Cross-category scientific research achievement intelligent registration method

Abstract

The invention discloses a cross-category scientific research achievement intelligent registration method, which comprises the following steps: a, data acquisition: acquiring an achievement text, and preprocessing the achievement text; b, data processing and storage: calculating a hash value of the preprocessed data, generating a digital fingerprint and a semantic vector, and performing block chain evidence storage; c, performing data analysis, and performing registration number discrimination, fingerprint comparison and cross-category semantic comparison; d, constructing a knowledge graph, constructing an achievement graph which takes achievements as nodes and takes similarity / derivation / time sequence as edges, and labeling a version chain and a time sequence chain; e, performing multi-dimensional scoring, performing scoring based on the knowledge graph and the block chain evidence storage data, and performing sorting output; according to the method, deep fusion and semantic-level duplicate checking of multi-source achievements are realized by constructing a unified achievement processing and traceability architecture fusing semantic fingerprints and block chain double-track evidence storage, and the capabilities of an achievement evolution path, representativeness, innovativeness and influence are improved by constructing a dynamic knowledge graph and combining a multi-dimensional scoring mechanism.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, specifically to a method for intelligent registration of scientific research results that can be applied across different categories. Background Technology

[0002] In the context of increasingly interdisciplinary and collaborative scientific research and innovation activities, the forms and sources of scientific research achievements are becoming more diverse. Besides traditional academic papers, patents, software copyrights, technical standards, and government science and technology reports have all become important knowledge outputs. However, existing scientific research achievement registration systems generally suffer from insufficient systematicity and struggle to cope with this complex situation. These systems are typically designed for single-category achievements, often using registration numbers or names for deduplication, making it difficult to identify deep technical connections and semantic evolutionary relationships between cross-category achievements (such as patents, software copyrights, and government projects). This results in a lack of unified clustering analysis and deduplication mechanisms when integrating achievement data from different sources, leading to duplication, fragmentation, and inconsistencies in macro-level summaries and presentations, making it difficult to form a coherent knowledge framework.

[0003] Furthermore, current evaluation and ranking mechanisms for research achievements are mostly limited to a single dimension, such as relying solely on journal impact factors or citation counts. They fail to comprehensively consider the decay effect over time (i.e., "freshness") and the relative position of the achievement within its similar cluster (i.e., "representativeness"). They also lack the organic integration of multi-dimensional indicators such as metadata integrity, social impact, and source credibility. Overall, they are unable to comprehensively and fairly reflect the comprehensive value of research achievements and cannot provide refined data support for research management decisions. Moreover, existing schemes have significant deficiencies in data credibility assurance and traceability capabilities, making achievement data easily tampered with and the modification history impossible to trace. They lack an immutable and credible evidence preservation mechanism and authoritative dissemination evidence, making them lack credibility when used for awards, promotion, or dispute arbitration. Summary of the Invention

[0004] The purpose of this invention is to provide a cross-category intelligent registration method for scientific research achievements that can achieve deep integration of multi-source achievements, credible semantic tracing, dynamic knowledge graph construction, and multi-dimensional intelligent evaluation. This method aims to solve the technical problems of existing achievement registration systems, such as difficulty in integrating cross-category achievements, lack of semantic association, weak credible evidence storage, and the single dimension of traditional evaluation mechanisms that cannot reflect the evolution path and comprehensive value of achievements due to insufficient systematicity.

[0005] To solve the above technical problems, the present invention adopts the following solution: A method for intelligent registration of scientific research achievements across categories includes the following steps: a. Data collection: Obtain the texts of patents, software copyrights, and completed government projects, and preprocess them. Extract key content such as title, author, abstract, and main text from user-submitted documents or texts, standardize them, and perform operations such as word segmentation and stop word removal on the text to prepare for the subsequent generation of semantic vectors. b. Data processing and storage: Calculate the hash value of the preprocessed data to generate a digital fingerprint and use an AI model to transform the data content into a semantic vector representing its meaning and store it on the blockchain; bind the digital fingerprint, semantic vector and result metadata summary hash and write it into the blockchain to form an immutable semantic traceability chain. c. Data analysis: Perform registration number discrimination, fingerprint comparison and cross-category semantic comparison to determine the similarity between the data and the results in the database, and conduct semantic-level plagiarism detection and innovation analysis so that texts with completely different wording but similar core ideas can also be detected, which is convenient for subsequent knowledge graph construction and scoring. d. Construct a knowledge graph, using achievements as nodes and similarity / derivation / chronological order as edges, and label version chains and chronological chains; establish connections based on semantic vector similarity, with higher similarity resulting in greater edge "weight"; derived edges indicate that one achievement is derived from another, or explicitly cites / references another achievement; chronological edges establish connections according to the publication time of achievements, showcasing the development trajectory of the field; version chains refer to multiple versions of the same research (such as preprints, conference papers, journal articles), linking them together to form a clear version evolution chain; chronological chains refer to linking key achievements in chronological order under a specific research topic, showcasing the research evolution of that topic.

