Enterprise and scientific achievement intelligent matching method and system based on portrait and semantic vector

Abstract

The invention discloses an enterprise and scientific research achievement intelligent matching method based on portraits and semantic vectors. The method comprises the following steps: firstly, acquiring multi-dimensional data of a target enterprise and an enterprise achievement transformation capability evaluation result, fusing to form an enterprise introduction text, generating potential technical requirement description by a large model, converting the potential technical requirement description into a first semantic vector, and forming an enterprise portrait; secondly, for each scientific and technological achievement in the scientific and technological achievement database, obtaining content description, achievement unit information and research and development personnel information of the scientific and technological achievement, forming a scientific and technological achievement portrait, and converting the content description into a second semantic vector; then traversing each potential technical requirement of the target enterprise, performing scientific and technological achievement matching by semantic vector similarity, and recalling to a preliminary screening achievement library; and finally, based on the enterprise portrait and the scientific and technological achievement portrait, re-evaluating the matching degree through a large model, and sorting the matching degree of the scientific and technological achievements in the preliminary screening achievement library. According to the invention, intelligent matching of enterprise demands and scientific research achievements can be efficiently and accurately realized.

Description

Technical Field

[0001] The present invention belongs to the field of personalized recommendation, and specifically relates to a method and system for intelligently matching enterprises and scientific research results based on portraits and semantic vectors. Background Art

[0002] With the development of science and technology and the increasing demand of enterprises for technological innovation, the matching of enterprises and scientific research results has become an important research direction. In the existing technology, the matching methods of enterprises and scientific research results usually rely on manual or simple keyword matching. However, these methods are difficult to accurately reflect the actual needs of enterprises and the potential value of scientific research results, resulting in low matching accuracy and efficiency. The limitations of existing matching methods are mainly reflected in several aspects: first, the understanding of enterprise needs is not deep enough, and it is difficult to fully consider the actual situation of enterprises in technology research and development, intellectual property rights and other aspects; second, the content of scientific research results cannot be deeply analyzed, resulting in the inability to accurately evaluate its application potential; third, there is a lack of effective intelligent means, and the matching process is inefficient.

[0003] In the field of scientific and technological services such as scientific and technological achievement transformation, traditional matching methods also show similar problems. Traditional methods mostly use text fuzzy matching and exhaustive methods, which cannot understand the contextual semantics of policies and achievements, and often lead to inaccurate and incorrect matching. In addition, existing personalized recommendation technologies, such as recommendation algorithms based on collaborative filtering and content filtering, are widely used in e-commerce, books, patents and other fields. However, in scientific and technological service platforms, there is a problem of data sparsity due to the lack of feedback information such as user ratings and evaluations. The collaborative filtering recommendation algorithm has the problem of low recommendation accuracy, and the traditional content filtering algorithm has limited semantic understanding ability when applied to unstructured text information.

[0004] Therefore, in the field of technology transformation services, how to achieve personalized recommendations for scientific and technological achievements is a technical problem that needs to be solved urgently. Summary of the invention

[0005] The purpose of the present invention is to solve the problems existing in the prior art and to provide a method and system for intelligent matching of enterprises and scientific research results based on portraits and semantic vectors.

[0006] The specific technical solutions adopted by the present invention are as follows:

[0007] In a first aspect, the present invention provides a method for intelligently matching enterprises and scientific research results based on portraits and semantic vectors, which includes:

[0008] S1: Obtain multi-dimensional data of the target enterprise, including enterprise business information, enterprise qualification information, and enterprise R&D information;

[0009] S2: For the target enterprise, an evaluation result of the enterprise's achievement transformation capability is generated based on the multi-dimensional data, and the multi-dimensional data and the evaluation result of the enterprise's achievement transformation capability are integrated to form a piece of enterprise introduction text, and the big model is driven by prompt words to generate several potential technical demand descriptions of the target enterprise based on the enterprise introduction text, and each potential technical demand description is converted into a first semantic vector by text vectorization; at the same time, the potential technical demand generated by the big model is combined with the multi-dimensional data to form a corporate portrait of the target enterprise;

[0010] S3. For each optional scientific and technological achievement in the scientific research achievement database, obtain the content description of the scientific and technological achievement, the achievement unit information and the R&D personnel information to form a scientific and technological achievement portrait, and convert the content description of the scientific and technological achievement into a second semantic vector by text vectorization;

[0011] S4. Traverse each potential technical demand of the target enterprise, use the similarity of the semantic vector to represent the matching degree, match the scientific and technological achievements in the scientific research achievement database, and recall all the potential technical demands obtained by matching to the initial screening achievement database;

[0012] S5. Drive the big model through prompt words to re-evaluate the matching degree between the enterprise portrait and the scientific and technological achievement portrait based on the enterprise portrait of the target enterprise and the scientific and technological achievement portrait of each scientific and technological achievement in the preliminary screening results library, and rank the matching degree of all scientific and technological achievements in the preliminary screening results library according to the matching degree obtained by the re-evaluation.

