Scientific and technological achievement and enterprise technology demand matching method and system

Abstract

The invention discloses a scientific and technological achievement and enterprise technology demand matching method and system, and relates to the field of data mining, and the method comprises the steps: constructing scientific and technological achievement original data into scientific and technological achievement structured information, and forming a first vector based on the scientific and technological achievement structured information; carrying out association storage on the scientific and technological achievement structured information and the corresponding first vector, and constructing a scientific and technological achievement library; constructing the enterprise demand original data into enterprise demand structured information, and forming a second vector based on the enterprise demand structured information; carrying out association storage on the enterprise demand structured information and the corresponding second vector, and constructing an enterprise demand library; and rewriting a user request into a retrieval request containing a structured cue word by utilizing a large language model, respectively executing vector recall and full-text retrieval recall, and fusing output results of the vector recall and the full-text retrieval recall to obtain a matching result set. The invention constructs an intelligent matching method for scientific and technological achievements and enterprise technology requirements with multivariate heterogeneous data processing, semantic understanding and cross-scene alignment capabilities.

Description

Technical Field

[0001] The present invention relates to the field of data processing, and in particular to a method and system for matching scientific and technological achievements with enterprise technical requirements. Background Art

[0002] In the field of scientific and technological achievement transformation, large-scale model technology provides a solution for breaking down the barriers between scientific research and industry. Its development background is closely related to the strategic need to deepen collaborative innovation between industry, academia, and research institutes.

[0003] From the perspective of enterprise development, facing increasingly fierce technological competition and ever-changing market demands, companies are no longer content with the traditional passive response model when encountering technical bottlenecks. Instead, they are increasingly inclined to proactively build a system of technical solutions to maintain a sustainable competitive advantage. Especially in the current environment where the speed of technological iteration is increasing exponentially, the ability to identify and match enterprise technological needs has become one of the factors that determine their innovation capabilities and market adaptability.

[0004] However, companies currently face numerous challenges in seeking external technical support. For one thing, many scientific and technological achievement databases still rely on users entering precise keywords for retrieval, lacking the semantic understanding and correlation analysis capabilities for complex technical requirements. This results in low information matching efficiency and accuracy. Furthermore, while some technical forums or academic exchange platforms can tap into potential expert resources through social networks, this process is often time-consuming and costly, and is limited by the breadth and depth of individual relationship networks, making it difficult to systematically and scalably meet the diverse technological demands of companies.

[0005] Therefore, it is urgent to build an intelligent matching platform between scientific and technological achievements and enterprise technology needs driven by big models, and use big models and other technologies to achieve accurate analysis of enterprise technology needs and efficient recommendation of scientific and technological achievements, thereby improving the efficiency of scientific and technological achievement transformation and helping enterprises achieve innovation-driven development.

[0006] There are several key technical bottlenecks in the current process of matching scientific and technological achievements with enterprise technology needs, which are mainly reflected in the following aspects: (1) Governance challenges of multi-source heterogeneous scientific and technological achievement data: The data sources of scientific and technological achievements are extensive and diverse, covering a variety of types such as patent texts, scientific research papers, business plans, project application materials, and experimental reports. These data contain both structured information (such as tables and fielded metadata) and a large amount of semi-structured (such as XML and JSON formats) and unstructured data (such as natural language descriptions and charts). This multi-source heterogeneous feature poses significant challenges to the unified collection, cleaning, standardization, and knowledge extraction of scientific and technological achievement databases, and seriously restricts the efficiency and quality of subsequent intelligent matching.

[0007] (2) The accuracy of label system matching is limited by the ability to express semantics: Currently, most platforms use a matching mechanism based on keywords or traditional tags, but this approach has obvious limitations when facing complex technical requirements. On the one hand, label systems often find it difficult to fully cover the multi-dimensional semantic features between scientific and technological achievements and corporate needs; on the other hand, conventional label matching is prone to the problem of "semantic collapse", that is, it is unable to effectively distinguish technical concepts with similar surface vocabulary but large semantic differences. Therefore, existing systems usually need to rely on manual intervention to correct and adjust labels, which is not only inefficient but also difficult to achieve large-scale application.

