preceding claims.

[0054] This invention discloses an expert matching method and system based on knowledge expansion and fusion using a large language model. First, basic expert information is collected and enhanced using a pre-trained large language model to obtain expanded expert information. Then, the basic and expanded information are integrated, analyzed, and fused using Roberta and a knowledge graph pre-trained model to form an expert fusion representation. Based on enterprise technical needs, the details of the needs are expanded using the large language model, and combined with the representation analysis of technical needs and enterprise information, a technical needs fusion representation is generated. Finally, experts are screened for keyword relevance to narrow the matching range. The fusion representations of experts and needs are compared, similarity is calculated and ranked, and relevant matching experts are recommended. This invention effectively performs skill matching analysis on external expert information and improves expert matching accuracy and recommendation effectiveness based on semantic analysis of technical needs and expert matching. Attached Figure Description

[0055] Figure 1 A flowchart of an expert matching method based on knowledge extension and fusion of a large language model according to the present invention is shown;

[0056] Figure 2 The flowchart of the expert fusion characterization information acquisition process of the present invention is shown;

[0057] Figure 3 The flowchart for obtaining the technical requirement fusion characterization information of the present invention is shown;

[0058] Figure 4 The diagram shows a block diagram of an expert matching system based on knowledge expansion and fusion of a large language model according to the present invention. Detailed Implementation

[0059] To better understand the above-mentioned objectives, features, and advantages of the present invention, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be noted that, unless otherwise specified, the embodiments and features described in these embodiments can be combined with each other.

[0060] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and therefore the scope of protection of the invention is not limited to the specific embodiments disclosed below.

[0061] In addressing the challenge of matching experts with industry technology needs across different fields, and considering the technical difficulties posed by the fact that both parties in the matching process are described by multimodal data such as text and images, a method and system for expert matching based on knowledge expansion and fusion of a large language model is proposed.

[0062] Figure 1 The flowchart of an expert matching method based on knowledge expansion and fusion of a large language model according to the present invention is shown.

[0063] like Figure 1 As shown, the first aspect of this invention provides an expert matching method based on knowledge expansion and fusion of a large language model, comprising:

[0064] S102, obtain basic information of experts in the target domain, generate first query information based on a pre-trained large language model using the basic information of experts, perform information query in the pre-trained large language model, and obtain expert extended information based on the query results;

[0065] S104 integrates the basic information and extended information of experts into intermediate information. The intermediate information is then represented and analyzed by the Roberta pre-trained representation model and the knowledge graph pre-trained representation model. The two representations are then spliced ​​and fused to obtain the expert fused representation information.

[0066] S106. Generate second query information based on the target company's technical needs and basic information. Perform enterprise demand extension query through the big language model and the second query information to obtain demand extension information. Perform characterization analysis and fusion splicing of technical needs, basic information, and extension information to obtain technical demand fusion characterization information.

[0067] S108: Based on keywords related to technical requirements, perform relevance analysis and filtering on all experts, and obtain an expert list based on the filtering results.

[0068] S110 performs matching analysis based on expert fusion representation information and technical demand fusion representation information, calculates matching results based on the similarity of representations, ranks them based on similarity, and makes expert recommendations based on the ranking results.

[0069] It should be noted that further information from experts is required to obtain more data on their technical capabilities, including but not limited to representative achievements and major industry applications.

[0070] According to an embodiment of the present invention, the steps of obtaining basic expert information in the target domain, generating first query information based on a pre-trained large language model using the basic expert information, performing information querying within the pre-trained large language model, and obtaining extended expert information based on the query results are as follows:

[0071] Retrieve basic information about experts in the target field from the expert database;

[0072] In the expert's basic information, determine whether there is any extended information. If not, call the preset prompt template and generate the first query information by combining it with the expert's basic information.

[0073] The first query information is sent to a pre-trained large language model for information retrieval, and the query results are converted into text information and expert extended information is obtained.

[0074] It should be noted that the basic expert information includes information obtained from multiple experts for subsequent screening and analysis. The pre-trained large language model can be a large language model such as GPT or BERT.

[0075] The following is a test example demonstrating the use of a preset prompt template:

[0076] Q: Please provide information on the technical capabilities of physics expert Yang Zhenning and list his major research achievements.

