A matching and recommendation method, device and equipment for scientific research achievement transformation

Abstract

This specification relates to the field of artificial intelligence technology and provides a matching and recommendation method, apparatus, and device for the transformation of scientific research results. The method includes: receiving user input information and determining the user's demand type based on the user input information; when the user demand type is a first request, semantically expanding the user input information to obtain a second request, the second request containing an expanded terminology set; searching in a pre-constructed knowledge graph according to the second request to obtain a first search result set; calculating the semantic similarity between each search result in the first search result set and the user input information, and determining a second search result set based on the semantic similarity; and outputting the second search result set to the user's corresponding terminal. Through this specification, the matching efficiency and accuracy of scientific research results and industrial needs can be improved.

Description

Technical Field

[0001] The embodiments in this specification relate to the field of artificial intelligence technology, and in particular to a matching and recommendation method, apparatus and equipment for the transformation of scientific research results. Background Technology

[0002] The commercialization of scientific research achievements serves as a crucial bridge connecting technological innovation with industrial applications. Its core objective is to accurately match research results such as papers, patents, and technical solutions with the actual technological needs of enterprises, which is of great significance for promoting industrial upgrading and improving the rate of technological innovation commercialization. With the continuous development of innovative achievements in the scientific research field and the diversification of technological demands from the industrial sector, differences exist in terminology. For example, enterprises often describe their needs using application-oriented expressions, while researchers describe their achievements using professional academic terminology. Due to the lack of effective semantic understanding and mapping mechanisms on the internet, simple keyword matching often fails to cover deeper relationships. Therefore, the efficient and accurate matching of scientific research achievements with enterprise needs has become an urgent issue for the industry.

[0003] Existing matching and recommendation schemes for scientific research output transformation input texts of research results and industry needs into a general language model. This model generates recommendation results relevant to user needs through semantic understanding. However, the language models used in this method are mostly trained on general corpora, lacking precise understanding of specialized terminology and logical relationships within the research field. This makes it difficult to accurately grasp the complex, cross-domain semantic connotations of research results and industry needs. Furthermore, this method fails to consider the differences in terminology between research and industry, leading to the omission of some core related results due to differing terminology. Consequently, the search results suffer from poor relevance and low matching efficiency, hindering the efficient and accurate matching of research results with industry needs. Therefore, there is an urgent need for a matching and recommendation method for scientific research output transformation that can improve the efficiency and accuracy of matching research results with industry needs. Summary of the Invention

[0004] In view of the above-mentioned problems in the prior art, the purpose of the embodiments of this specification is to provide a matching and recommendation method, apparatus and equipment for the transformation of scientific research results, so as to solve the problem that it is difficult to achieve efficient and accurate matching between scientific research results and industrial needs in the prior art.

[0005] To solve the above-mentioned technical problems, the specific technical solutions of the embodiments in this specification are as follows: On the one hand, embodiments of this specification provide a matching and recommendation method for the transformation of scientific research results, the method comprising: Receive user input information and determine the type of user request based on the user input information; When the user request type is a first request, the user input information is semantically expanded to obtain a second request, which includes an expanded terminology set. Based on the second request, a search is performed in the pre-built knowledge graph to obtain a first search result set; Calculate the semantic similarity between each search result in the first search result set and the user input information, and determine the second search result set based on the semantic similarity; The second search result set is output to the user's corresponding terminal.

[0006] Further, determining the type of user need based on the user input information includes: The user input information is converted into a structured query; Map the structured query to function parameters; The corresponding target function is determined based on the function parameters; The user requirement type is determined based on the function name corresponding to the target function call.

[0007] Furthermore, the construction process of the knowledge graph includes: Acquire knowledge data in the target domain; Extract multiple preset types of entities and semantic relationships between entities from the target domain knowledge data; Semantic mapping is performed on the domain terms associated with each extracted entity to generate a set of semantic descriptions corresponding to each domain term. The set of semantic descriptions contains semantic descriptions at multiple semantic levels. A knowledge graph is constructed based on the entities, the semantic relationships between entities, and the set of semantic descriptions.

[0008] Furthermore, semantic expansion of the user input information includes: Extract domain terms from the structured query; The domain terms are input into a large language model to generate an initial extended term set, which contains extended terms at multiple semantic levels. Each extended term in the initial extended term set is semantically validated in the knowledge graph; The extended terms that pass semantic validation will be used as the extended term set for the corresponding domain terms.

[0009] Furthermore, the semantic level includes at least the professional terms corresponding to the domain terminology, the standard industry terms corresponding to the professional terms, and the colloquial expressions corresponding to the standard industry terms.

[0010] Further, the retrieval in the pre-built knowledge graph according to the second request includes: Determine the terminology set corresponding to the user input information, the terminology set including domain terms and extended terminology sets; If any term in the term set has a matching result in the knowledge graph, it is used as a search term; If any term in the term set does not have a matching result in the knowledge graph, then one or more semantically similar terms are determined from the knowledge graph as retrieval items; The search is performed on the knowledge graph according to the search terms, and the retrieved entity information is used to form a first search result set.