[0006] e. Multidimensional scoring: Based on the knowledge graph in step d and the blockchain-stored data in step b, scoring is performed and output in chronological order. The scoring types include freshness / innovation score (S_new(i)) and representativeness score (S_rep(i)), and can be expanded to include completeness score (S_comp), impact score (S_impact), and credibility score (S_cred). The completeness score S_comp, impact score S_impact, and credibility score S_cred are all normalized to [0,1]. The output score after scoring should contain at least two types. The freshness / innovation score is used to evaluate the novelty and innovation of the results. The representativeness score is used to evaluate the importance and typicality of the results in their research field or cluster. It also evaluates the completeness of the results' metadata (authors, institutions, references, etc.). The impact score refers to the number of traditional citations or the number of connections to other important nodes in the graph. The credibility score includes whether it has been peer-reviewed, whether it has been included by an authoritative institution, and whether the blockchain-stored information is complete.

[0007] Furthermore, step a includes intelligent data entry assistance; this assistance includes automatically predicting categories, marking potential duplication and inheritance relationships, automatically extracting project numbers / related documents and completing fields, and performing incremental on-chain evidence storage when an update is triggered. The intelligent data entry assistance is mainly used for conflict / duplication checks and field completion. When entering results, based on the already entered content and semantic analysis, it automatically recommends or completes other field information and prompts whether the results are duplicated or highly similar to existing results in the database, thus avoiding duplicate data entry.

[0008] Furthermore, the intelligent data entry assistance system features a comprehensive visualization interface. This interface outputs clustering, ranking, and multi-dimensional dashboard views for computer display and retrieval, and provides visualization of the evidence storage chain and evolutionary path to enable traceability and auditing. The clustering view automatically groups similar results into a category, forming a view of research hotspots. The ranking view displays the results list sorted by different dimensions such as time, score, and influence. The multi-dimensional dashboard view, like a dashboard, comprehensively displays various statistical information and analysis results. The visualization of the evidence storage chain and evolutionary path can intuitively display the evidence storage information of the results on the blockchain, as well as the related "version chain" and "time sequence chain."

[0009] Furthermore, The freshness / innovation score (S_new(i)) in step e is calculated based on a combination of time decay factor, similarity suppression term, and ranking decay term. The overall calculation formula is as follows: f_time(i)=exp(-λΔt_i),λ=ln2 / τ_{1 / 2} p_sim(i)=1-β*(max(similarity(i)))^γ p_rank(i)=1 / (1+α(rank-1))^η S_new(i)=clip(f_time(i)*p_sim(i)*p_rank(i),0,1) In the formula, f_time(i) is the time decay factor, λ is the decay coefficient, and Δt_i is the time difference between the current time and the publication time of result i; p_sim(i) is the originality factor, and max(similarity(i)) represents the maximum semantic similarity between result i and its previous results; p_rank(i) is the ranking decay factor, and rank represents the quality ranking of result i within its similarity cluster; clip(x,0,1) is the cutoff function, which takes values ​​at the endpoints when the result exceeds the range [0,1]; parameters β, γ, α, and η are adjustment parameters with values ​​in the range (0,1), used to control the degree of influence of each factor.