[0008] (3) Matching misalignment caused by cross-scenario semantic alignment: When enterprises propose technical requirements, they usually describe them based on their actual business scenarios, emphasizing practical factors such as functional implementation, performance indicators, and engineering feasibility; while scientific and technological achievements focus more on theoretical or laboratory-level expressions such as technical principles, innovations, and experimental verification. This difference in expression perspective leads to semantic drift or semantic misalignment when using mechanisms such as text matching and vector matching for technical matching, making it difficult to achieve accurate semantic alignment between scientific and technological achievements and enterprise needs, thereby affecting the relevance and practicality of recommendation results. Summary of the Invention

[0009] This application addresses at least one shortcoming of the prior art and provides a method for matching scientific and technological achievements with enterprise technology needs, as follows: A method for matching scientific and technological achievements with enterprise technology needs includes the following steps: The original data of scientific and technological achievements are constructed into scientific and technological achievement structured information, and a first vector is formed based on the scientific and technological achievement structured information; the scientific and technological achievement structured information and its corresponding first vector are associated and stored to construct a scientific and technological achievement database; The original enterprise demand data is constructed into enterprise demand structured information, and a second vector is formed based on the enterprise demand structured information; the enterprise demand structured information and its corresponding second vector are associated and stored to construct an enterprise demand library; In response to a user request, a large language model is used to rewrite the user request into a retrieval request containing structured prompt words. Vector recall and full-text retrieval recall are performed based on the retrieval request respectively. The output results of the vector recall and full-text retrieval recall are fused to obtain a matching result set. A pre-trained sorting model is called, and the user request and the recall results in the corresponding matching result set are used as input. The local matching relationship and global semantic association between the user request and the recall results are analyzed, and a score representing the degree of matching between the user request and the recall results is output. The results are sorted according to the score to obtain a second sorted list, and the matching results are displayed based on the second sorted list.

[0010] Preferably, the full-text search and recall method includes: Perform keyword matching search on all enterprise demand structured information in the enterprise demand library based on the rewritten search request, and / or Based on the rewritten search request, a keyword matching search is performed in all the structured information of scientific and technological achievements in the scientific and technological achievements database.

[0011] Preferably, the vector recall method includes: Call the vector model to calculate the search request rewritten by the large language model to obtain the query vector; Using a vector similarity algorithm, we search the vector space of the enterprise demand database with the query vector, calculate the vector dot product, and recall the top-order results that are most semantically similar to the query vector. A vector similarity algorithm is used to search the vector space of the scientific and technological achievements database with the query vector, calculate the vector dot product, and recall several pre-sequence results that are semantically closest to the query vector.

[0012] Preferably, before displaying the matching results, the method further includes filtering the recall results below a threshold in the second sorted list to obtain a third sorted list, and displaying the matching results based on the third sorted list.

[0013] Preferably, before displaying the matching results, the following steps are also included: language model analysis and scoring: Call the large language model, input the third ranked list, construct prompt words to analyze the correlation between the user request and the matching results, output the final matching score, and display the matching results based on the final matching score; Among them, the correlation analysis includes the correlation between user requests and matching results based on production technology, application fields, and materials.

[0014] Preferably, the structured information of scientific and technological achievements also includes the following data fields: source institution / unit, output time, personnel, maturity and first details; The enterprise demand structured information also includes the following data fields: enterprise name and collection / generation time.

[0015] Preferably, the process also includes data deduplication and merging steps. The structured information of scientific and technological achievements is compared based on the source organization / unit, output time, personnel and technical field, as well as the semantic analysis and comparison of the first title and abstract, to generate a unique identifier of a multi-field combination, and duplicate records are identified and merged based on the unique identifier of the multi-field combination.

[0016] Preferably, the scientific and technological achievement database stores scientific and technological achievement data and scientific and technological achievement data groups, and the scientific and technological achievement data groups include a grouping mark and one or more scientific and technological achievement data associated with the grouping mark.

[0017] Preferably, the method for generating a matching result set further includes fusing the output results of the vector recall and the full-text search recall and combining them with the output time of the scientific and technological achievements or the generation time of the enterprise needs to obtain the matching result set.

[0018] Preferably, the method for generating a matching result set further includes executing vector recall and full-text search recall based on the search request to respectively recall scientific and technological achievements and enterprise needs, fusing the scientific and technological achievements recalled by the vector recall and the full-text search, and fusing the enterprise needs recalled by the vector recall and the full-text search; According to the output time of scientific and technological achievements and the generation time of enterprise needs, combined with the maturity of scientific and technological achievements, the integrated scientific and technological achievements and enterprise needs are sorted to obtain a set of matching results.