[0013] As a preferred embodiment of the first aspect, in S1, the multi-dimensional data of the target enterprise includes enterprise business information, enterprise qualification information, enterprise R&D statistical data and enterprise intellectual property data.

[0014] As a preferred embodiment of the first aspect, in S2, the evaluation results of the enterprise's achievement transformation capability include evaluation scores in three dimensions: innovation input, achievement output and operating conditions.

[0015] As a preferred embodiment of the above-mentioned first aspect, the enterprise introduction text is composed of a combination of enterprise profile, business scope, enterprise qualifications and enterprise achievement transformation capability evaluation results.

[0016] As a preferred embodiment of the above-mentioned first aspect, each potential technical requirement description generated by the large model consists of two parts: a technical requirement title and a technical requirement description content.

[0017] As a preferred embodiment of the above-mentioned first aspect, the content description of the scientific and technological achievements consists of two parts: the title of the scientific and technological achievements and the abstract of the content of the scientific and technological achievements.

[0018] As a preferred embodiment of the above-mentioned first aspect, the degree of matching between the potential technical requirements and the scientific and technological achievements is the cosine similarity between the first semantic vector and the second semantic vector corresponding to the two; scientific and technological achievements in the scientific research results database whose degree of matching with any potential technical requirements exceeds a threshold need to be recalled to the preliminary screening results library.

[0019] In a second aspect, the present invention provides an intelligent matching system for enterprises and scientific research results based on portraits and semantic vectors, which includes:

[0020] Enterprise data acquisition module, used to obtain multi-dimensional data of target enterprises, including enterprise business information, enterprise qualification information and enterprise R&D information;

[0021] An enterprise portrait generation module is used to generate an enterprise achievement transformation capability evaluation result for a target enterprise based on the multi-dimensional data, and fuse the multi-dimensional data and the enterprise achievement transformation capability evaluation result to form an enterprise introduction text, drive the big model to generate several potential technical requirement descriptions of the target enterprise based on the enterprise introduction text through prompt words, and convert each potential technical requirement description into a first semantic vector through text vectorization; at the same time, the potential technical requirements generated by the big model are combined with the multi-dimensional data to form an enterprise portrait of the target enterprise;

[0022] The scientific and technological achievement portrait module is used to obtain the content description, achievement unit information and R&D personnel information of each optional scientific and technological achievement in the scientific research achievement database to form a scientific and technological achievement portrait, and convert the content description of the scientific and technological achievement into a second semantic vector by text vectorization;

[0023] The results initial screening and matching module is used to traverse each potential technical demand of the target enterprise, use the similarity of semantic vectors to represent the matching degree, match scientific and technological achievements in the scientific research results database, and recall all the potential technical demands obtained by matching to the initial screening results database;

[0024] The achievement refinement ranking module is used to drive the large model through prompt words to re-evaluate the matching degree between the enterprise portrait and the scientific and technological achievement portrait based on the enterprise portrait of the target enterprise and the scientific and technological achievement portrait of each scientific and technological achievement in the preliminary screening achievement library, and rank the matching degree of all scientific and technological achievements in the preliminary screening achievement library according to the matching degree obtained by the re-evaluation.

[0025] In a third aspect, the present invention provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the method for intelligent matching of enterprises and scientific research results based on portraits and semantic vectors as described in any of the schemes in the first aspect above is implemented.

[0026] In a fourth aspect, the present invention provides a computer electronic device comprising a memory and a processor;

[0027] The memory is used to store computer programs;

[0028] The processor is used to implement the method for intelligent matching of enterprises and scientific research results based on portraits and semantic vectors as described in any of the solutions of the first aspect above when executing the computer program.

[0029] Compared with the prior art, the present invention has the following beneficial effects:

[0030] (1) Improve the accuracy and efficiency of matching: The present invention forms a corporate portrait by acquiring and calculating multi-dimensional data of the enterprise, including industrial and commercial information, the proportion of R&D personnel, the proportion of R&D funds, intellectual property output, etc., and uses a large language model to generate the potential technical R&D needs of the enterprise. At the same time, multi-dimensional data such as the content, unit of achievement and inventor of scientific and technological achievements are collected to generate a portrait of scientific and technological achievements. The semantic vector of the portrait is calculated using a vector model, and the data is first retrieved by calculating the relevance of the vector. Then, the large model is used to precisely sort and score the recall results to match the most relevant content. This method can accurately reflect the actual needs of the enterprise and the potential value of scientific research results, greatly improving the accuracy and efficiency of matching.