[0011] Furthermore, before outputting the second search result set to the user's corresponding terminal, the method further includes: For each search result in the second search result set, verify whether there is a tracing path in the knowledge graph that is associated with the user input information; The search results of the tracing paths associated with the user input information are used to form a candidate result set; Calculate the confidence score of each candidate result in the candidate result set relative to the user input information. The confidence score is used to reflect the degree to which the search results solve the user's needs. If the confidence score is lower than the preset threshold, the step of determining the user's need type based on the user input information is repeated. If the step of determining the user's need type based on user input information is repeated more than a preset number of times, the retrieval process will be terminated and a preset prompt message will be output.

[0012] On the other hand, embodiments of this specification provide a matching and recommendation device for the transformation of scientific research results, the device comprising: A receiving module is used to receive user input information and determine the type of user request based on the user input information; An extension module is used to semantically extend the user input information to obtain a second request when the user request type is a first request, wherein the second request contains an extended terminology set; The retrieval module is used to perform a retrieval in a pre-built knowledge graph according to the second request and obtain a first retrieval result set; The calculation module is used to calculate the semantic similarity between each search result in the first search result set and the user input information, and to determine the second search result set based on the semantic similarity. The output module is used to output the second search result set to the user's corresponding terminal.

[0013] In another aspect, embodiments of this specification also provide a computer device, including a memory, a processor, and a computer program stored in the memory, wherein the computer program, when executed by the processor, performs instructions of any of the methods described above.

[0014] In another aspect, embodiments of this specification also provide a computer-readable storage medium having a computer program stored thereon, the computer program being executed by a processor of a computer device to perform instructions for any of the methods described above.

[0015] In another aspect, embodiments of this specification also provide a computer program product, which, when run by the processor of a computer device, executes instructions for any of the methods described above.

[0016] By adopting the above technical solution, the matching and recommendation method for scientific research results transformation provided in the embodiments of this specification addresses the problems of inaccurate professional semantic understanding, omission of matching results due to differences in cross-domain terminology systems, and insufficient matching efficiency and accuracy of existing general language models in matching scientific research results transformation. It first identifies the type of user needs, then semantically expands specific needs to cover different terminology expressions, avoiding the omission of important related results. Next, it retrieves complex cross-domain semantics and professional logical relationships based on a pre-constructed scientific research knowledge graph. Finally, it optimizes the results through semantic similarity filtering. This significantly improves the accuracy and efficiency of matching scientific research results with industry needs, effectively solving the matching obstacles caused by semantic deviations and terminology differences between the scientific research and industry fields, and providing strong support for the rapid and efficient transformation of scientific research results.

[0017] The above description is merely an overview of some embodiments of the technical solutions in this specification. In order to better understand the technical means of some embodiments of this specification and to implement them in accordance with the content of the specification, and to make the above and other objects, features and advantages of the embodiments of this specification more apparent and understandable, preferred embodiments are described below in detail with reference to the accompanying drawings. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in the embodiments or prior art of this specification, the drawings used in the description of the embodiments or prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this specification. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1 The diagram illustrates a schematic of an implementation system for a matching and recommendation method for the transformation of scientific research results, as described in some embodiments of this specification. Figure 2 The diagram illustrates the steps of a matching and recommendation method for the transformation of scientific research results in some embodiments of this specification. Figure 3 This specification illustrates an overall flowchart of a matching and recommendation method for the transformation of scientific research results, as shown in some embodiments of this specification. Figure 4 This specification illustrates the steps for determining the type of user requirement in some embodiments; Figure 5 This specification illustrates schematic diagrams of steps for semantic expansion of user input information in some embodiments; Figure 6 The present specification illustrates the steps of knowledge graph construction in some embodiments. Figure 7 This specification illustrates schematic diagrams of steps for performing a retrieval in a knowledge graph based on a second request in some embodiments; Figure 8 This specification illustrates the steps for rearranging the first search results in some embodiments; Figure 9 This specification illustrates the steps for performing source path verification and confidence level verification on the second search result in some embodiments. Figure 10 This specification shows a schematic diagram of the structure of an apparatus in some embodiments; Figure 11 A schematic diagram of the structure of a computer device is shown in this specification.

[0020] Explanation of symbols in the attached drawings: 101. Terminal; 102. Server; 1001. Receiver module; 1002. Extension Module; 1003. Search module; 1004. Calculation Module; 1005. Output module; 1102. Computer equipment; 1104. Processor; 1106. Memory; 1108. Drive mechanism; 1110. Input / output module; 1112. Input devices; 1114. Output devices; 1116. Presentation device; 1118. Graphical User Interface; 1120. Network interface; 1122. Communication link; 1124. Communication bus. Detailed Implementation

[0021] The technical solutions in the embodiments of this specification will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this specification, and not all embodiments. Based on the embodiments in this specification, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this specification.

[0022] It should be noted that the terms "first," "second," etc., used in this specification, claims, and the foregoing drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover a non-exclusive inclusion; for example, a process, method, apparatus, product, or device that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or devices.

[0023] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties. Furthermore, the acquisition, storage, use, and processing of data in the technical solutions described in the embodiments of this application all comply with relevant regulations.

[0024] It should be noted that in the embodiments of this specification, certain software, components, models and other existing solutions in the industry may be mentioned. These should be regarded as exemplary and are only intended to illustrate the feasibility of implementing the technical solutions of the embodiments of this specification. However, it does not mean that the applicant has used or necessarily used such solutions.