[0019] The present invention also proposes a system for matching scientific and technological achievements with enterprise technical requirements, including the following structure: A scientific and technological achievement database construction unit: constructs the original scientific and technological achievement data into scientific and technological achievement structured information, forms a first vector based on the scientific and technological achievement structured information; associates and stores the scientific and technological achievement structured information with the corresponding first vector, and constructs a scientific and technological achievement database; Enterprise demand library construction unit: constructs the original enterprise demand data into enterprise demand structured information, forms a second vector based on the enterprise demand structured information; associates and stores the enterprise demand structured information with its corresponding second vector to construct the enterprise demand library; Matching unit: In response to user requests, the large language model is used to rewrite the user request into a retrieval request containing structured prompt words, and vector recall and full-text retrieval recall are performed based on the retrieval request respectively. The output results of vector recall and full-text retrieval recall are integrated to obtain a matching result set, and the pre-trained sorting model is called. The user request and the recall results in the corresponding matching result set are used as input to analyze the local matching relationship and global semantic association between the user request and the recall results, and output a score indicating the degree of matching between the user request and the recall results. The results are sorted according to the score to obtain a sorted list, and the matching results are displayed based on the sorted list.

[0020] The beneficial effects of the present invention are as follows: a method and system for matching scientific and technological achievements with enterprise technology needs based on a large language model has been established, which has the capabilities of multi-dimensional heterogeneous data processing, semantic understanding, and cross-scenario alignment. For those who propose technology needs, the natural language interaction form breaks through the format limitations of traditional technology retrieval, supports enterprises to describe fuzzy needs in natural language, and rewrites retrieval requests containing structured prompt words using a large language model. The intelligent analysis replaces manual retrieval, which can automatically complete the analysis process that traditional technology brokers spend weeks on. The multimodal semantic understanding and dynamic vector matching technology significantly improves the efficiency of matching enterprise technology needs with scientific research results, and its application effect demonstrates its value in the technology innovation ecosystem. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative labor.

[0022] Figure 1 It is a flow chart of the method for matching scientific and technological achievements with enterprise technology needs; Figure 2 It is a flow chart for constructing the scientific and technological achievements database and the enterprise demand database; Figure 3 This is a flow chart of the multi-stage matching mechanism disclosed in this embodiment. DETAILED DESCRIPTION

[0023] The following will be combined with the accompanying drawings to clearly and completely describe the technical solutions in the embodiments of this application. Obviously, the embodiments described are only part of the embodiments of this application, not all of them. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without making creative efforts are within the scope of protection of this application.

[0024] The terms "first", "second", etc. in the claims and specification of this application are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the terms used in this way can be interchangeable under appropriate circumstances. This is merely a way of distinguishing when describing objects with the same properties in the embodiments of this application. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, so that a process, method, system, product or apparatus that includes a series of units is not necessarily limited to those units, but may include other units not expressly listed or inherent to these processes, methods, products or apparatuses.

[0025] Example 1: A method for matching scientific and technological achievements with enterprise technology needs, such as Figures 1 to 3 , including the following steps: Step S1: constructing scientific and technological achievement original data into scientific and technological achievement structured information, wherein the scientific and technological achievement structured information includes a first title and an abstract, and establishing a full-text index for each of the first title and the abstract; Step S2: splicing the first title and the corresponding abstract into a first text content, and converting the first text content into a first vector; associating and storing the scientific and technological achievement structured information with the corresponding first vector, and constructing a scientific and technological achievement database; Step S3: constructing the original enterprise demand data into enterprise demand structured information, the enterprise demand structured information including the second title and the second detail, and establishing full-text indexes for the second title and the second detail respectively; Step S4: concatenate the second title and the corresponding second details into second text content, and convert the second text content into a second vector; associate and store the enterprise demand structured information with the corresponding second vector to construct an enterprise demand library; Step S5: In response to the user request, the large language model is used to rewrite the user request into a retrieval request containing structured prompt words, and vector recall and full-text retrieval recall are respectively performed based on the retrieval request. The output results of the vector recall and full-text retrieval recall are fused to obtain a matching result set, and the pre-trained sorting model is called. The user request and the recall results in the corresponding matching result set are used as input, and the local matching relationship and global semantic association between the user request and the recall results are analyzed. A score representing the degree of matching between the user request and the recall result is output, and the results are sorted according to the score to obtain a second sorted list. The matching results are displayed based on the second sorted list.