[0010] Furthermore, The representative score (S_rep(i)) in step e is calculated based on the centrality index or vector density metric in the graph. It is used to measure the representativeness and coverage of a result within similar clusters. The calculation formula is as follows: S_rep(i)=norm(C(i)) In the formula, C(i) is the centrality index of result i in the knowledge graph, which can be the average similarity, degree centrality, betweenness centrality or eigenvector centrality; norm(x) is the normalization function used to map the result to the interval [0,1].

[0011] Furthermore, The comprehensive evaluation score S_total output in step e is calculated using a weighted linear synthesis method, as shown in the following formula: S_total=w_new*S_new+w_rep*S_rep+w_comp*S_comp+w_impact*S_impact+w_cred*S_cred In the formula, w_new, w_rep, w_comp, w_impact, and w_cred are weights, set by relevant domain experts, to reflect the importance of different dimensions in the final evaluation; all weights satisfy the constraint w_new+w_rep+w_comp+w_impact+w_cred=1; the scoring result S_total is normalized to [0,1] and output in chronological order to generate the final ranking result.

[0012] Furthermore, the blockchain-based evidence stored in step b includes a registration number or project number, a digest hash, a semantic fingerprint, a model version, and a timestamp, which are used to ensure that the data is tamper-proof and traceable throughout the entire chain.

[0013] Furthermore, in step c, the cross-category semantic comparison calculates the cosine similarity between the resulting semantic fingerprints based on a vector space model. The calculation formula is as follows: Sim(A,B)=(VA·VB) / (‖VA‖‖VB‖) In the formula, VA and VB represent the semantic vectors corresponding to result A and result B, respectively. The symbol "·" represents the vector dot product. ‖VA‖ and ‖VB‖ represent the magnitude of the corresponding vectors. The similarity value ranges from [0,1]. The larger the value, the closer the semantic content of the two results is.

[0014] When mapping the regional industrial chain, it is also necessary to calculate and output the industrial fit score S_fit(i). The calculation formula is: S_fit(i)=norm(Sim(A,B)), which is used to evaluate the degree of matching and transformation potential between scientific research results and market demand, industrial application prospects, etc., where norm(·) is a normalization function to map the result to [0,1].

[0015] The beneficial effects of this invention are as follows: This invention effectively solves the problems of semantic conflicts and data credibility assurance across categories of research results by constructing a unified research results processing and traceability architecture that integrates semantic fingerprinting and blockchain dual-track notarization. It achieves deep integration and semantic-level deduplication of multi-source results such as patents, software copyrights, and scientific and technological reports. By constructing a dynamic knowledge graph with temporal, derivative, and similarity relationships as its core, and combining it with a multi-dimensional scoring mechanism that integrates time decay, similar cluster suppression, and graph centrality calculation, it significantly improves the comprehensive evaluation capability of the evolution path, representativeness, innovation, and influence of research results, overcoming the limitations of traditional single-dimensional evaluation. Ultimately, it realizes cross-category integration, credible notarization, intelligent analysis, and industry mapping of scientific research results, providing comprehensive, accurate, and reliable data support for scientific research management and decision-making. Attached Figure Description

[0016] Figure 1 This is a schematic diagram of the cross-category achievement registration process structure of the present invention; Figure 2 This is a schematic diagram of the version chain identification process structure of the present invention; Figure 3 This is a schematic diagram of the progressive relationship identification process structure of the present invention; Figure 4 This is a schematic diagram of the regional industrial chain mapping of the present invention; Figure 5 This is a schematic diagram of the traceability process of the results of this invention. Detailed Implementation

[0017] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0018] The following is a detailed description of the intelligent registration method for cross-category scientific research achievements according to the present invention, with reference to the accompanying drawings: A method for intelligent registration of scientific research achievements across categories includes the following steps: a. Data collection: Obtain the results texts of patents, software copyrights, and completed government projects, and perform preprocessing. b. Data processing and storage: Calculate the hash value of the preprocessed data, generate digital fingerprints and semantic vectors, and store them on the blockchain. c. Data analysis, including registration number identification, fingerprint comparison, and cross-category semantic comparison, to determine the similarity between the data and the results in the database; d. Construct a knowledge graph, building a graph with results as nodes and similarity / derivation / temporal sequence as edges, and labeling version chains and temporal sequence chains; e. Multidimensional scoring: Based on the knowledge graph in step d and the blockchain-stored data in step b, scores are performed and output in chronological order. The scoring types include freshness / innovation score (S_new(i)) and representativeness score (S_rep(i)), and can be expanded to include completeness score (S_comp), impact score (S_impact) and credibility score (S_cred).