[0025] like Figure 1The diagram illustrates an implementation system for a matching and recommendation method for scientific research results transformation, as described in this specification. The system includes a terminal 101 and a server 102, with a communication connection established between them to enable data interaction. The terminal 101 collects and sends user input information to the server 102. The server 102 receives the user input information and processes it based on the method described in this specification. Specifically, this includes: determining the user's request type based on the user input information; when the user's request type is a first request, semantically expanding the user input information to obtain a second request, the second request containing an expanded terminology set; searching a pre-constructed knowledge graph based on the second request to obtain a first search result set; calculating the semantic similarity between each search result in the first search result set and the user input information, and determining a second search result set based on the semantic similarity. The server 102 returns the generated second search result set to the terminal 101. The terminal 101 displays the results to the user and receives user operation instructions, forming a closed-loop interaction.

[0026] In the embodiments of this specification, the server 102 may be an independent physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, content delivery networks (CDN), and big data and artificial intelligence platforms.

[0027] In an optional embodiment, terminal 101 may be an electronic device, including but not limited to self-service terminal equipment, desktop computers, tablet computers, laptops, smart wearable devices, etc. Optionally, the operating system running on the electronic device may include, but is not limited to, Android, iOS, Linux, Windows, etc. Of course, terminal 101 is not limited to the aforementioned physical electronic devices; it may also be software running on the aforementioned electronic devices.

[0028] In addition, it should be noted that, Figure 1 The example shown is merely one application environment provided by this disclosure. In practical applications, it may include multiple terminals 101, and this specification does not impose any restrictions.

[0029] Figure 2This is a schematic diagram illustrating the steps of a matching and recommendation method for the transformation of scientific research results, as provided in the embodiments of this specification. This specification provides the operational steps of the method described in the embodiments or flowcharts, but based on conventional or non-creative labor, it may include more or fewer operational steps. The order of steps listed in the embodiments is merely one possible execution order among many, and does not represent the only possible execution order. In actual system or device products, the methods shown in the embodiments or drawings can be executed sequentially or in parallel. Specifically, as shown in the figures... Figure 1 As shown, when applied to the server side described above, the method may include: S201: Receive user input information and determine the user's request type based on the user input information; S202: When the user request type is a first request, the user input information is semantically expanded to obtain a second request, the second request containing an expanded term set; S203: Based on the second request, a search is performed in the pre-built knowledge graph to obtain a first search result set; S204: Calculate the semantic similarity between each search result in the first search result set and the user input information, and determine the second search result set based on the semantic similarity; S205: Output the second search result set to the user's corresponding terminal.

[0030] By adopting the above technical solution, the matching and recommendation method for scientific research results transformation provided in the embodiments of this specification addresses the problems of inaccurate professional semantic understanding, omission of matching results due to differences in cross-domain terminology systems, and insufficient matching efficiency and accuracy of existing general language models in matching scientific research results transformation. It first identifies the type of user needs, then semantically expands specific needs to cover different terminology expressions, avoiding the omission of important related results. Next, it retrieves complex cross-domain semantics and professional logical relationships based on a pre-constructed scientific research knowledge graph. Finally, it optimizes the results through semantic similarity filtering. This significantly improves the accuracy and efficiency of matching scientific research results with industry needs, effectively solving the matching obstacles caused by semantic deviations and terminology differences between the scientific research and industry fields, and providing strong support for the rapid and efficient transformation of scientific research results.

[0031] In the embodiments of this specification, user input information may include speech or natural language text, such as "find AI quality inspection technologies to improve the yield rate of production lines." If it is speech, it is first converted into text form before further processing. After receiving user input information, the system uses a large language model and function call mechanism to identify the user's intent, determine the type of user need, and classify the request into retrieval needs or dialogue needs, such as... Figure 3As shown, when a user's request is identified as a retrieval type, the natural language request is parsed into a structured query request and expanded to narrow the gap between scientific research terminology and industry needs. Then, based on the expanded keywords, a search and matching process is performed in the knowledge graph to filter out the result set most relevant to the user's request. In the embodiments of this specification, the knowledge graph is a pre-constructed four-layer structure knowledge graph containing experts, achievements, applications, and needs. It has unified the modeling and association of entities (such as experts, scientific research achievements, and application scenarios) and their relationships scattered in heterogeneous data from multiple sources such as papers, patents, and project announcements, and has established a three-level semantic mapping at different semantic levels. When a user's request is identified as a dialogue type, a natural language answer is directly generated using a large language model for generalized question-and-answer consultation. At the same time, the confidence of the search results is evaluated. When the confidence is low, a reflection mechanism is triggered to automatically re-identify the intent. If more than two reflections are performed, it is considered that the graph content is insufficient to answer the user's question, and the answer is "No relevant content was found. Please try a different question." By diverting the above-mentioned demands, we can avoid using a uniform processing flow for all demands, ensuring the accuracy of search-type demands while also taking into account the flexibility of dialogue-type demands.

[0032] In the embodiments of this specification, refer to Figure 4 Step S201 determines the user's demand type based on the user's input information, specifically including the following steps: S401: Convert the user input information into a structured query; S402: Map the structured query to function parameters; S403: Determine the corresponding target calling function based on the function parameters; S404: Determine the user requirement type based on the function name corresponding to the target call function.