[0031] (2) In-depth exploration of the actual value of enterprise needs and scientific research results: The present invention generates the potential technology research and development needs of enterprises through a large language model, which can comprehensively consider the actual situation of enterprises in technology research and development, intellectual property rights, etc. For scientific and technological achievements, the semantic vector of the portrait is calculated using a vector model, which can deeply analyze the content of the scientific research results and accurately evaluate their application potential. This method can deeply explore the actual value of enterprise needs and scientific research results and improve the accuracy of matching results.

[0032] (3) Solving multiple limitations in the prior art: In the prior art, the matching method between enterprises and scientific research results usually relies on manual or simple keyword matching, resulting in low matching accuracy and efficiency. The present invention uses a large language model and a vector model to effectively solve these limitations and improve matching accuracy and efficiency.

[0033] (4) Intelligent matching means to quickly respond to needs: The present invention uses a large model to accurately sort and score the recall results, which can quickly respond to the matching requests of enterprises' technical needs and scientific research results, improve the efficiency of the matching process, and avoid the problem of lack of intelligent means and low efficiency of the matching process in the prior art.

[0034] (5) Improve the performance of recommendation and clustering systems: The present invention has advantages in processing the quantitative representation of complex texts and multimodal feature fusion, and can effectively improve the performance of recommendation and clustering systems. This method improves the quantitative representation ability of complex texts through multi-semantic filtering and multimodal feature representation, and solves the problem of low recommendation and clustering performance of matching systems in the prior art. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] Figure 1 A schematic diagram of the steps of the intelligent matching method between enterprises and scientific research results based on portraits and semantic vectors;

[0036] Figure 2 A schematic diagram for automatically combining and generating the first prompt word;

[0037] Figure 3 A schematic diagram for automatically combining and generating a second prompt word;

[0038] Figure 4 This is a schematic diagram of the module composition of the intelligent matching system between enterprises and scientific research results based on portraits and semantic vectors;

[0039] Figure 5 It is a schematic diagram of the structure of computer electronic equipment. DETAILED DESCRIPTION

[0040] In order to make the above-mentioned purpose, features and advantages of the present invention more obvious and easy to understand, the specific implementation mode of the present invention is described in detail below in conjunction with the accompanying drawings. In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention can be implemented in many other ways different from those described herein, and those skilled in the art can make similar improvements without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below. The technical features in each embodiment of the present invention can be combined accordingly without conflicting with each other.

[0041] In the description of the present invention, it should be understood that the terms "first" and "second" are only used for the purpose of distinguishing descriptions, and should not be understood as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Therefore, the features defined as "first" and "second" may explicitly or implicitly include at least one of the features.

[0042] In order to solve the problem of matching enterprise needs with scientific research results, the present invention proposes a method for intelligently matching enterprises and scientific research results based on portraits and semantic vectors. It forms an enterprise portrait by acquiring and calculating data of various dimensions of the enterprise, and uses a large language model to generate the enterprise's potential technology research and development needs; at the same time, it collects multi-dimensional data such as the content, unit of the achievement, and inventor of the scientific and technological achievements to generate a portrait of the scientific and technological achievements; for two portraits, a vector model is used to calculate the semantic vector of the portrait. In the matching process, the data is first retrieved by calculating the relevance using a vector, and then the recall results are precisely sorted and scored using a large model to match the most relevant content. This method not only improves the accuracy and efficiency of matching, but also can deeply understand the actual value of enterprise needs and scientific research results, solving the problem that the existing technology is difficult to efficiently and accurately achieve intelligent matching between enterprises and scientific research results.

[0043] The specific implementation of the above-mentioned method for intelligent matching of enterprises and scientific research results based on portraits and semantic vectors of the present invention is described in detail below.

[0044] like Figure 1 As shown, in a preferred embodiment of the present invention, a method for intelligently matching enterprises and scientific research results based on portraits and semantic vectors is provided, which includes a total of 5 steps from S1 to S5.

[0045] S1: Obtain multi-dimensional data of the target enterprise, including business information, enterprise qualification information and enterprise R&D information.

[0046] It should be noted that the specific data format and content in the multi-dimensional data can be adjusted according to the actual data source connected, as long as it can cover the business information, enterprise qualification information and enterprise R&D information of the enterprise. The business information of the enterprise in the present invention refers to the business registration information of the enterprise, and the business scope contained therein can reflect the distribution of the business areas of the enterprise. The enterprise qualification information in the present invention refers to some certification qualifications related to R&D and scientific and technological achievements of the enterprise, such as the qualifications of specialized and new small and medium-sized enterprises, the qualifications of single champion enterprises (little giants) in the manufacturing industry, the qualifications of national intellectual property advantage enterprises, and the qualifications of key high-tech enterprises in the National Torch Program. The enterprise R&D information in the present invention refers to some R&D statistical data and intellectual property conditions of the enterprise itself.