[0026] As an example disclosure, the use of a large language model to rewrite user requests mentioned in this invention can employ the conversational model DeepSeek-V3 and the inference model DeppSeek-R1. DeepSeek-V3 is primarily used for tasks such as identifying user intent and rewriting user requests; DeepSeek-R1 is primarily used for tasks such as analyzing the correlation between matching results and user intent and scoring the matching results.

[0027] When matching scientific and technological achievements with enterprise technical needs, user-entered descriptions of technical requirements or achievements often contain issues such as ambiguity, incomplete keywords, and non-standard terminology, which can affect the accuracy of subsequent retrieval and matching. To address this, the system uses a large language model to semantically understand and optimize the structure of original user requests, rewriting them into search requests containing structured prompts to improve matching accuracy.

[0028] The DeepSeek-V3 large language model is used as the core semantic understanding engine. By writing structured prompt words, the model is guided to complete the standardized rewriting and semantic enhancement of the user's original user request.

[0029] For example, the interactive prompts for the large model are as follows: Role You are a professional technical requirements analysis assistant; Task Please complete the following tasks based on the original technical requirements or scientific and technological achievements descriptions entered by the user below: 1. Understand user intent and identify key elements; 2. Provide semantic clarification and supplement for unclear or ambiguous content; 3. Rewrite the requirement using more standardized and technical terms; 4. Make the sentence more fluent, easier to understand and search while maintaining the original meaning; User input: Please output the rewritten description.

[0030] Finally, by semantically understanding and structurally optimizing the original user request, it is rewritten into a search request containing structured prompt words. The structured prompt words for enterprise needs can include the problems to be solved, performance indicators, application scenarios, company names, etc. The structured prompt words for scientific and technological achievements can include the technical content, innovation points and application value, output time, personnel information, maturity, etc. of the scientific and technological achievements.

[0031] Specifically, the method of performing vector recall and full-text search recall based on the search request includes: Based on the search request, vector recall and full-text search are performed to respectively recall a plurality of preceding results, thereby obtaining a first set of candidate scientific and technological achievements and a second set of candidate scientific and technological achievements, and / or a first set of candidate enterprise requirements and a second set of candidate enterprise requirements. The first set of candidate scientific and technological achievements and the second set of candidate scientific and technological achievements are fused to obtain a set of matching results, and / or the first set of candidate enterprise requirements and the second set of candidate enterprise requirements are fused to obtain a set of matching results.

[0032] As a matching method ultimately applied to a scientific and technological achievement / enterprise technology demand matching platform, relevant options are configured. If the user request is to recall the corresponding scientific and technological achievements, a list of scientific and technological achievements is output according to the scientific and technological achievement and enterprise technology demand matching method, which are named the first scientific and technological achievement candidate set and the second scientific and technological achievement candidate set in this preferred solution.

[0033] If the user's request is to recall the corresponding enterprise technology needs, the corresponding enterprise information and enterprise technology needs content will be output according to the matching method between scientific and technological achievements and enterprise technology needs.

[0034] The corresponding matching result set is finally output and displayed on the platform page according to the matching result set.

[0035] Others recall both scientific and technological achievements and enterprise technology needs based on user requests. For example, searching for "environmental pollution control" will display a list of both achievement recalls and demand recalls.

[0036] For example, full-text search is based on the full-text index of enterprise needs and scientific and technological achievements built on Elasticsearch. Using the rewritten search request, keyword matching is performed. Through mechanisms such as word segmentation, synonym expansion, and Boolean combination queries, the database can quickly filter out candidate results that are highly relevant to the user's needs at the lexical level. Because the rewritten search request includes structured prompts, keyword matching is performed. Through mechanisms such as word segmentation, synonym expansion, and Boolean combination queries, the database can quickly filter out candidate results that are highly relevant to the user's needs at the lexical level.

[0037] For example, vector recall first calls the vector model BGE-Large-zh-v1.5 to calculate the vector of the rewritten search request. Since the rewritten search request has structured prompt words, the query vector is obtained based on the structured prompt words. Then, the vector similarity algorithm provided by Elasticsearch is used to search in the vector space, calculate the vector dot product (Dot Product), and recall the results with the closest semantics.

[0038] As a preferred solution, a multi-stage matching mechanism is disclosed, which specifically also includes a secondary sorting method, and the secondary sorting method includes: Before displaying the matching results, step 6 may also be included: filtering the recalled results below a threshold in the second sorted list to obtain a third sorted list, and displaying the matching results based on the third sorted list.