[0077] A: Chen-Ning Yang is an outstanding physicist whose major research achievements include, but are not limited to, the following aspects.

[0078] 1. Mirror symmetry breaking theory: He co-proposed this theory with Tsung-Dao Lee, stating that mirror symmetry can be broken in weak interactions. This theory, proposed in 1957, earned them the Nobel Prize in Physics.

[0079] 2. Polaron theory: In collaboration with T.D. Lee, he proposed a theory about polarons, a type of meson.

[0080] 3. Microscopic structure of matter and high-energy physics: Yang Zhenning has made many contributions to high-energy physics, including the understanding of particle physics and weak interactions.

[0081] 4. Statistical Mechanics and Nonequilibrium Physics: He also made some important contributions to statistical mechanics and nonequilibrium physics.

[0082] 5. Education and Academic Leadership: Yang Zhenning not only has outstanding achievements in scientific research, but also has a profound influence in the field of education. He has taught at many well-known universities and mentored many students.

[0083] Overall, Yang Zhenning is renowned for his pioneering work in theoretical physics and particle physics, as well as his contributions to physics education.

[0084] Information is extracted and text-transformed from the answers to obtain relevant expert supplementary information.

[0085] Figure 2 The flowchart of the expert fusion characterization information acquisition process of the present invention is shown;

[0086] like Figure 2 The above is a simplified flowchart of the analysis and acquisition of expert fusion representation information; the specific process is described in the embodiments of the present invention.

[0087] According to an embodiment of the present invention, the process of integrating basic expert information and extended expert information into intermediate information, performing representation analysis on the intermediate information using a Roberta pre-trained representation model and a knowledge graph pre-trained representation model respectively, and then concatenating and fusing the two representations to obtain expert fused representation information includes:

[0088] The basic information and extended information of experts are integrated into textual information to form intermediate information;

[0089] Construct a pre-trained representation model based on Roberta;

[0090] The intermediate information is de-noised, and the encoding is segmented using Roberta's bytes.

[0091] The intermediate information is encoded and imported into a pre-trained representation model. The representation data is obtained through forward propagation and selection of a preset output layer.

[0092] By summing all the output representation data, a first text-based representation is obtained;

[0093] The basic and extended information of experts are imported into the semantic analysis model to perform semantic analysis in three dimensions: entity, relation, and attribute, and triple data is generated.

[0094] An expert knowledge graph was constructed using triplet data.

[0095] In the expert knowledge graph, the TransE algorithm is used to vectorize entity and relation data, resulting in graph vector representation data.

[0096] In the graph vector representation data, a similarity distance value is set, and knowledge reasoning is performed on the vectors corresponding to entity data and relation data. The reasoning process is as follows: based on the cosine similarity method, the distance between the selected vector and other domain vectors is calculated. If the distance is lower than the similarity distance value, it is determined that the selected vector has related knowledge. The association information between the corresponding domain vector and the selected vector is recorded. The association analysis is performed on all vectors in the graph vector representation data, and the obtained association information is used as knowledge graph reasoning data.

[0097] Based on knowledge graph reasoning data, knowledge data is supplemented to the expert knowledge graph. At the same time, the triple data in the expert knowledge graph and the knowledge graph structure are updated.

[0098] By using a knowledge graph pre-trained representation model, text representation analysis based on entities, relationships, and attributes is performed on the expert knowledge graph to generate a second representation based on the knowledge graph.

[0099] It should be noted that the Roberta (Robustly Optimized BERT Pretraining Approach) is a deep learning model based on the Transformer architecture for natural language processing tasks. The semantic analysis model is a CNN-based semantic analysis model trained using historical expert text information. Knowledge graph reasoning data is generally relational reasoning data. The knowledge graph pre-trained representation model is a BERT-based semantic representation analysis model that extracts textual semantic representations from triples (entities, relations, attributes) in the knowledge graph.

[0100] It is worth mentioning that in the process of expert skill analysis and demand matching, traditional technologies are often limited to keyword matching of expert text information, which is less accurate and effective in complex demand analysis. Therefore, this invention integrates two representations (information text and knowledge graph) to achieve multi-dimensional mining and analysis of expert skill information. Furthermore, in the knowledge graph, a distributed representation learning algorithm (TransE) is used for knowledge reasoning to mine the expert's associated skill information and related technical information. Subsequently, the representations are spliced ​​and fused to achieve accurate matching between expert information and demand, thereby improving the recommendation effect.