[0047] In a preferred embodiment of the present invention, the multi-dimensional data of the target enterprise includes enterprise business information, enterprise qualification information, enterprise R&D statistics data and enterprise intellectual property data. These dimensional data can be collected and obtained in the following ways:

[0048] 1) Obtaining the business information and qualifications of the enterprise:

[0049] Get the business information of the enterprise from the local enterprise database. The specific fields include: enterprise name, unified social credit code, legal representative, registered capital, establishment date, business scope, business status, registered address, business term, registration authority, enterprise type, registration authority, approval date, certificate validity period, shareholder information, etc.

[0050] Obtain the qualification data of the enterprise, whether it is a specialized, innovative and new small and medium-sized enterprise, a single champion enterprise in the manufacturing industry (little giant), a national intellectual property advantage enterprise, a key high-tech enterprise in the National Torch Program, etc.

[0051] 2) Acquisition of enterprise R&D statistical data:

[0052] The number of R&D personnel, their educational background and professional title, the ratio of R&D personnel to the total number of employees, etc.; the amount of the company's R&D investment, the ratio of R&D investment to total revenue, the annual rate of increase in R&D investment, etc.; the situation of the key R&D project team achieved by the company, etc.

[0053] 3) Obtaining enterprise intellectual property data:

[0054] Obtain intellectual property data such as corporate patents, papers, and software copyrights.

[0055] Patents specifically include the following fields: patent name, patent type (such as invention patent, utility model patent, design patent), patent application number / patent number, patent application date, patent authorization date, patent owner, patent status (such as valid, expired, pending), patent legal status (such as authorized, unauthorized, transferred), patent technology field, and patent summary.

[0056] The paper specifically includes the following fields: paper title, paper author (including company employees), publication date, name of publishing journal / conference, journal / conference level (such as SCI, EI, core journals, etc.), paper abstract, research field, and number of citations.

[0057] Software copyright specifically includes the following fields: software name, software version number, copyright registration number, registration date, copyright owner, software type (such as application software, system software, etc.), software function description, software development environment and technical characteristics.

[0058] However, it should be noted that the specific content and form of the above-mentioned various types of data are merely a better exemplary preferred method and can be adjusted according to actual needs.

[0059] S2: For the target enterprise, an evaluation result of the enterprise's ability to transform results is generated based on the multi-dimensional data, and the multi-dimensional data and the evaluation result of the enterprise's ability to transform results are integrated to form a company introduction text. The big model (LLM) is driven by prompt words (Prompt) to generate several potential technical requirement descriptions of the target enterprise based on the company introduction text, and each potential technical requirement description is converted into a first semantic vector by text vectorization; at the same time, the potential technical requirements generated by the big model are combined with the multi-dimensional data to form a corporate portrait of the target enterprise.

[0060] It should be noted that the enterprise introduction text is a fusion of multi-dimensional data and the evaluation results of the enterprise's achievement transformation capability. The specific fusion method can be adjusted according to the actual situation, so as to reflect the multi-dimensional information of the enterprise and the current enterprise achievement transformation capability. Among them, the fused multi-dimensional data can be extracted from the data collected in step S1, and part of the data can be used or all of it can be fused. The enterprise achievement transformation capability evaluation results in the present invention are the results obtained by evaluating the achievement transformation capability of the target enterprise. In practical applications, the evaluation dimensions and indicators of enterprise achievement transformation can be designed according to expert knowledge or prior knowledge, and then the evaluation scores of each dimension are determined by comprehensive scoring.

[0061] In a preferred embodiment of the present invention, the enterprise achievement transformation capability evaluation results can be generated through the enterprise achievement transformation capability evaluation model. The evaluation model uses three-level indicators for scoring, and the first-level indicators are innovation input, achievement output and business conditions. The indicator system and evaluation criteria are as follows:

[0062] Table 1 Index system and evaluation criteria in the enterprise achievement transformation capability evaluation model

[0063]

[0064]

[0065]

[0066] Using the multidimensional data of enterprises collected in S1, we score according to the model in Table 1 above to obtain the scores of all three-level indicators, and then calculate the scores of the secondary indicators and the primary indicators in turn (the score of the secondary indicator is the sum of the scores of the subordinate three-level indicators. If the sum of the scores exceeds the upper limit of the score of the secondary indicator, the upper limit is taken; similarly, the score of the primary indicator is the sum of the scores of the subordinate secondary indicators). Finally, the scores of the three first-level indicators are used as the evaluation results of the enterprise's ability to transform achievements.

[0067] In an embodiment of the present invention, the enterprise introduction text is composed of a combination of the enterprise profile, business scope, enterprise qualifications and the enterprise achievement transformation capability evaluation results. The enterprise profile can be extracted from the industrial and commercial information (or obtained from the official website of the enterprise), and the enterprise achievement transformation capability evaluation results include evaluation scores in three dimensions: innovation input, achievement output and operating conditions.