[0033] Understandably, unlike traditional classification rules based on manual design (such as keyword matching, rule templates, etc.), the requirement type determination in this embodiment is based on a large language model, introducing a function call mechanism to automatically map the user's natural language input into structured function call parameters. By using the function name and parameter structure selected by the LLM, the user's specific intent type can be automatically determined. First, based on the LLM's semantic parsing capabilities, the preprocessed natural language input is converted into a standardized structured format, such as JSON. During the conversion process, the LLM automatically extracts key parameters from the input, such as the requirement topic, target type, and constraints, and organizes them according to preset field rules. For example, if a user inputs: "Our company wants to find an AI quality inspection technology that can improve the yield rate of the production line," it will be converted into [{"content":"AI quality inspection technology","data_type":[{"patent","achievement"}]}], where the "content" field corresponds to the requirement topic, and the "data_type" field corresponds to the result type. This approach solves the problems of ambiguity and non-standardization in natural language input, providing a unified processing foundation for subsequent function parameter mapping and intent recognition, and avoiding deviations in subsequent processes caused by chaotic input formats. Then, based on multiple callable functions pre-registered with the LLM, the LLM automatically reads the core fields and corresponding content from the structured query and accurately fills them into the corresponding parameter fields of the target function. Each function corresponds to an intent category, such as `retrieve_paper_infomation` corresponding to the intent of paper retrieval, and each function contains a clear parameter structure, with key parameters such as publication time ranges (`date_ranges`), authors (`authors`), and paper topic (`content`). The field types and meanings of the parameters correspond one-to-one with the core fields of the structured query. For example, corresponding to the user input above: "Our company wants to find an AI quality inspection technology that can improve the yield rate of the production line," after being converted into a structured query in step S401, the fields in this structured query are mapped to the parameter values ​​corresponding to multiple callable functions, obtaining a parameter set corresponding to the user's needs. This aligns the fields of the structured query with the pre-set functions, ensuring that the functions can accurately receive and recognize the constraints of the user's needs, providing data support for the subsequent matching of the target function. Step S403 compares the parameter set after the structured query mapping with the parameter characteristics of each preset function to filter out functions that match the parameter fields as target calling functions. In some embodiments, if the parameter set matches multiple functions simultaneously, the LLM will further determine the uniqueness of the target calling function by combining parameter weights. Step S404 By identifying the name of the target calling function, its corresponding intent category can be directly determined, thereby determining the user's demand type.For example, if the target function to be called is retrieve_paper_infomation, then the user's request type is determined to be a paper retrieval request, which belongs to the retrieval type. If the target function to be called is general_consultation, then it is determined to be a dialogue type request.

[0034] In the embodiments of this specification, refer to Figure 5 Step S202 semantically expands the user input information, specifically including the following steps: S501: Extract domain terms from the structured query; S502: Input the domain terms into the large language model to generate an initial extended term set, which contains multiple semantic-level extended terms; S503: Perform semantic verification on each extended term in the knowledge graph for each extended term in the initial extended term set; S504: The extended terms that pass semantic validation will be used as the extended term set of the corresponding domain terms.

[0035] Understandably, in retrieval scenarios, semantic expansion of structured queries, compared to many existing technologies that use static dictionaries or thesaurus for keyword expansion, leverages the powerful generation and semantic recognition capabilities of large-scale models to expand semantics through prompt word engineering, facilitating subsequent graph verification. Semantic expansion primarily involves expanding the content field in intent recognition, generating relevant terms at different semantic levels through a large language model. In this embodiment, the semantic levels include at least professional terms (level_3) corresponding to domain terminology, standard industry terms (level_2) corresponding to professional terms, and common expressions (level_1) corresponding to the standard industry terms. Popular expressions are defined as simplified terms or common words used in everyday communication by the general public. They are usually understandable without requiring a professional background, are simple and clear, and can be widely disseminated. Standard industry terms are defined as terms that are generally accepted and standardized in a particular industry or field. Although they are specialized, they are more widely understood by industry practitioners and are usually defined by industry standardization organizations (such as ISO and IEEE). Professional terms are defined as terms with specific meanings in a particular discipline or industry that can only be accurately understood by experts in the relevant field. They are usually more abstract and complex and require highly specialized background knowledge to explain. By expanding field terms to the above three levels of related terms, we can cover diverse expressions from the research, industry, and public perspectives, thus filling the semantic gaps in different fields. For example, AI quality inspection technology can be expanded into a set of related synonyms, professional terms, or cutting-edge terms: {"input": "AI quality inspection technology","level_1": ["computer vision", "machine vision", "defect detection"],"level_2": ["industrial camera", "image processing algorithm", "CNN"],"level_3": ["image recognition based on convolutional networks", "few-shot learning", "unsupervised defect detection"]}.