[0019] Step a includes intelligent data entry assistance; the intelligent data entry assistance includes automatically predicting categories, marking potential duplication and inheritance relationships, automatically extracting project numbers / related documents and completing fields, and performing incremental on-chain evidence storage when an update is triggered.

[0020] The intelligent data entry assistant has a comprehensive visualization interface; the comprehensive visualization interface outputs clustering, sorting and multi-dimensional dashboard views for computer display and retrieval, and provides visualization of evidence storage chain and evolution path to achieve traceability and auditing.

[0021] The freshness / innovation score (S_new(i)) in step e is calculated based on a combination of time decay factor, similarity suppression term, and ranking decay term. The overall calculation formula is as follows: f_time(i)=exp(-λΔt_i),λ=ln2 / τ_{1 / 2} p_sim(i)=1-β*(max(similarity(i)))^γ p_rank(i)=1 / (1+α(rank-1))^η S_new(i)=clip(f_time(i)*p_sim(i)*p_rank(i),0,1) In the formula, f_time(i) is the time decay factor, λ is the decay coefficient, and Δt_i is the time difference between the current time and the publication time of result i; p_sim(i) is the originality factor, and max(similarity(i)) represents the maximum semantic similarity between result i and its previous results; p_rank(i) is the ranking decay factor, and rank represents the quality ranking of result i within its similarity cluster; clip(x,0,1) is the cutoff function, which takes values ​​at the endpoints when the result exceeds the range [0,1]; parameters β, γ, α, and η are adjustment parameters with values ​​in the range (0,1), used to control the degree of influence of each factor.

[0022] The representative score (S_rep(i)) in step e is calculated based on the centrality index or vector density metric in the graph. It is used to measure the representativeness and coverage of a result within similar clusters. The calculation formula is as follows: S_rep(i)=norm(C(i)) In the formula, C(i) is the centrality index of result i in the knowledge graph, which can be the average similarity, degree centrality, betweenness centrality or eigenvector centrality; norm(x) is the normalization function used to map the result to the interval [0,1].

[0023] The comprehensive evaluation score S_total output in step e is calculated using a weighted linear synthesis method, as shown in the following formula: S_total=w_new*S_new+w_rep*S_rep+w_comp*S_comp+w_impact*S_impact+w_cred*S_cred In the formula, w_new, w_rep, w_comp, w_impact, and w_cred are weights, set by relevant domain experts, to reflect the importance of different dimensions in the final evaluation; all weights satisfy the constraint w_new+w_rep+w_comp+w_impact+w_cred=1; the scoring result S_total is normalized to [0,1] and output in chronological order to generate the final ranking result.

[0024] The blockchain-based evidence stored in step b includes a registration number or project number, a digest hash, a semantic fingerprint, a model version, and a timestamp, which are used to ensure that the data is tamper-proof and traceable throughout the entire chain.

[0025] In step c, the cross-category semantic comparison calculates the cosine similarity between the resulting semantic fingerprints based on a vector space model. The calculation formula is as follows: Sim(A,B)=(VA·VB) / (‖VA‖‖VB‖) In the formula, VA and VB represent the semantic vectors corresponding to result A and result B, respectively. The symbol "·" represents the vector dot product, and ||VA|| and ||VB|| represent the magnitudes of the corresponding vectors. The similarity value ranges from [0,1], with a larger value indicating that the semantic content of the two results is closer. Cross-category semantic comparison can identify results with different wording but highly similar technical cores, which may be different expressions or slight improvements of the same technology, or results with logical derivation relationships, such as a project report leading to a patent. "Cross-unit, cross-text, and cross-category" judgments can be made to determine whether they are different versions or evolutionary paths of the same technological achievement. Example 1

[0026] When using this invention for cross-category achievement registration, the following methods are employed: Figure 1 The registration process, as shown, begins with researchers submitting materials to obtain the results (project reports, software copyrights, patent documents). The registration summary states: In 2022, the research unit completed the major project "Integration and Demonstration of Key Technologies for Smart Agriculture in the Autonomous Region," with innovation point A being the integrated demonstration of multi-source sensors and a cloud platform. In 2023, the unit registered "Agricultural Traceability Data Governance Platform," with innovation point B being blockchain-based evidence storage and automated data governance. In 2024, the unit was granted the patent "On-Chain Governance Method for Agricultural Product Traceability," with innovation point C being cross-node consensus and smart contract verification mechanisms.