[0068] In addition, the enterprise introduction text needs to be generated through prompt word engineering and input into the large model (LLM), driving the large model to generate several potential technical requirements descriptions of the target enterprise based on the enterprise introduction text. The specific prompt word design of the large model can be designed according to the general principles of prompt word engineering and continuously optimized through measured data. The present invention does not limit the specific prompt words. For a company introduction text, the number and format of potential technical requirements descriptions output by the large model can be controlled as needed.

[0069] In a preferred embodiment of the present invention, when generating potential technical requirements of an enterprise, a conventional large model (such as GPT-4, etc.) or an industrial network chain large model obtained by fine-tuning the industrial chain data can be called to integrate the company profile (which can be obtained from industrial and commercial information), business scope, enterprise qualifications, and enterprise achievement transformation capability evaluation results into a piece of enterprise introduction text {company_info}, which is merged into the first prompt word (Prompt) template and input into the large model. The large model generates and predicts potential technical requirements. Each potential technical requirement description consists of two parts: the technical requirement title and the technical requirement description content. In an embodiment of the present invention, a first prompt word template can be pre-constructed, and a placeholder for the company introduction text {company_info} can be set in the template to achieve automatic combination and generation of prompt words, such as Figure 2 The first prompt word template used in the embodiment of the present invention is shown in Table 2:

[0070] Table 2 Prompt word template

[0071]

[0072]

[0073] Finally, two examples of potential enterprise needs are generated as shown in Table 3:

[0074] Table 3 Examples of potential enterprise needs

[0075]

[0076]

[0077] In addition, after the large model generates a description of potential technical requirements, each potential technical requirement description can be converted into a first semantic vector by text vectorization. The text vectorization of the present invention can be implemented using existing models, such as BERT, or it can be implemented by calling large models such as GPT. In an embodiment of the present invention, the text generation vector model preferably uses the BGE-Large-zh-v1.5 model open sourced by the Beijing Zhiyuan Artificial Intelligence Research Institute. The model weights can be downloaded from the ModelScope website at: https: / / modelscope.cn / models / BAAI / bge-large-zh-v1.5. Therefore, the first semantic vector can be calculated separately for each potential technical requirement description and stored in the Elasticsearch library.

[0078] In addition, the description of potential technical requirements generated by the large model also needs to be combined with the multi-dimensional data to form a corporate portrait of the target enterprise. In an embodiment of the present invention, the established corporate portrait includes multiple dimensions including corporate R&D direction portrait, corporate transformation capability portrait, corporate R&D team portrait, corporate technological advancement portrait, corporate demand portrait, and corporate basic information portrait. Each dimension needs to set multiple label types to represent the portrait type. In an example, the corporate portrait established for a company located in Hangzhou is shown in Table 4, where each label needs to be assigned according to the actual situation of the company. For example, for the dimension of corporate R&D direction, it includes two labels, namely, the national high-tech enterprise field and the five major industrial ecosystems in Hangzhou. It needs to be assigned according to the national high-tech enterprise field to which the company actually belongs and the five major industrial ecosystems in Hangzhou.

[0079] Table 4 Exemplary corporate portraits

[0080]

[0081] It should be noted that the specific dimensions and labels in the above-mentioned corporate portraits can be reasonably adjusted according to actual needs.

[0082] S3. For each optional scientific and technological achievement in the scientific research achievement database, obtain the content description of the scientific and technological achievement, the achievement unit information and the R&D personnel information to form a scientific and technological achievement portrait, and convert the content description of the scientific and technological achievement into a second semantic vector through text vectorization.

[0083] The content description of different scientific and technological achievements in the present invention may have different fields, but for general achievements, it basically includes two parts: the title of the scientific and technological achievement and the abstract of the scientific and technological achievement content. Moreover, the information of these two parts can generally translate the content of the achievement, so in the embodiment of the present invention, the content description of the scientific and technological achievement required for calculating the second semantic vector can be composed of the title of the scientific and technological achievement and the abstract of the scientific and technological achievement content.

[0084] In a preferred embodiment of the present invention, the collected multi-dimensional data such as the content of scientific and technological achievements, achievement units and inventors mainly include the following aspects:

[0085] Information on the content of scientific and technological achievements: name of the achievement, type of achievement (such as new products, new technologies, new processes, new materials, etc.), field to which the achievement belongs (such as electronic information, biomedicine, new energy, environmental protection technology, etc.), brief introduction to the achievement (including technical principles, innovative points, application prospects, etc.), technical level of the achievement (such as internationally leading, internationally advanced, domestically leading, domestically advanced, etc.), maturity of the achievement (such as research and development stage, trial production stage, industrialization stage, etc.).

[0086] Information on the unit of the achievement: unit name, nature of the unit (such as enterprise, research institute, university, etc.), unit address, unit contact information, unit profile (including research fields, R&D capabilities, main achievements, etc.).