[0036] In the embodiments described in this specification, the expanded terms will be validated through a knowledge graph to ensure that the expanded terms are consistent with the original semantics, thereby ensuring the usability of the expanded results. Specifically, the validation logic includes two steps. The first step is existence validation, which aims to check whether the expanded terms exist in the entity system of the knowledge graph, such as whether the terms correspond to entities such as achievements, technologies, and application scenarios in the graph. The second step is semantic consistency validation, which aims to verify the semantic relevance between the expanded terms and the original domain terms through entity relationships in the graph, ensuring that the core connotation of the terms has not changed. This addresses the illusion problem that may exist in large language models, filtering out invalid terms that are irrelevant to the domain, semantically inconsistent, or have no corresponding entities in the graph. For example, if AI quality inspection technology is mistakenly expanded to medical image diagnosis, it will be removed because there is no associated path in the graph. At the same time, through graph constraints, it is ensured that the expanded terms can be directly used for subsequent retrieval, improving retrieval efficiency. The final expanded term set is the set of valid terms after validation in step S503. This term set not only covers the usage habits of different fields, bridging the semantic gap between scientific research and industry, but also ensures semantic accuracy through knowledge graph validation, avoiding invalid expansions.

[0037] In the embodiments of this specification, refer to Figure 6 The construction process of a knowledge graph specifically includes the following steps: S601: Acquire knowledge data in the target domain; S602: Extract multiple entities of preset types and semantic relationships between entities from the target domain knowledge data; S603: Perform semantic mapping on the domain terms associated with each extracted entity to generate a semantic description set corresponding to each domain term, wherein the semantic description set contains semantic descriptions at multiple semantic levels; S604: A knowledge graph is constructed based on the entities, the semantic relationships between entities, and the set of semantic descriptions.

[0038] It is understandable that target domain knowledge data refers to multi-source heterogeneous data covering the entire process of scientific research achievement transformation, including scientific research data such as academic papers, patent texts, scientific research project applications, and technical achievement appraisal reports, which carry professional technical information; industry data such as enterprise technology demand announcements, project cooperation tenders, and industry application cases, which carry practical application demand information; and related resource data such as expert and scholar information, scientific research institution information, and technology transformation project records, which carry entity association information. After obtaining the above data, the obtained raw data is standardized, including unstructured text cleaning and format unification, such as standardizing fields like patent number and paper publication time, removing redundant symbols, and correcting ambiguous expressions, to avoid deviations in map construction caused by chaotic data formats. In the embodiments of this specification, the entities extracted in step S602 include expert entities, achievement entities, application entities, and demand entities. Specifically, LLM and prompt word engineering are used. By inputting preset extraction rules into the LLM, such as identifying the technology name, applicant, and inventor from the patent text, entities that meet the preset types are automatically extracted from the data. Semantic relation extraction leverages LLM's semantic understanding capabilities to identify the inherent logical connections between entities and transform them into standardized relational representations, forming entity-relationship-entity triples. Examples include the "proposal" relationship between experts and achievements, the "application" relationship between achievements and application scenarios, and the "matching" relationship between needs and achievements. To reduce the differences in terminology between scientific research and industry and bridge the semantic gap, step S603 in this embodiment performs semantic mapping on professional terms representing technical directions, achievement attributes, and demand themes within entities. For each domain term, LLM automatically generates the aforementioned three-level semantic description, forming a complete semantic description set. For example, the semantic description set for the gradient descent algorithm is: {professional term: "gradient descent algorithm", standard industry term: "neural network optimization method", colloquial expression: "intelligent learning method"}. After generation, semantic consistency verification is performed on each level of semantic description to ensure that the generated three-level descriptions do not exhibit semantic drift. Step S604 The knowledge graph is a four-layered architecture containing entities of experts, results, applications and needs. Specifically, the extracted entities are added to the graph according to the four-layer classification, and the association paths between entities are established through semantic relationships. Then, the domain terms contained in each entity are associated with the corresponding set of semantic descriptions.

[0039] In this embodiment of the specification, to ensure the timeliness and accuracy of the knowledge graph, an update method combining periodic batch import of new data and real-time knowledge extraction is used. Specifically, the latest data is periodically collected from paper databases and patent databases, knowledge extraction is performed using prompt word engineering, and three-level semantic expansion is applied to the extracted domain terms. Newly extracted knowledge undergoes rigorous verification through graph relationship validation before being added to the knowledge graph to ensure its authenticity and rationality. If the LLM cannot clearly determine the validity of some data, it automatically marks that data and places it in a manual review queue. The specific verification process is as follows: First, a keyword search operation is performed on the newly extracted knowledge. The search target is the entities already stored in the knowledge graph, including but not limited to included technical terms, scientific research terms, patent information, and paper information. If no entity or relationship between entities corresponding to the keyword is matched in the knowledge graph after the search, the system automatically adds the new knowledge to the knowledge graph and constructs a corresponding association path for it. If a relevant entity or relationship between entities is matched after the search, the vector similarity between the keywords of the newly extracted knowledge and the keywords corresponding to the matched entities is calculated based on the semantic vector model, and the vector similarity is sorted. If the vector similarity is less than a preset threshold, it is determined that the new knowledge conflicts with the knowledge already stored in the knowledge graph. The conflict includes but is not limited to cross-domain terminology conflicts, emerging terminology adaptation conflicts, and semantic conflicts caused by contextual differences. In this case, the new knowledge needs to be submitted to the manual review queue for review. Through the above dynamic update mechanism, the embodiments of this specification can continuously and efficiently integrate the latest scientific research results and industry dynamics into the knowledge graph without large-scale manual intervention, thereby ensuring the real-time performance and credibility of the knowledge graph recommendation results.