[0027] Preprocessing: The system extracts keywords such as "smart agriculture", "data governance", and "on-chain governance"; and performs word segmentation, stop word removal, and standardization on the text.

[0028] Data processing is performed, the hash value of the preprocessed abstract is calculated, a digital fingerprint is generated, and an AI model is used to transform the data content into a semantic vector representing its meaning. Information such as the result number, abstract hash, semantic vector fingerprint, model version and timestamp are stored on the blockchain to form an immutable traceability record.

[0029] The data of the three achievements were analyzed, and semantic comparisons were performed across categories. The similarity was calculated using the vector cosine similarity formula: Sim(A,B)=(VA·VB) / (‖VA‖‖VB‖). The similarity was: project and software copyright ≈ 0.78, project and patent ≈ 0.72, software copyright and patent ≈ 0.84. A semantic chain of "project → software copyright → patent" was established and an evolutionary time chain was formed.

[0030] A knowledge graph system is constructed, with results as nodes and semantic similarity relationships and temporal order as edges. The system automatically labels the relationship between version chains and temporal chains to obtain a "temporal evolution graph".

[0031] Multidimensional scoring is performed based on the knowledge graph in step d and the blockchain-stored data in step b. The scoring characteristics are high patent representativeness and low project novelty. Based on the calculation formula of this invention, the parameters are set as λ=ln2 / τ1 / 2 (τ1 / 2=3 years), β=0.5, γ=1, α=0.5, η=1, and the year is calculated based on the current year (2025). The novelty score is f_time(i)=e^{-λΔt_i}; the representativeness score is S_rep(i)=(C(i)-Cmin) / (Cmax-Cmin) → (0.0, 0.38, 1.0); the comprehensive score is S_total=w_new*S_new+w_rep*S_rep, where w_new=0.4, w_rep=0.6; C(project)=0.65, C(software copyright)=0.70. C(patent) = 0.82, which, after normalization, are 0.65, 0.70, and 0.82 respectively.

[0032] Substituting the values ​​into the calculation, the scoring results are as follows: Project freshness S_new(i)≈0.25, representativeness S_rep(i)≈0.65, and overall score S_total≈0.54; Software copyright freshness S_new(i)≈0.20, representativeness S_rep(i)≈0.70, and overall score S_total≈0.56; Patent freshness S_new(i)≈0.23, representativeness S_rep(i)≈0.82, and overall score S_total≈0.60.

[0033] Results with a comprehensive score of 0.6 or higher are recommended as "cluster representative results" for display and are output through a comprehensive visualization interface. The output dashboard mainly displays the evolutionary chain, score table, and recommendation tags; the score table content is as follows. Results Freshness score representative score Overall score Recommended level project 0.25 (Low) 0.65 (Medium) 0.54 (Not recommended) Only retain the timing chain Soft copyright 0.20 (Low) 0.70 (Medium) 0.56 (Not recommended) Auxiliary display patent 0.23 (Low) 0.82 (High) 0.60 (Recommended) Cluster representative results The registration results in this embodiment are synchronously mapped to the corresponding regional industrial chain map nodes, and aggregated and displayed with the results of the same theme at the nodes, which facilitates comparison and promotion from the perspective of industrial chain links. The dashboard displays cluster representatives and evolution chains. Example 2

[0034] When using this invention for version chain identification, the following method is employed: Figure 2 The registration process shown first involves researchers submitting materials to obtain the original and improved versions of the software copyright documents. The registration abstract is as follows: Registered in 2021 as "Intelligent Image Processing Platform," with the innovation being the integration of image acquisition and classification functions. Registered in 2023 as "Intelligent Image Processing Platform (Improved Version)," with the innovation being the addition of deep learning enhancement and automatic annotation mechanisms.