[0101] According to an embodiment of the present invention, the step of integrating basic expert information and extended expert information into intermediate information, performing representation analysis on the intermediate information using a Roberta pre-trained representation model and a knowledge graph pre-trained representation model respectively, and then splicing and fusing the two representations to obtain expert fused representation information further includes:

[0102] Obtain the first and second representations;

[0103] Based on the splicing and fusion method, the first representation and the second representation are spliced ​​together into a single representation data, and expert fusion representation information is obtained.

[0104] It should be noted that the splicing and fusion method can preserve the original representation information to the greatest extent, enabling more comprehensive and accurate retrieval and analysis in the future.

[0105] Figure 3 The flowchart for obtaining the technical requirement fusion characterization information of the present invention is shown;

[0106] like Figure 3 The above is a simplified flowchart for the analysis and acquisition of technical requirement fusion characterization information; the specific process is described in the embodiments of the present invention.

[0107] According to an embodiment of the present invention, the step of generating second query information based on the target enterprise's technical needs and basic enterprise information, performing enterprise demand expansion query through a large language model and the second query information to obtain demand expansion information, and performing characterization analysis and fusion splicing of technical needs, basic enterprise information, and demand expansion information to obtain fused characterization information of technical needs, specifically:

[0108] Obtain the target company's technology needs and basic company information;

[0109] A preset prompt template is invoked to generate second query information based on the target company's technical needs and basic company information;

[0110] The second query information is sent to a pre-trained large language model for information retrieval, and the query results are converted into text information to obtain the demand expansion information.

[0111] Semantic representation analysis is performed based on technical requirements, basic enterprise information, and extended requirement information. The resulting representations are then fused and combined to obtain fused representation information of technical requirements.

[0112] It should be noted that in this invention, text representations can be extracted using methods such as word embedding or sentence vectors. The extended requirement information includes textual information such as technical characteristics and talent competency requirements within the enterprise's needs.

[0113] The following is a test example demonstrating the process of obtaining extended requirement information through the second query after invoking the preset template:

[0114] Q: I have a technical requirement to develop an expert recommendation system. Could you please help me find information on the technologies required and the technical requirements for the personnel?

[0115] A: Developing an expert recommendation system involves multiple technical fields and talent requirements. The following are the main technical information and talent requirements to consider:

[0116] Technical Information:

[0117] 1. Data Collection and Processing:

[0118] - Web scraping and data cleaning: Obtain data from various data sources (such as academic paper databases, expert social networks, internal enterprise systems, etc.) and perform cleaning, deduplication, and standardization.

[0119] 2. Data storage and management:

[0120] - Database design and management: Select appropriate database technologies (such as relational databases or NoSQL databases), design data models for storing experts and related data, and ensure efficient data retrieval and updating.

[0121] 3. Information retrieval and recommendation algorithms:

[0122] - Natural Language Processing (NLP): Used to extract key information from texts such as expert papers and project descriptions.

[0123] - Information retrieval technology: Building search engines or indexes for retrieving expert information.

[0124] - Recommendation system algorithms: Use machine learning algorithms (such as collaborative filtering, content recommendation, deep learning models, etc.) to recommend suitable experts to users.

[0125] 4. Front-end development:

[0126] - User Interface Design and Development: Design a user-friendly interface that allows users to easily enter queries, view recommended expert lists and detailed information.

[0127] 5. Backend Development:

[0128] - Application logic development: Implement the core functions of the system, handle user requests, call data processing and recommendation algorithms, and provide API interfaces for front-end calls.

[0129] 6. Security and Privacy:

[0130] - Data security and privacy protection: Ensure the security of user data and expert information, and comply with relevant laws, regulations and privacy policies.

[0131] Talent and technical requirements:

[0132] 1. Data Engineer:

[0133] - Proficient in data acquisition, cleaning, and processing techniques.

[0134] - Familiar with various data storage and management systems, and able to design and optimize database structures.

[0135] 2. Machine Learning Engineer:

[0136] - Possess a deep understanding and practical experience with recommendation system algorithms and technologies.

[0137] - Proficient in programming languages ​​such as Python and machine learning frameworks (such as TensorFlow and PyTorch).