[0040] In the embodiments of this specification, refer to Figure 7 The retrieval process, based on the second request, involves the following steps within a pre-built knowledge graph: S701: Determine the terminology set corresponding to the user input information, the terminology set including domain terms and extended terminology set; S702: If any term in the term set has a matching result in the knowledge graph, then it shall be used as a search term; S703: If any term in the term set does not have a matching result in the knowledge graph, then one or more semantically similar terms are determined from the knowledge graph as retrieval items; S704: Perform a search in the knowledge graph according to the search terms, and form a first search result set from the retrieved entity information.

[0041] It is understood that the embodiments in this specification extract domain terms from the structured query corresponding to the user input information, and use LLM to semantically expand the extracted domain terms to obtain an expanded terminology set. Then, through entity matching verification of the knowledge graph, directly usable search keywords are selected to ensure the accuracy and efficiency of the retrieval. The criteria for determining the matching results are: any term in the terminology set has semantic consistency with an entity already stored in the knowledge graph, that is, the term directly exists in the entity name or semantic description set of the knowledge graph. When a term meets the above matching conditions, it is directly used as a search term because the term has passed the structured verification of the knowledge graph, its association with the entity in the graph is clear, and its semantics are unambiguous. It can be directly used for subsequent retrieval to quickly locate relevant entities, avoiding retrieval delays caused by additional processing, while ensuring the accuracy of the search results. If a term does not meet the above matching conditions, one or more semantically similar terms are determined from the knowledge graph as search terms. The method for determining semantically similar terms includes: first, based on the semantic relationships between entities in the knowledge graph, candidate terms whose semantic domain is consistent with that of the target term are selected; then, the vector similarity between the target term and the candidate terms is calculated using a semantic vector model, and terms with similarity greater than a preset threshold are selected as semantically similar terms. After determining the search term, starting from the search term, the semantic relationships between entities in the knowledge graph are traversed to extract all entities that are directly or indirectly related to the search term. Entity information includes complete attribute data and relationship data of the entity. For achievement entities, this includes patent number, paper title, publication time, core technical features, etc.; for application entities, this includes application scenario name, applicable industry, technical requirements, etc.; for demand entities, this includes demand issuer, demand description, deadline, etc., thus ensuring the completeness of entity information. The first search result set is a collection of all related entity information. This set will be deduplicated and finally stored in a structured format to facilitate similarity calculation and result rearrangement in subsequent steps.

[0042] The first search result set generated through the above retrieval process effectively addresses the issue of missing key results due to the semantic gap, ensuring the comprehensiveness of the search. However, since this retrieval primarily relies on term matching and entity association filtering, it does not further explore the deep semantic fit between user needs and candidate entities. Therefore, some candidate entities may have weak relevance to user needs, and the ranking may lack semantic priority. To further improve matching accuracy and select the research results that best match user needs, the first search result set needs to be rearranged. This process is achieved through deep semantic analysis using a semantic vector model, referring to... Figure 8 Specifically, this includes steps S801 to S803: S801: Convert the user input information and the information of each entity in the first search result set into a semantic vector; S802: Calculate the semantic similarity between the semantic vector corresponding to each entity and the semantic vector corresponding to the user input information; S803: The first search result set is rearranged according to semantic similarity, and the second search result set is obtained by filtering.

[0043] For entities with semantic similarity greater than a preset threshold, they are sorted from highest to lowest semantic similarity. Entities with higher scores indicate a stronger semantic match with the user's needs and are ranked higher, ensuring users receive the most relevant recommendations first. Entities with semantic similarity less than the preset threshold indicate a weaker semantic connection to the user's needs and are removed to avoid invalid results wasting resources and negatively impacting user experience. In this way, through the reordering and filtering of search results, the accuracy of recommendations is ensured while improving the efficiency of users obtaining relevant information, effectively solving the problems of poor relevance and excessive invalid information in existing technologies.

[0044] In the embodiments described in this specification, considering the stringent requirements for the accuracy, traceability, and practical applicability of recommendation results in scientific research achievement transformation scenarios, relying solely on semantic similarity cannot fully guarantee the reliability of the results. There may be instances where some entities, while semantically similar to the user's needs, lack clear logical connections in the knowledge graph or fail to address the user's core needs. To further ensure the compliance, verifiability, and practical value of the recommendation results, before outputting the second search result set to the user's terminal, source path verification and confidence checks are required to ensure that the final output results include clear sources and logical basis, enabling users to understand the result generation process and verify the authenticity of the information. Figure 9 Specifically, this is achieved through the following steps S901 to S905: S901: For each search result in the second search result set, verify whether there is a tracing path in the knowledge graph that is associated with the user input information; S902: Form a candidate result set from the retrieval results of the tracing paths associated with the user input information; S903: Calculate the confidence score of each candidate result in the candidate result set relative to the user input information, wherein the confidence score is used to reflect the degree to which the search results solve the user's needs; S904: If the confidence score is lower than the preset threshold, then the step of determining the user's need type based on the user input information is executed again; S905: If the step of determining the user's demand type based on the user input information is re-executed more than a preset number of times, the retrieval process is terminated and a preset prompt message is output.