[0035] Data processing is performed, including calculating hash values ​​for preprocessed summaries, generating digital fingerprints, and using AI models to transform data content into semantic vectors representing its meaning, which are then stored on the blockchain. The data is analyzed and the similarity is calculated; if the similarity is approximately 0.88, the system identifies it as a version relationship and establishes a version chain. Multidimensional scoring, based on the knowledge graph in step d and the blockchain-stored data in step b, is used to score the data. The scoring characteristics are high representativeness of the original version and high freshness of the improved version. The scoring results are as follows: Original version freshness ≈ 0.18, representativeness ≈ 0.80, overall score ≈ 0.55; Improved version freshness ≈ 0.40, representativeness ≈ 0.60, overall score ≈ 0.57. The score table is shown below: Results Freshness score representative score Overall score Recommended level Original 0.18 (Low) 0.80 (High) 0.55 (Not recommended) Cluster representative results Improved version 0.40 (Medium) 0.60 (Medium) 0.57 (Not recommended) Auxiliary display This embodiment systematically locates the results to the corresponding links in the industrial chain and aggregates and displays them with similar results, with representative annotations for horizontal comparison. The dashboard displays the original version plus the improved chain, and the calculation formula is the same as in Embodiment 1. Example 3

[0036] When using this invention to identify progressive relationships, the following methods are employed: Figure 3 The registration process shown first involves researchers submitting materials to obtain two patent documents. The registration abstracts are as follows: A 2020 patent, "Energy Monitoring Method Based on the Internet of Things," is authorized, with its innovation being IoT terminal data collection and cloud uploading. A 2023 patent, "Energy Monitoring Method Based on the Internet of Things and Blockchain," is authorized, with its innovation being the introduction of blockchain evidence storage and cross-node sharing mechanisms.

[0037] Data processing is performed, including calculating hash values ​​for preprocessed summaries, generating digital fingerprints, and using AI models to transform data content into semantic vectors representing its meaning, which are then stored on the blockchain. The data is analyzed, and semantic comparisons are performed across categories to calculate similarity: if the similarity is ≈0.82, the system identifies it as progressive innovation and establishes derivative edges. Multidimensional scoring, based on the knowledge graph in step d and the blockchain-stored data in step b, is used to score the patents. The scoring characteristics are high representativeness of the basic patents and high freshness of the progressive patents. The scoring results are as follows: Basic patent freshness ≈ 0.12, representativeness ≈ 0.82, overall score ≈ 0.58; Progressive patent freshness ≈ 0.55, representativeness ≈ 0.70, overall score ≈ 0.62. The score table is shown below: Results Freshness score representative score Overall score Recommended level Basic Patents 0.12 (Low) 0.82 (High) 0.58 (Not recommended) Cluster representative results Progressive patents 0.55 (Medium) 0.70 (Medium) 0.62 (Recommended) Recommended display The version chain and derivative chain generated in this embodiment form a visual chain segment on the industry chain map, which supports the display and summary of each link. The dashboard displays the basic patent root node + progressive patents, and the calculation formula is the same as in embodiment 1. Example 4

[0038] This invention is used to illustrate the application and scoring calculation method of the present invention in the context of regional industrial chain mapping.

[0039] This embodiment demonstrates how the system, based on blockchain-stored evidence data, uses semantic vector analysis and industry fit calculation to achieve automatic matching and quantitative evaluation of scientific research results and industry chain nodes.