[0138] 3. Software Engineer:

[0139] - Proficient in front-end and back-end development technologies, including front-end technologies such as HTML / CSS / JavaScript, and back-end frameworks such as Django and Flask.

[0140] - Capable of designing and implementing scalable and stable applications.

[0141] 4. Information retrieval expert:

[0142] - Understand the principles and implementation of information retrieval technology and search engines.

[0143] Familiarity with search engine optimization (SEO) techniques is a plus.

[0144] 5. Security Expert:

[0145] - Be familiar with best practices for cybersecurity and data privacy protection.

[0146] - Capable of assessing and improving the system's security and privacy protection capabilities.

[0147] 6. Project management and teamwork skills:

[0148] -Able to effectively manage project progress and team collaboration, ensuring projects are delivered on time and meet requirements.

[0149] According to an embodiment of the present invention, the process of performing relevance analysis and filtering on all experts based on keywords related to technical requirements, and obtaining an expert list based on the filtering results, specifically involves:

[0150] The technical requirements are converted into text and segmented into words, and relevant technical keywords are selected through word frequency calculation.

[0151] In the expert database, domain keywords for each expert are extracted. Based on the correlation analysis between domain keywords and technical keywords, experts with a correlation of 0 are filtered out, and an expert list is obtained based on the filtering results.

[0152] It should be noted that, in the semantic analysis of the expert information corresponding to a relevance of 0, there are no experts whose domain keywords do not contain any technical keywords or related technical terms.

[0153] According to an embodiment of the present invention, the matching analysis based on expert fusion representation information and technical demand fusion representation information, calculating the matching result based on the similarity of the representations, ranking them based on the similarity, and making expert recommendations based on the ranking results, specifically includes:

[0154] Obtain the corresponding expert fusion representation information from the expert list and mark it as the current expert fusion representation information;

[0155] The current expert fusion representation information and the technology demand fusion representation information are matched and analyzed based on similarity.

[0156] Similarity is calculated based on a preset distance calculation method, and the similarity is used as the matching degree to sort the expert list and obtain the sorting result;

[0157] Based on the sorting results, a preset number of expert information items are sent from the expert list to preset user terminals for expert recommendation.

[0158] It should be noted that the expert recommendations refer to a predetermined number of experts ranked highly. The predetermined distance calculation methods include Euclidean distance and cosine similarity.

[0159] Figure 4 The diagram shows a block diagram of an expert matching system based on knowledge expansion and fusion of a large language model according to the present invention.

[0160] A second aspect of this invention also provides an expert matching system based on the knowledge expansion and fusion of a large language model. This system includes: an expert information expansion module, a technical requirement expansion module, a prompt template management module, an expert information knowledge graph representation module, an expert information text representation module, an expert information fusion representation module, a technical requirement fusion representation module, a similarity calculation module, a matching filtering module, and a matching result output module. The system operates by implementing the following steps:

[0161] Obtain basic information about experts in the target domain, generate first query information based on a pre-trained large language model using the basic information of experts, perform information query in the pre-trained large language model, and obtain extended information about experts based on the query results;

[0162] The basic information and extended information of experts are integrated into intermediate information. The intermediate information is represented and analyzed by the Roberta pre-trained representation model and the knowledge graph pre-trained representation model respectively. The two representations are then spliced ​​and fused to obtain the expert fused representation information.

[0163] Based on the target company's technical needs and basic information, a second query is generated. The company's needs are then expanded using a large language model and the second query. The technical needs, basic information, and expanded information are then represented, analyzed, and fused together to obtain the fused representation information of the technical needs.

[0164] Based on keywords related to technical needs, a relevance analysis and filtering were performed on all experts, and an expert list was obtained based on the filtering results.

[0165] Matching analysis is performed based on expert fusion representation information and technology demand fusion representation information. Matching results are calculated based on the similarity of the representations, ranked based on the similarity, and expert recommendations are made based on the ranking results.

[0166] The expert matching system based on the knowledge extension and fusion of a large language model specifically includes:

[0167] The expert information extension module is responsible for calling the interface of the pre-trained large language model (such as GPT, BERT, etc.) through the prompt template to obtain the corresponding expert extension information, and storing the expert extension information in the database as the expert's extension information;

[0168] The technical requirement extension module uses the prompt module to call the pre-trained large language model interface to obtain extended information about the technical requirements and store it in the database. For non-confidential technical requirements, basic enterprise information can be added to the prompt template.