[0039] For example, the pre-trained ranking model can use the BGE-Reranker-v2-M3 model, combining semantic similarity and contextual features to re-score and re-rank recall results. For example, given the input (user request, recall result title + summary), the output is a score representing the degree of match, used to determine the relevance between the recall result and the user request. For example, the relevance score can be output as a decimal between 0 and 1, with a set relevance threshold.

[0040] More preferably, the multi-stage matching mechanism further includes step 7: large language model analysis and scoring before displaying the matching results, specifically: Call the large language model, input the third ranked list, construct prompt words to analyze the correlation between the user request and the matching results, output the final matching score, and display the matching results based on the final matching score; Among them, the correlation analysis includes the correlation between user requests and matching results based on production technology, application fields, and materials.

[0041] More specifically, the correlation score is calculated and the correlation is graded according to the score range, such as high correlation: [80, 100], medium-high correlation: [60, 80], and medium correlation: [40, 60].

[0042] Conventional label matching is prone to the "semantic collapse" problem, meaning it's unable to effectively distinguish technical concepts that share similar surface vocabulary but significant semantic differences. For example, in the energy storage materials sector, a company's technical requirements may encompass multiple dimensions, including "material modification," "application scenario adaptation," and "cost control." Traditional single labels or shallow classification methods struggle to capture these complex implicit semantic associations.

[0043] Specifically, the full-text search and recall method includes: Perform keyword matching search on all enterprise demand structured information in the enterprise demand library based on the rewritten search request, and / or Based on the rewritten search request, a keyword matching search is performed in all the structured information of scientific and technological achievements in the scientific and technological achievements database.

[0044] Specifically, the vector recall method includes: Call the vector model to calculate the search request rewritten by the large language model to obtain the query vector; Using a vector similarity algorithm, we search the vector space of the enterprise demand database with the query vector, calculate the vector dot product, and recall the top-order results that are most semantically similar to the query vector. Using a vector similarity algorithm, we search the vector space of the scientific and technological achievements database using the query vector, calculate the vector dot product, and recall the top-order results that are most semantically similar to the query vector. We use a combination of full-text search and vector recall, using keyword matching and semantic similarity matching strategies to leverage their respective strengths in accuracy and generalization.

[0045] Exemplarily, the vector model uses the BGE-Large-zh-v1.5 model as the core vectorization engine. The present invention applies BGE-Large-zh-v1.5 to the semantic encoding of key fields such as the title, abstract, technical description of scientific and technological achievements, as well as the description of enterprise technical requirements, to generate high-quality dense vector representations, ensuring the semantic alignment accuracy between scientific and technological achievements and enterprise requirements.

[0046] For example, to enhance the integrity and contextual coherence of the semantic representation of scientific and technological achievements, the first title and corresponding abstract are concatenated into a complete textual content. This is used as the semantic input to the vector model BGE-Large-zh-v1.5, generating a 1024-dimensional dense vector representation to characterize the core semantic features of scientific and technological achievements. The structured information of each scientific and technological achievement is associated and stored with its corresponding vector representation, building a unified scientific and technological achievement database to provide efficient support for matching scientific and technological achievements with enterprise technology needs.

[0047] For the original data required by enterprises, it is necessary to perform scientific and technological achievement data collection and management, including data acquisition, cleaning and management, field extraction, structured storage, data deduplication and merging. As an innovative improvement and optimization of the present invention, the steps of field extraction, structured storage, and data deduplication and merging are further described as follows: The structured information of scientific and technological achievements also includes the following data fields: source organization / unit, output time, personnel, maturity and first details.

[0048] The title includes: the name of the scientific and technological achievement or the project title; the abstract includes: a brief description of the technical content, innovations, and application value of the scientific and technological achievement; the source institution / unit includes: the main output unit of the achievement; the output time includes: the time of publication, completion, or authorization of the achievement; the personnel includes: a list of the main researchers or inventors; the maturity includes: the technical maturity level; the first details include: a detailed technical description of the achievement, a link to an attachment, or an extended information portal. As shown in Table 1 below: Table 1: Scientific and technological achievements data model table

[0049] The enterprise demand structured information includes the following data fields: second title, second details, enterprise name, and collection / generation time.