[0087] R&D personnel / team information: name of the R&D personnel (e.g., inventor of a patent), position / title of the R&D personnel, department / team to which the R&D personnel belongs, description of the R&D personnel’s contribution (e.g., proposing innovative ideas, being responsible for technology research and development, etc.), and contact information of the R&D personnel.

[0088] Therefore, based on the above information, the scientific and technological achievement portrait can be composed of the achievement title, achievement summary, and achievement unit. At the same time, for the achievement content information composed of the achievement title and achievement summary in the achievement portrait, the above BGE-Large-zh-v1.5 model can also be used to calculate the vector and store it in the Elasticsearch library.

[0089] S4. Traverse each potential technical demand of the target enterprise, use the similarity of the semantic vector to represent the matching degree, match the scientific and technological achievements in the scientific research achievement database, and recall all the potential technical demands obtained by matching to the initial screening achievement database.

[0090] It should be noted that the similarity of semantic vectors can be calculated in a variety of ways. In the present invention, the matching degree between potential technical requirements and scientific and technological achievements can be calculated by the cosine similarity between the first semantic vector and the second semantic vector corresponding to the two. The calculation formula for calculating the semantic similarity between the enterprise portrait and the scientific and technological achievement portrait using cosine similarity is as follows:

[0091] cosine_similarity(A,B)=(A dot B) / (||A||*||B||)

[0092] Where: A and B are two vectors to be compared; dot represents the dot product of the vectors, also known as the inner product, which is the sum of the products of corresponding elements of the two vectors; ||A|| and ||B|| represent the modulus (or length) of the vector, which can be obtained by calculating the sum of the squares of each vector element and then taking the square root.

[0093] The value range of the above cosine similarity cosine_similarity(A,B) is between -1 and 1:

[0094] When the cosine similarity is equal to 1, it means that the two vectors have exactly the same direction in the multidimensional space, that is, they are completely similar.

[0095] When the cosine similarity is equal to 0, it means that there is no linear relationship between the two vectors and they are unrelated.

[0096] When the cosine similarity is equal to -1, it means that the directions of the two vectors in the multidimensional space are exactly opposite, that is, they are completely different.

[0097] Based on this cosine similarity, the degree of match between each potential technical requirement and all optional scientific and technological achievements in the database can be evaluated. In practical applications, a similarity threshold can be set, and scientific and technological achievements in the scientific research achievement database that match any potential technical requirement more than the threshold can be recalled to the preliminary screening results library. Of course, if there are too many scientific and technological achievements that match more than the threshold, a maximum recall number can also be set, and then some of the scientific and technological achievements with the highest matching degree can be recalled.

[0098] S5. Drive the big model through prompt words to re-evaluate the matching degree between the enterprise portrait and the scientific and technological achievement portrait based on the enterprise portrait of the target enterprise and the scientific and technological achievement portrait of each scientific and technological achievement in the preliminary screening results library, and rank the matching degree of all scientific and technological achievements in the preliminary screening results library according to the matching degree obtained by the re-evaluation.

[0099] It should be noted that step S5 is equivalent to further fine-ranking the recalled scientific and technological achievements using the reasoning ability of the big model, so that they can better match the actual needs of the enterprise. In an embodiment of the present invention, all the fields of the scientific and technological achievement portrait {achievement_portrait} and the corporate portrait {company_portrait} in the preliminary screening results library recalled by vector similarity can be input into the big model, allowing the big model to determine the degree of match between the two portraits and give a score. This process also requires the construction of a corresponding prompt template through a prompt word project. In an embodiment of the present invention, a second prompt word template can be constructed in advance, and placeholders for the scientific and technological achievement portrait {achievement_portrait} and the corporate portrait {company_portrait} can be set in the template to achieve automatic combination and generation of prompt words, such as Figure 3 The second prompt word template used in the embodiment of the present invention is shown in Table 5:

[0100] Table 5 Prompt word template

[0101]

[0102]

[0103] Finally, the most relevant matching results after large-scale model analysis can be output as scientific and technological achievements recommended to enterprises, or all recalled scientific and technological achievements can be re-ranked according to the refined ranking results and recommended to enterprises in order.

[0104] It should be noted that the method steps shown in S1 to S5 above can essentially be implemented in the form of a computer program.