[0169] The prompt template management module is responsible for managing the information templates that call the pre-trained large language model interface, including adding, deleting, modifying and querying template information;

[0170] The expert information knowledge graph representation module is responsible for extracting keywords from expert information, constructing a knowledge graph, and generating the expert's knowledge graph representation (i.e., the second representation) through a knowledge graph pre-training model.

[0171] The expert information text representation module is responsible for integrating basic expert information and extended expert information, and generating a text representation (i.e., the first representation) of the expert information through a pre-trained representation model.

[0172] The expert information fusion and representation module is responsible for fusing the expert's knowledge graph representation with the text representation;

[0173] The technology requirement fusion and characterization module is responsible for splicing together the enterprise's technology requirements, basic enterprise information, and expanded information to generate the enterprise's technology requirement fusion and characterization information.

[0174] The similarity calculation module is responsible for calculating the similarity between expert representations and demand representations (similarity calculation can use Euclidean distance or cosine similarity) to obtain the expert similarity ranking based on specific technical requirements.

[0175] The matching and filtering module filters out experts whose domain keywords do not contain any technical requirement keywords, based on the keywords of technical requirements. This ensures that at least one expert in the matched expert information contains a technical requirement keyword, serving as a correction mechanism for the model.

[0176] The matching result output module outputs expert matching results based on user needs, using similarity as the matching value for recommendations by users or downstream tasks. It can output a preset number of expert recommendations.

[0177] A third aspect of the present invention also provides a computer-readable storage medium comprising an expert matching program based on the knowledge extension and fusion of a large language model, wherein when the expert matching program based on the knowledge extension and fusion of a large language model is executed by a processor, it implements the steps of the expert matching method based on the knowledge extension and fusion of a large language model as described in any of the preceding claims.

[0178] This invention discloses an expert matching method and system based on knowledge expansion and fusion using a large language model. First, basic expert information is collected and enhanced using a pre-trained large language model to obtain expanded expert information. Then, the basic and expanded information are integrated, analyzed, and fused using Roberta and a knowledge graph pre-trained model to form an expert fusion representation. Based on enterprise technical needs, the details of the needs are expanded using the large language model, and combined with the representation analysis of technical needs and enterprise information, a technical needs fusion representation is generated. Finally, experts are screened for keyword relevance to narrow the matching range. The fusion representations of experts and needs are compared, similarity is calculated and ranked, and relevant matching experts are recommended. This invention effectively performs skill matching analysis on external expert information and improves expert matching accuracy and recommendation effectiveness based on semantic analysis of technical needs and expert matching.

[0179] In the several embodiments provided in this application, it should be understood that the disclosed devices and methods can be implemented in other ways. The device embodiments described above are merely illustrative. For example, the division of units is only a logical functional division, and in actual implementation, there may be other division methods, such as: multiple units or components can be combined, or integrated into another system, or some features can be ignored or not executed. In addition, the coupling, direct coupling, or communication connection between the various components shown or discussed can be through some interfaces, and the indirect coupling or communication connection between devices or units can be electrical, mechanical, or other forms.

[0180] The units described above as separate components may or may not be physically separate. The components shown as units may or may not be physical units. They may be located in one place or distributed across multiple network units. Some or all of the units may be selected to achieve the purpose of this embodiment according to actual needs.

[0181] In addition, in the various embodiments of the present invention, each functional unit can be integrated into one processing unit, or each unit can be a separate unit, or two or more units can be integrated into one unit; the integrated unit can be implemented in hardware or in the form of hardware plus software functional units.

[0182] Those skilled in the art will understand that all or part of the steps of the above method embodiments can be implemented by hardware related to program instructions. The aforementioned program can be stored in a computer-readable storage medium. When the program is executed, it performs the steps of the above method embodiments. The aforementioned storage medium includes various media capable of storing program code, such as mobile storage devices, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0183] Alternatively, if the integrated units of this invention are implemented as software functional modules and sold or used as independent products, they can also be stored in a computer-readable storage medium. Based on this understanding, the technical solutions of the embodiments of this invention, or the parts that contribute to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the methods described in the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as mobile storage devices, ROM, RAM, magnetic disks, or optical disks.