[0050] Specifically, as shown in Table 2 below, the second heading includes the brief name of the enterprise's technology needs; The second details include a specific description of the technical requirements, including detailed information such as the problem to be solved, performance indicators, application scenarios, etc.; the company name includes the name of the company or unit with the technical requirements; the collection / generation time includes the time when the requirements are recorded or generated.

[0051] Table 2: Enterprise requirement data fields

[0052] The steps for data deduplication and merging are as follows: The structured information of scientific and technological achievements is compared based on the source institution / unit, output time, personnel, and technical field, as well as the semantic analysis and comparison of the title and abstract. A unique identifier of a multi-field combination is generated. Duplicate records are identified and merged based on the unique identifier of the multi-field combination. For example, the unique identifier can be a hash value generated based on the multi-field combination.

[0053] Specifically, the source institutions / units are compared by comparing the names of the units that report or publish the results and their unified social credit codes and other information to determine whether they come from the same legal entity, thereby identifying possible duplicate submitted results.

[0054] Personnel comparison includes inventor information comparison and researcher comparison. For example, for patent data, the inventor's name, ID number, work unit and other information are comprehensively compared to identify whether there are similar or identical results completed by the same invention team between different results.

[0055] The consistency of technical field classification is judged by combining the technical field to which the scientific and technological achievements belong (such as artificial intelligence, biomedicine, new materials, etc.) and keyword information to judge the consistency of scientific and technological achievements in research direction and assist in identifying duplicates.

[0056] Compare the production time of scientific and technological achievements, sort and compare them by the release time or appraisal time of scientific and technological achievements, and avoid the same achievement being entered multiple times in different time periods.

[0057] Semantic analysis of titles and abstracts uses natural language processing technology to perform text similarity analysis on the titles and abstracts of achievements, identifying achievement entries with similar expressions but slightly different fields.

[0058] As a preferred option, in the data collection and governance stage, the original multi-source heterogeneous scientific and technological achievement data is converted into scientific and technological achievement structured information, field extraction and structured warehousing operations are performed according to Table 1, and data deduplication and merging are performed.

[0059] Based on these processes, the need for rapid grouping can be achieved. Furthermore, the scientific and technological achievement database stores scientific and technological achievement data and scientific and technological achievement data groups, and the scientific and technological achievement data groups include grouping flags and one or more scientific and technological achievement data associated with the grouping flags.

[0060] The grouping mark includes personnel information (inventor information or researcher information), the technical field to which the achievement belongs, the completion time of the achievement, the source organization / unit, etc. The specific grouping mark can be determined based on the fields in the structured information of scientific and technological achievements.

[0061] As a preferred approach, we propose an optimized method for generating a matching result set, further applying structured information to the sorting and selection of search results. This method, based on the aforementioned steps, further incorporates the output of vector recall and full-text search recall, then combines the time of scientific and technological achievement production or the time of enterprise demand generation to generate a matching result set. This method provides a more timely matching result set.

[0062] As a preferred embodiment, for further applying structured information to the sorting and selection of search results, an optimized matching result set generation method is proposed. A method for matching scientific and technological achievements with enterprise needs is further disclosed. When the user request content does not clearly recall scientific and technological achievements or enterprise needs, both scientific and technological achievements and enterprise needs are recalled. An optimized matching result set generation method is further proposed. Based on the steps mentioned above, it also includes the steps of performing vector recall and full-text search recall based on the search request to recall scientific and technological achievements and enterprise needs respectively, fusing the scientific and technological achievements recalled by the vector recall and full-text search, and fusing the enterprise needs recalled by the vector recall and full-text search; sorting the fused scientific and technological achievements and enterprise needs according to the output time of the scientific and technological achievements and the generation time of the enterprise needs, combined with the maturity of the scientific and technological achievements, to obtain a matching result set. After recalling the scientific and technological achievements and enterprise needs, the scientific and technological achievements list and the enterprise needs list are sorted according to the output time of the scientific and technological achievements and the generation time of the enterprise needs, combined with the maturity of the scientific and technological achievements, and setting sorting prompt words to standardize the relevance of the sorting results of the scientific and technological achievements list and the enterprise needs list.

[0063] Based on this, we further combine the specific vector recall method and the specific full-text search recall method to output and display the matching results. The final displayed matching results, whether they are scientific and technological achievements or corporate needs, have several preceding data with high timeliness, and the preceding scientific and technological achievements and corporate needs are also highly correlated with each other, making them more suitable for users with enterprise-school and enterprise-research docking needs.