[0045] It is understood that the confidence score is an indicator used to quantify the degree to which candidate results actually address the user's core needs. The score ranges from 0 to 1; the closer the score is to 1, the better the candidate result meets the user's needs; the closer it is to 0, the weaker the solution. In this embodiment, the confidence score is calculated considering multiple dimensions, with each dimension weighted and summed according to preset weights to obtain the final score. In this embodiment, the calculation dimensions may include: source path validity, semantic similarity, entity relevance, and requirement completeness, etc. Entity relevance measures the matching degree between the entity corresponding to the candidate result and the entity associated with the user's needs, while requirement completeness measures whether the candidate result covers the core constraints in the user's needs. By setting corresponding calculation rules and weights for each dimension, the confidence score can be calculated. If the confidence score of a candidate result is lower than a preset threshold, a reflection mechanism is triggered, and the user need type identification step is re-executed to optimize the intent recognition accuracy and subsequent retrieval process. If the user demand type identification step is re-executed more than the preset number of times, and no result with a satisfactory confidence score and a valid source path is obtained, it is determined that there are insufficient research results in the current knowledge graph to match the user's needs. At this point, the entire search process is terminated, and a preset prompt message is output to the user's corresponding terminal, such as: "No relevant content found. Please try a different question." This ensures the controllability of the system process and the integrity of the user experience. After obtaining the final candidate results, they are output along with the source path in the knowledge graph. The recommended results are presented in natural language, with clear sources and logical justifications, enabling users to understand the result generation process and verify the authenticity of the information.

[0046] Based on the aforementioned matching and recommendation method for the transformation of scientific research results, this specification also provides a corresponding matching and recommendation device for the transformation of scientific research results. The device may include a system (including a distributed system), software (application), module, component, server, client, etc., using the method described in this specification, combined with necessary hardware implementation. Based on the same innovative concept, the devices in one or more embodiments provided in this specification are as described in the following embodiments. Since the implementation schemes and methods for solving the problem by the devices are similar, the implementation of specific devices in this specification can refer to the implementation of the aforementioned method, and repeated details will not be elaborated further. As used below, the terms "unit" or "module" can refer to a combination of software and / or hardware that implements a predetermined function. Although the devices described in the following embodiments are preferably implemented in software, hardware implementation, or a combination of software and hardware, is also possible and contemplated.

[0047] Specifically, Figure 10 This is a schematic diagram of the module structure of one embodiment of a matching and recommendation device for the transformation of scientific research results provided in this specification. (Refer to...) Figure 10As shown in the embodiments of this specification, a matching and recommendation device for the transformation of scientific research results includes: The receiving module 1001 is used to receive user input information and determine the user's demand type based on the user input information; Extension module 1002 is used to semantically expand the user input information to obtain a second request when the user request type is a first request, the second request containing an extended term set; The retrieval module 1003 is used to perform a retrieval in a pre-built knowledge graph according to the second request to obtain a first retrieval result set; The calculation module 1004 is used to calculate the semantic similarity between each search result in the first search result set and the user input information, and to determine the second search result set based on the semantic similarity. The output module 1005 is used to output the second search result set to the user's corresponding terminal.

[0048] The beneficial effects obtained by the apparatus provided in the embodiments of this specification are consistent with the beneficial effects obtained by the methods described above, and will not be repeated here.

[0049] Reference Figure 11 As shown, based on the matching and recommendation method for the transformation of scientific research results described above, an embodiment of this specification also provides a computer device 1102, wherein the above method runs on the computer device 1102. The computer device 1102 may include one or more processors 1104, such as one or more central processing units (CPUs), each processing unit capable of implementing one or more hardware threads. The computer device 1102 may also include any memory 1106 for storing any kind of information such as code, settings, data, etc. Non-limitingly, for example, the memory 1106 may include any type of RAM, any type of ROM, flash memory, hard disk, optical disk, etc. More generally, any memory can use any technology to store information. Further, any memory can provide volatile or non-volatile retention of information. Further, any memory can represent a fixed or removable component of the computer device 1102. In one case, when the processor 1104 executes associated instructions stored in any memory or combination of memories, the computer device 1102 can perform any operation of the associated instructions. The computer device 1102 also includes one or more drive mechanisms 1108 for interacting with any memory, such as a hard disk drive mechanism, an optical disk drive mechanism, etc.

[0050] Computer device 1102 may also include an input / output module 1110 (I / O) for receiving various inputs (via input device 1112) and providing various outputs (via output device 1114). A specific output mechanism may include a presentation device 1116 and an associated graphical user interface (GUI) 1118. In other embodiments, the input / output module 1110 (I / O), input device 1112, and output device 1114 may be omitted, and the device may function solely as a computer device within a network. Computer device 1102 may also include one or more network interfaces 1120 for exchanging data with other devices via one or more communication links 1122. One or more communication buses 1124 couple the components described above together.

[0051] Communication link 1122 can be implemented in any way, such as via a local area network, a wide area network (e.g., the Internet), a point-to-point connection, or any combination thereof. Communication link 1122 may include any combination of hardwired links, wireless links, routers, gateway functions, name servers, etc., governed by any protocol or combination of protocols.

[0052] Corresponding to, for example Figures 2 to 9 In addition to the method shown, embodiments of this specification also provide a computer-readable storage medium storing a computer program that, when executed by a processor, performs the steps of the above-described method.