[0184] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in the present invention should be included within the scope of protection of the present invention.

Claims

1. An expert matching method based on knowledge expansion and fusion of a large language model, characterized in that, include: Obtain basic information about experts in the target domain, generate first query information based on a pre-trained large language model using the basic information of experts, perform information query in the pre-trained large language model, and obtain extended information about experts based on the query results; The basic information and extended information of experts are integrated into intermediate information. The intermediate information is represented and analyzed by the Roberta pre-trained representation model and the knowledge graph pre-trained representation model respectively. The two representations are then spliced ​​and fused to obtain the expert fused representation information. Based on the target company's technical needs and basic information, a second query is generated. The company's needs are then expanded using a large language model and the second query. The technical needs, basic information, and expanded information are then represented, analyzed, and fused together to obtain the fused representation information of the technical needs. Based on keywords related to technical needs, a relevance analysis and filtering were performed on all experts, and an expert list was obtained based on the filtering results. Matching analysis is performed based on expert fusion representation information and technical requirement fusion representation information. Matching results are calculated based on the similarity of the representations, ranked according to similarity, and then used for expert recommendations based on the ranking results. Specifically, this also includes: Obtain the corresponding expert fusion representation information from the expert list and mark it as the current expert fusion representation information; The current expert fusion representation information and the technology demand fusion representation information are matched and analyzed based on similarity. Similarity is calculated based on a preset distance calculation method, and the similarity is used as the matching degree to sort the expert list and obtain the sorting result; Based on the sorting results, a preset number of expert information items are sent from the expert list to preset user terminals for expert recommendation.

2. The expert matching method based on knowledge expansion and fusion of a large language model according to claim 1, characterized in that, The process of obtaining basic expert information in the target domain, generating first query information based on a pre-trained large language model using this basic expert information, performing information retrieval within the pre-trained large language model, and obtaining extended expert information based on the query results specifically involves: Retrieve basic information about experts in the target field from the expert database; In the expert's basic information, determine whether there is any extended information. If not, call the preset prompt template and generate the first query information by combining it with the expert's basic information. The first query information is sent to a pre-trained large language model for information retrieval, and the query results are converted into text information and expert extended information is obtained.

3. The expert matching method based on knowledge expansion and fusion of a large language model according to claim 2, characterized in that, The process involves integrating basic expert information and extended expert information into intermediate information. This intermediate information is then represented and analyzed using both a Roberta pre-trained representation model and a knowledge graph pre-trained representation model. The two representations are then concatenated and fused to obtain the expert fused representation information, including: The basic information and extended information of experts are integrated into textual information to form intermediate information; Construct a pre-trained representation model based on Roberta; The intermediate information is de-noised, and the encoding is segmented using Roberta's bytes. The intermediate information is encoded and imported into a pre-trained representation model. The representation data is obtained through forward propagation and selection of a preset output layer. By summing all the output representation data, a first text-based representation is obtained; The basic and extended information of experts are imported into the semantic analysis model to perform semantic analysis in three dimensions: entity, relation, and attribute, and triple data is generated. An expert knowledge graph was constructed using triplet data. In the expert knowledge graph, the TransE algorithm is used to vectorize entity and relation data, resulting in graph vector representation data. In the graph vector representation data, a similarity distance value is set, and knowledge reasoning is performed on the vectors corresponding to entity data and relation data. The reasoning process is as follows: based on the cosine similarity method, the distance between the selected vector and other domain vectors is calculated. If the distance is lower than the similarity distance value, it is determined that the selected vector has related knowledge. The association information between the corresponding domain vector and the selected vector is recorded. The association analysis is performed on all vectors in the graph vector representation data, and the obtained association information is used as knowledge graph reasoning data. Based on knowledge graph reasoning data, knowledge data is supplemented to the expert knowledge graph. At the same time, the triple data in the expert knowledge graph and the knowledge graph structure are updated. By using a knowledge graph pre-trained representation model, text representation analysis based on entities, relationships, and attributes is performed on the expert knowledge graph to generate a second representation based on the knowledge graph.