[0064] As a matching method ultimately applied to a scientific and technological achievement / enterprise technology demand matching platform, relevant options are configured. If the user requests to recall the scientific and technological achievements produced by Research Institute A from 2024 to 2025, the corresponding simultaneously associated scientific and technological achievement data will be called according to the grouping flag and the final matching result will be output.

[0065] Also disclosed is an electronic device comprising a memory and a processor, wherein the memory is used to store one or more computer instructions, wherein the one or more computer instructions are executed by the processor to implement the above-mentioned method for matching scientific and technological achievements with enterprise technical needs.

[0066] Those skilled in the art will clearly understand that, for ease and brevity of description, the specific operating process of the electronic device described above can refer to the corresponding process in the aforementioned method embodiment and will not be repeated here. Also disclosed is a computer-readable storage medium storing a computer program that, when executed by a computer, implements the aforementioned method for matching scientific and technological achievements with enterprise technical needs.

[0067] Exemplarily, a computer program may be divided into one or more modules / units, one or more modules / units being stored in a memory and executed by a processor, and the input interface and output interface completing the I / O interface transmission of data to complete the present invention. One or more modules / units may be a series of computer program instruction segments capable of completing specific functions, and the instruction segments are used to describe the execution process of the computer program in a computer device.

[0068] The computer device may be a desktop computer, laptop, PDA, cloud server, or other computing device. The computer device may include, but is not limited to, a memory and a processor. Those skilled in the art will appreciate that this embodiment is merely an example of a computer device and does not limit the computer device. The computer device may include more or fewer components, or a combination of certain components, or different components. For example, the computer device may also include an input device, a network access device, a bus, and the like.

[0069] The processor may be a central processing unit (CPU), other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA), other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor, etc.

[0070] Memory can be an internal storage unit of a computer device, such as a computer device's hard drive or memory. Memory can also be an external storage device of a computer device, such as a plug-in hard drive, a Smart Media Card (SMC), a Secure Digital (SD) card, or a Flash Card. Furthermore, memory can include both internal and external storage units of a computer device. Memory is used to store computer programs and other programs and data required by the computer device. Memory can also be used to temporarily store data on an output device. The aforementioned storage media include various media that can store program code, such as USB flash drives, mobile hard drives, read-only memories (ROMs), random access memories (RAMs), disks, or optical disks.

[0071] The above descriptions merely illustrate some preferred embodiments of the present invention and the underlying technical principles. Those skilled in the art should understand that the scope of the invention encompassed by the embodiments of the present invention is not limited to technical solutions formed by specific combinations of the aforementioned technical features. It also encompasses other technical solutions formed by any combination of the aforementioned technical features or their equivalents, without departing from the aforementioned inventive concept. For example, a technical solution formed by replacing the aforementioned features with (but not limited to) technical features with similar functions disclosed in the embodiments of the present invention.

Claims

1. A method for matching scientific and technological achievements with enterprise technology needs, characterized in that: The steps include: The original data of scientific and technological achievements are constructed into scientific and technological achievement structured information, and a first vector is formed based on the scientific and technological achievement structured information; the scientific and technological achievement structured information and its corresponding first vector are associated and stored to construct a scientific and technological achievement database; Constructing the original enterprise demand data into enterprise demand structured information, and forming a second vector based on the enterprise demand structured information; The enterprise demand structured information is associated with the corresponding second vector and stored to build an enterprise demand library; In response to user requests, the large language model is used to rewrite the user requests into retrieval requests containing structured prompt words. Based on the retrieval request, vector recall and full-text retrieval recall are performed respectively, and the output results of vector recall and full-text retrieval recall are fused to obtain a matching result set. The pre-trained sorting model is called, and the user request and the recall results in the corresponding matching result set are used as input. The local matching relationship and global semantic association between the user request and the recall results are analyzed, and a score representing the degree of matching between the user request and the recall results is output. The results are sorted according to the score to obtain a second sorted list, and the matching results are displayed based on the second sorted list.

2. The method for matching scientific and technological achievements with enterprise technology needs according to claim 1, characterized in that: The full-text search and recall method includes: Perform keyword matching search on all enterprise demand structured information in the enterprise demand library based on the rewritten search request, and / or Based on the rewritten search request, a keyword matching search is performed in all the structured information of scientific and technological achievements in the scientific and technological achievements database.