[0105] like Figure 4 As shown, based on the same inventive concept, the present invention also provides a system for intelligently matching enterprises and scientific research achievements based on portraits and semantic vectors, corresponding to the method for intelligently matching enterprises and scientific research achievements based on portraits and semantic vectors provided in the above embodiment, which includes:

[0106] Enterprise data acquisition module, used to obtain multi-dimensional data of target enterprises, including enterprise business information, enterprise qualification information and enterprise R&D information;

[0107] An enterprise portrait generation module is used to generate an enterprise achievement transformation capability evaluation result for a target enterprise based on the multi-dimensional data, and fuse the multi-dimensional data and the enterprise achievement transformation capability evaluation result to form an enterprise introduction text, drive the big model to generate several potential technical requirement descriptions of the target enterprise based on the enterprise introduction text through prompt words, and convert each potential technical requirement description into a first semantic vector through text vectorization; at the same time, the potential technical requirements generated by the big model are combined with the multi-dimensional data to form an enterprise portrait of the target enterprise;

[0108] The scientific and technological achievement portrait module is used to obtain the content description, achievement unit information and R&D personnel information of each optional scientific and technological achievement in the scientific research achievement database to form a scientific and technological achievement portrait, and convert the content description of the scientific and technological achievement into a second semantic vector by text vectorization;

[0109] The results initial screening and matching module is used to traverse each potential technical demand of the target enterprise, use the similarity of semantic vectors to represent the matching degree, match scientific and technological achievements in the scientific research results database, and recall all the potential technical demands obtained by matching to the initial screening results database;

[0110] The achievement refinement ranking module is used to drive the large model through prompt words to re-evaluate the matching degree between the enterprise portrait and the scientific and technological achievement portrait based on the enterprise portrait of the target enterprise and the scientific and technological achievement portrait of each scientific and technological achievement in the preliminary screening achievement library, and rank the matching degree of all scientific and technological achievements in the preliminary screening achievement library according to the matching degree obtained by the re-evaluation.

[0111] In addition, if Figure 5 As shown, based on the same inventive concept, the present invention also provides a computer electronic device corresponding to the method for intelligently matching enterprises and scientific research results based on portraits and semantic vectors provided in the above embodiment, which includes a memory and a processor;

[0112] The memory is used to store computer programs;

[0113] The processor is used to implement the above-mentioned intelligent matching method between enterprises and scientific research results based on portraits and semantic vectors when executing the computer program;

[0114] In addition, the logic instructions in the above-mentioned memory can be implemented in the form of software functional units and can be stored in a computer-readable storage medium when sold or used as an independent product. Based on such an understanding, the technical solution of the present invention, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product, which is stored in a storage medium and includes several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention.

[0115] In addition, based on the same inventive concept, the present invention provides a computer-readable storage medium corresponding to the method for intelligent matching of enterprises and scientific research results based on portraits and semantic vectors, and a computer program is stored on the storage medium. When the computer program is executed by the processor, it can implement the method for intelligent matching of enterprises and scientific research results based on portraits and semantic vectors as described above.

[0116] In addition, based on the same inventive concept, the present invention provides a computer program product, including a computer program / instruction, which, when executed by a processor, can implement the intelligent matching method between enterprises and scientific research results based on portraits and semantic vectors as described above.

[0117] Specifically, in the computer-readable storage medium of each of the above embodiments, the stored computer program is executed by the processor to perform the above steps S1 to S5.

[0118] It is understandable that the above storage medium may include random access memory (RAM) or non-volatile memory (NVM), such as at least one disk storage. The storage medium may also be a U disk, a mobile hard disk, a magnetic disk or an optical disk, etc., which can store program codes.

[0119] It can be understood that the above-mentioned processor can be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it can also be a digital signal processor (DSP), an application specific integrated circuit (ASIC), a field programmable gate array (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components.

[0120] It should also be noted that those skilled in the art can clearly understand that for the convenience and simplicity of description, the specific working process of the system described above can refer to the corresponding process in the aforementioned method embodiment, and will not be repeated here. In the various embodiments provided in this application, the division of steps or modules in the system and method is only a logical function division, and there may be other division methods in actual implementation, such as multiple modules or steps can be combined or integrated together, and a module or step can also be split.

[0121] The above-described embodiments are only some preferred implementations of the present invention, but are not intended to limit the present invention. A person skilled in the relevant technical field may make various changes and modifications without departing from the spirit and scope of the present invention. Therefore, any technical solution obtained by equivalent replacement or equivalent transformation falls within the protection scope of the present invention.