4. The expert matching method based on knowledge expansion and fusion of a large language model according to claim 3, characterized in that, The process of integrating basic expert information and extended expert information into intermediate information, performing representation analysis on the intermediate information using both the Roberta pre-trained representation model and the knowledge graph pre-trained representation model, and then concatenating and fusing the two representations to obtain the expert fused representation information, further includes: Obtain the first and second representations; Based on the splicing and fusion method, the first representation and the second representation are spliced ​​together into a single representation data, and expert fusion representation information is obtained.

5. The expert matching method based on knowledge expansion and fusion of a large language model according to claim 4, characterized in that, The process involves generating second query information based on the target company's technological needs and basic information. Then, using a large language model, a further query is performed to expand the company's needs, yielding expanded demand information. Finally, the technological needs, basic company information, and expanded demand information are represented, analyzed, and fused together to obtain integrated technological demand representation information. Specifically: Obtain the target company's technology needs and basic company information; A preset prompt template is invoked to generate second query information based on the target company's technical needs and basic company information; The second query information is sent to a pre-trained large language model for information retrieval, and the query results are converted into text information to obtain the demand expansion information. Semantic representation analysis is performed based on technical requirements, basic enterprise information, and extended requirement information. The resulting representations are then fused and combined to obtain fused representation information of technical requirements.

6. The expert matching method based on knowledge expansion and fusion of a large language model according to claim 5, characterized in that, Based on the keywords related to technical requirements, a relevance analysis and filtering were performed on all experts, and an expert list was obtained based on the filtering results, specifically as follows: The technical requirements are converted into text and segmented into words, and relevant technical keywords are selected through word frequency calculation. In the expert database, domain keywords for each expert are extracted. Based on the correlation analysis between domain keywords and technical keywords, experts with a correlation of 0 are filtered out, and an expert list is obtained based on the filtering results.

7. An expert matching system based on knowledge expansion and fusion of a large language model, characterized in that, The system includes: an expert information extension module, a technical requirement extension module, a prompt template management module, an expert information knowledge graph representation module, an expert information text representation module, an expert information fusion representation module, a technical requirement fusion representation module, a similarity calculation module, a matching filtering module, and a matching result output module. The system operates by performing the following steps: Obtain basic information about experts in the target domain, generate first query information based on a pre-trained large language model using the basic information of experts, perform information query in the pre-trained large language model, and obtain extended information about experts based on the query results; The basic information and extended information of experts are integrated into intermediate information. The intermediate information is represented and analyzed by the Roberta pre-trained representation model and the knowledge graph pre-trained representation model respectively. The two representations are then spliced ​​and fused to obtain the expert fused representation information. Based on the target company's technical needs and basic information, a second query is generated. The company's needs are then expanded using a large language model and the second query. The technical needs, basic information, and expanded information are then represented, analyzed, and fused together to obtain the fused representation information of the technical needs. Based on keywords related to technical needs, a relevance analysis and filtering were performed on all experts, and an expert list was obtained based on the filtering results. Matching analysis is performed based on expert fusion representation information and technical requirement fusion representation information. Matching results are calculated based on the similarity of the representations, ranked according to similarity, and then used for expert recommendations based on the ranking results. Specifically, this also includes: Obtain the corresponding expert fusion representation information from the expert list and mark it as the current expert fusion representation information; The current expert fusion representation information and the technology demand fusion representation information are matched and analyzed based on similarity. Similarity is calculated based on a preset distance calculation method, and the similarity is used as the matching degree to sort the expert list and obtain the sorting result; Based on the sorting results, a preset number of expert information items are sent from the expert list to preset user terminals for expert recommendation.

8. The expert matching system based on knowledge expansion and fusion of a large language model according to claim 7, characterized in that, The process of obtaining basic expert information in the target domain, generating first query information based on a pre-trained large language model using this basic expert information, performing information retrieval within the pre-trained large language model, and obtaining extended expert information based on the query results specifically involves: Retrieve basic information about experts in the target field from the expert database; In the expert's basic information, determine whether there is any extended information. If not, call the preset prompt template and generate the first query information by combining it with the expert's basic information. The first query information is sent to a pre-trained large language model for information retrieval, and the query results are converted into text information and expert extended information is obtained.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium includes an expert matching program based on the knowledge extension and fusion of a large language model. When the expert matching program based on the knowledge extension and fusion of a large language model is executed by a processor, it implements the steps of the expert matching method based on the knowledge extension and fusion of a large language model as described in any one of claims 1 to 6.

Images