3. The method for matching scientific and technological achievements with enterprise technology needs according to claim 1, characterized in that: The vector recall method includes: Call the vector model to calculate the search request rewritten by the large language model to obtain the query vector; Using a vector similarity algorithm, we search the vector space of the enterprise demand database with the query vector, calculate the vector dot product, and recall the top-order results that are most semantically similar to the query vector. A vector similarity algorithm is used to search the vector space of the scientific and technological achievements database with the query vector, calculate the vector dot product, and recall several pre-sequence results that are semantically closest to the query vector.

4. The method for matching scientific and technological achievements with enterprise technology needs according to claim 1, characterized in that: Before displaying the matching results, the method further includes filtering the recall results below a threshold in the second sorted list to obtain a third sorted list, and displaying the matching results based on the third sorted list.

5. A method for matching scientific and technological achievements with enterprise technology needs according to claim 4, characterized in that: Before displaying the matching results, the following steps are also included: language model analysis and scoring: Call the large language model, input the third ranked list, construct prompt words to analyze the correlation between the user request and the matching results, output the final matching score, and display the matching results based on the final matching score; Among them, the correlation analysis includes the correlation between user requests and matching results based on production technology, application fields, and materials.

6. The method for matching scientific and technological achievements with enterprise technology needs according to claim 1, characterized in that: The structured information of scientific and technological achievements includes a first title and an abstract, and full-text indexes are established for the first title and the abstract respectively; the first title and the corresponding abstract are spliced ​​into a first text content, and the first text content is converted into a first vector; the structured information of enterprise needs includes a second title and a second detail, and full-text indexes are established for the second title and the second detail respectively; the second title and the corresponding second detail are spliced ​​into a second text content, and the second text content is converted into a second vector.

7. A method for matching scientific and technological achievements with enterprise technology needs according to claim 6, characterized in that: It also includes data deduplication and merging steps. The structured information of scientific and technological achievements is compared based on the source organization / unit, output time, personnel and technical field, as well as the semantic analysis and comparison of the first title and abstract, to generate a unique identifier of a multi-field combination, and duplicate records are identified and merged based on the unique identifier of the multi-field combination.

8. The method for matching scientific and technological achievements with enterprise technology needs according to claim 1, characterized in that: The scientific and technological achievement database stores scientific and technological achievement data and scientific and technological achievement data groups. The scientific and technological achievement data groups include grouping marks and one or more scientific and technological achievement data associated with the grouping marks.

9. The method for matching scientific and technological achievements with enterprise technology needs according to claim 6, characterized in that: The method for generating a matching result set further includes fusing the output results of the vector recall and the full-text search recall and combining them with the output time of the scientific and technological achievements or the generation time of the enterprise needs to obtain a matching result set.

10. The method for matching scientific and technological achievements with enterprise technology needs according to claim 6, characterized in that: The method for generating a matching result set further includes executing vector recall and full-text search recall based on the search request to respectively recall scientific and technological achievements and enterprise needs, fusing the scientific and technological achievements recalled by the vector recall and the full-text search, and fusing the enterprise needs recalled by the vector recall and the full-text search; According to the output time of scientific and technological achievements and the generation time of enterprise needs, combined with the maturity of scientific and technological achievements, the integrated scientific and technological achievements and enterprise needs are sorted to obtain a set of matching results.

11. A system for matching scientific and technological achievements with enterprise technology needs, characterized by: Includes the following structure: A scientific and technological achievement database construction unit: constructs the original scientific and technological achievement data into scientific and technological achievement structured information, forms a first vector based on the scientific and technological achievement structured information; associates and stores the scientific and technological achievement structured information with the corresponding first vector, and constructs a scientific and technological achievement database; Enterprise demand library construction unit: constructs the original enterprise demand data into enterprise demand structured information, and forms a second vector based on the enterprise demand structured information; The enterprise demand structured information is associated with the corresponding second vector and stored to build an enterprise demand library; Matching unit: In response to user requests, the large language model is used to rewrite the user request into a retrieval request containing structured prompt words, vector recall and full-text retrieval recall are performed based on the retrieval request respectively, the output results of the vector recall and full-text retrieval recall are integrated to obtain a matching result set, a pre-trained sorting model is called, the user request and the corresponding recall results in the matching result set are used as input, the local matching relationship and global semantic association between the user request and the recall results are analyzed, a score representing the degree of matching between the user request and the recall results is output, the results are sorted according to the score to obtain a second sorted list, and the matching results are displayed based on the second sorted list.

Images