Claims

1. A method for intelligently matching enterprises and scientific research results based on portraits and semantic vectors, characterized in that: include: S1: Obtain multi-dimensional data of the target enterprise, including enterprise business information, enterprise qualification information, and enterprise R&D information; S2: For the target enterprise, an evaluation result of the enterprise's achievement transformation capability is generated based on the multi-dimensional data, and the multi-dimensional data and the evaluation result of the enterprise's achievement transformation capability are integrated to form a piece of enterprise introduction text, and the big model is driven by prompt words to generate several potential technical demand descriptions of the target enterprise based on the enterprise introduction text, and each potential technical demand description is converted into a first semantic vector by text vectorization; at the same time, the potential technical demand generated by the big model is combined with the multi-dimensional data to form a corporate portrait of the target enterprise; S3. For each optional scientific and technological achievement in the scientific research achievement database, obtain the content description of the scientific and technological achievement, the achievement unit information and the R&D personnel information to form a scientific and technological achievement portrait, and convert the content description of the scientific and technological achievement into a second semantic vector by text vectorization; S4. Traverse each potential technical demand of the target enterprise, use the similarity of the semantic vector to represent the matching degree, match the scientific and technological achievements in the scientific research achievement database, and recall all the potential technical demands obtained by matching to the initial screening achievement database; S5. Drive the big model through prompt words to re-evaluate the matching degree between the enterprise portrait and the scientific and technological achievement portrait based on the enterprise portrait of the target enterprise and the scientific and technological achievement portrait of each scientific and technological achievement in the preliminary screening results library, and rank the matching degree of all scientific and technological achievements in the preliminary screening results library according to the matching degree obtained by the re-evaluation.

2. The method for intelligently matching enterprises and scientific research results based on portraits and semantic vectors as claimed in claim 1, characterized in that: In S1, the multi-dimensional data of the target enterprise includes enterprise business information, enterprise qualification information, enterprise R&D statistics data and enterprise intellectual property data.

3. The method for intelligently matching enterprises and scientific research results based on portraits and semantic vectors as claimed in claim 1, characterized in that: In S2, the evaluation results of the enterprise's achievement transformation capability include evaluation scores in three dimensions: innovation input, achievement output and operating conditions.

4. The method for intelligently matching enterprises and scientific research results based on portraits and semantic vectors as claimed in claim 1, characterized in that: The enterprise introduction text is composed of the enterprise profile, business scope, enterprise qualifications and the evaluation results of the enterprise's ability to transform achievements.

5. The method for intelligently matching enterprises and scientific research results based on portraits and semantic vectors according to claim 1, characterized in that: Each potential technical requirement description generated by the big model consists of two parts: the technical requirement title and the technical requirement description content.

6. The method for intelligently matching enterprises and scientific research results based on portraits and semantic vectors according to claim 1, characterized in that: The content description of the scientific and technological achievements consists of two parts: the title of the scientific and technological achievements and the summary of the scientific and technological achievements.

7. The method for intelligently matching enterprises and scientific research results based on portraits and semantic vectors according to claim 1, characterized in that: The degree of matching between the potential technical requirements and the scientific and technological achievements is the cosine similarity between the first semantic vector and the second semantic vector corresponding to the two; scientific and technological achievements in the scientific research achievement database whose degree of matching with any potential technical requirement exceeds the threshold need to be recalled to the preliminary screening achievement library.

8. An intelligent matching system for enterprises and scientific research results based on portraits and semantic vectors, characterized in that: include: Enterprise data acquisition module, used to obtain multi-dimensional data of target enterprises, including enterprise business information, enterprise qualification information and enterprise R&D information; An enterprise portrait generation module is used to generate an enterprise achievement transformation capability evaluation result for a target enterprise based on the multi-dimensional data, and fuse the multi-dimensional data and the enterprise achievement transformation capability evaluation result to form an enterprise introduction text, drive the big model to generate several potential technical requirement descriptions of the target enterprise based on the enterprise introduction text through prompt words, and convert each potential technical requirement description into a first semantic vector through text vectorization; at the same time, the potential technical requirements generated by the big model are combined with the multi-dimensional data to form an enterprise portrait of the target enterprise; The scientific and technological achievement portrait module is used to obtain the content description, achievement unit information and R&D personnel information of each optional scientific and technological achievement in the scientific research achievement database to form a scientific and technological achievement portrait, and convert the content description of the scientific and technological achievement into a second semantic vector by text vectorization; The results initial screening and matching module is used to traverse each potential technical demand of the target enterprise, use the similarity of semantic vectors to represent the matching degree, match scientific and technological achievements in the scientific research results database, and recall all the potential technical demands obtained by matching to the initial screening results database; The achievement refinement ranking module is used to drive the large model through prompt words to re-evaluate the matching degree between the enterprise portrait and the scientific and technological achievement portrait based on the enterprise portrait of the target enterprise and the scientific and technological achievement portrait of each scientific and technological achievement in the preliminary screening achievement library, and rank the matching degree of all scientific and technological achievements in the preliminary screening achievement library according to the matching degree obtained by the re-evaluation.

9. A computer-readable storage medium, characterized in that: The storage medium stores a computer program, and when the computer program is executed by the processor, the method for intelligent matching of enterprises and scientific research results based on portraits and semantic vectors as described in any one of claims 1 to 7 is implemented.

10. A computer electronic device, characterized in that: including memory and processor; The memory is used to store computer programs; The processor is used to implement the intelligent matching method between enterprises and scientific research results based on portraits and semantic vectors as described in any one of claims 1 to 7 when executing the computer program.

Images