[0053] This specification also provides computer-readable instructions, wherein when a processor executes the instructions, the program therein causes the processor to perform the following... Figures 1 to 9 The method shown.

[0054] This specification also provides a computer program product, including at least one instruction or at least one program segment, wherein the at least one instruction or the at least one program segment is loaded and executed by a processor to achieve the following: Figures 1 to 9 The method shown.

[0055] It should be understood that in the various embodiments of this specification, the sequence number of each process does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this specification.

[0056] It should also be understood that, in the embodiments of this specification, the term "and / or" is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, and B existing alone. Additionally, the character " / " in this specification generally indicates that the preceding and following related objects have an "or" relationship.

[0057] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed in this specification can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of each example have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this specification.

[0058] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0059] In the several embodiments provided in this specification, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the couplings or direct couplings or communication connections shown or discussed may be indirect couplings or communication connections through some interfaces, devices, or units, or they may be electrical, mechanical, or other forms of connection.

[0060] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of the embodiments described in this specification, depending on actual needs.

[0061] Furthermore, the functional units in the various embodiments of this specification can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0062] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this specification, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this specification. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0063] This specification uses specific embodiments to illustrate the principles and implementation methods of this specification. The descriptions of the above embodiments are only for the purpose of helping to understand the methods and core ideas of this specification. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of this specification. Therefore, the content of this specification should not be construed as a limitation of this specification.

Claims

1. A matching and recommendation method for the transformation of scientific research results, characterized in that, The method includes: Receive user input information and determine the type of user request based on the user input information; When the user request type is a first request, the user input information is semantically expanded to obtain a second request, which includes an expanded terminology set. Based on the second request, a search is performed in the pre-built knowledge graph to obtain a first search result set; Calculate the semantic similarity between each search result in the first search result set and the user input information, and determine the second search result set based on the semantic similarity; The second search result set is output to the user's corresponding terminal.

2. The method according to claim 1, characterized in that, The user's demand type is determined based on the user input information, including: The user input information is converted into a structured query; Map the structured query to function parameters; The corresponding target function is determined based on the function parameters; The user requirement type is determined based on the function name corresponding to the target function call.

3. The method according to claim 2, characterized in that, The process of constructing the knowledge graph includes: Acquire knowledge data in the target domain; Extract multiple preset types of entities and semantic relationships between entities from the target domain knowledge data; Semantic mapping is performed on the domain terms associated with each extracted entity to generate a set of semantic descriptions corresponding to each domain term. The set of semantic descriptions contains semantic descriptions at multiple semantic levels. A knowledge graph is constructed based on the entities, the semantic relationships between entities, and the set of semantic descriptions.

4. The method according to claim 3, characterized in that, Semantic expansion of the user input information includes: Extract domain terms from the structured query; The domain terms are input into a large language model to generate an initial extended term set, which contains extended terms at multiple semantic levels. Each extended term in the initial extended term set is semantically validated in the knowledge graph; The extended terms that pass semantic validation will be used as the extended term set for the corresponding domain terms.

5. The method according to any one of claims 3 or 4, characterized in that, The semantic level includes at least the professional terms corresponding to the domain terminology, the standard industry terms corresponding to the professional terms, and the colloquial expressions corresponding to the standard industry terms.

6. The method according to claim 3, characterized in that, Retrieval in a pre-built knowledge graph according to the second request includes: Determine the terminology set corresponding to the user input information, the terminology set including domain terms and extended terminology sets; If any term in the term set has a matching result in the knowledge graph, it is used as a search term; If any term in the term set does not have a matching result in the knowledge graph, then one or more semantically similar terms are determined from the knowledge graph as retrieval items; The search is performed on the knowledge graph according to the search terms, and the retrieved entity information is used to form a first search result set.

7. The method according to claim 1, characterized in that, Before outputting the second search result set to the user's corresponding terminal, the method further includes: For each search result in the second search result set, verify whether there is a tracing path in the knowledge graph that is associated with the user input information; The search results of the tracing paths associated with the user input information are used to form a candidate result set; Calculate the confidence score of each candidate result in the candidate result set relative to the user input information. The confidence score is used to reflect the degree to which the search results solve the user's needs. If the confidence score is lower than the preset threshold, the step of determining the user's need type based on the user input information is repeated. If the step of determining the user's need type based on the user input information is repeated more than a preset number of times, the retrieval process will be terminated and a preset prompt message will be output.

8. A matching and recommendation device for the transformation of scientific research results, characterized in that, The device includes: A receiving module is used to receive user input information and determine the type of user request based on the user input information; An extension module is used to semantically extend the user input information to obtain a second request when the user request type is a first request, wherein the second request contains an extended terminology set; The retrieval module is used to perform a retrieval in a pre-built knowledge graph according to the second request and obtain a first retrieval result set; The calculation module is used to calculate the semantic similarity between each search result in the first search result set and the user input information, and to determine the second search result set based on the semantic similarity. The output module is used to output the second search result set to the user's corresponding terminal.

9. A computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the method as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, implements the method as described in any one of claims 1 to 7.

11. A computer program product, characterized in that, It includes at least one instruction or at least one program segment, said at least one instruction or said at least one program segment being loaded and executed by a processor to implement the method as claimed in any one of claims 1 to 7.

Images