[0040] When using this invention for regional industrial chain mapping, the system first retrieves the registered achievements from the blockchain-stored data. Based on existing achievements registrations, the system extracts the registered achievements information from various research institutions within the Guangxi Zhuang Autonomous Region and maps these achievements to advantageous industrial chain nodes in Guangxi. Taking the aluminum industry as an example, the system automatically clusters achievements by region and technological theme, forming a multi-level industrial chain node structure: achievements registered in Baise mainly focus on upstream bauxite mining and resource management; achievements in Nanning lean towards midstream alumina and electrolytic aluminum process optimization; achievements in Liuzhou cover deep processing and recycled aluminum applications; and achievements in Beihai and Fangchenggang focus more on port logistics and export links. (See Table 1 below) hierarchy Node type Typical areas Results direction Upstream Bauxite Mining and Resource Management Baise Mineral resource monitoring and forecasting Midstream Alumina and electrolytic aluminum process optimization Nanning Electrolysis process and energy saving optimization Downstream Deep processing and recycled aluminum applications Liuzhou Aluminum product manufacturing and recycling exit Port logistics and foreign trade nodes Beihai, Fangchenggang Intelligent warehousing and export management Table 1 System Processing: The system first constructs a "Guangxi Aluminum Industry Chain Map," which mainly includes a three-layer relationship structure of "region-industry-outcome." It also includes four levels: upstream (bauxite), midstream (alumina / electrolytic aluminum), downstream (deep processing / recycling), and ports / foreign trade. Aggregated outcome data for each region is embedded into the map nodes, and semantic association edges are established between each outcome and industry node. The system automatically completes the mapping between "outcome data" and "industry chain nodes."

[0041] Scoring Calculation: The system provides scores for freshness, representativeness, and industry fit to measure the value of the achievement at the corresponding industry chain node. Example Data: Using λ=ln2 / 3≈0.231, β=0.5, γ=1, α=0.5, η=1; a mineral resource management patent (achievement X) from the Baise node, titled "Intelligent Management Method for Mineral Resources Based on Remote Sensing," has the following semantic matching result with the industry node "Mineral Resource Management": Sim(A,B)=(VA·VB) / (‖VA‖‖VB‖)=0.90, industry fit score: S_fit(i)=norm(Sim(A,B))=0.90, centrality C(i)=0.70, freshness S_new(i)=0.25, and representativeness S_rep(i)=0.3. 8. With a comprehensive score S_total = 0.68, it is strongly recommended as a "strongly recommended result in the regional industrial chain." A certain result from the Liuzhou node, "Control System V2.0 for Recycled Aluminum Processing Equipment" (result Y), has the following similarity with the corresponding industry node: Sim(A,B) = 0.76, industry fit score: S_fit(i) = 0.76, centrality C(i) = 0.65, freshness S_new(i) = 0.23, representativeness S_rep(i) = 0, and a comprehensive score S_total = 0.53. It is generally recommended as a "strongly recommended result in the regional industrial chain." Typical values ​​for the weighted comprehensive formula are w_new = 0.3, w_rep = 0.3, and w_fit = 0.4. The system makes recommendations based on whether the comprehensive score is ≥ 0.65 (as shown in Table 2 below). Results with a score ≥ 0.65 are marked as "strongly recommended results in the regional industrial chain" and highlighted in the visualization map.

[0042] Results S_new S_rep S_fit S_tota Recommended level X 0.25 0.38 0.90 0.68 Highly recommended Y 0.23 0 0.76 0.53 General Recommendation Table 2 Displayed through a comprehensive visual interface, such as Figure 4 The map showing the Guangxi aluminum industry chain uses different colors to mark upstream, midstream, downstream, and port / foreign trade nodes. It intuitively displays the distribution of various nodes and achievements in the regional industry chain. Each node summarizes the number of patents, software copyrights, and projects, and users can click to view representative achievements. The system combines "representativeness score," "novelty score," and "industry fit score" to form a regional achievement promotion dashboard, providing intuitive basis for the promotion of scientific research results and industry docking.

[0043] The results of this embodiment are displayed in nodes by using a regional industrial chain map. Users can expand representative achievement cards at the nodes and perform comparisons and recommendations. Example 5

[0044] When using this invention for tracing the source of results, the following methods are employed: Figure 5The registration process shown first extracts the materials and documents of three achievements, and registers the following abstracts: Achievement A (registered in 2023): "Research and Industrialization of Key Technologies for Intelligent Integrated Domestic Sewage Treatment Equipment," with innovations in high-performance PVDF hollow fiber membrane preparation and hydrophilic modification technology; Achievement B (registered in 2023): "A Sewage Treatment Equipment Based on Prison BIM Technology," with innovations in adjustable pore size filtration structure and BIM integrated monitoring system; and Achievement C (registered in 2024): "Intelligent Remote Network Management, Monitoring and Maintenance Control System V1.0 for Sewage Treatment Facility Operation," with innovations in a distributed intelligent management and control platform for sewage treatment based on 5G+IoT. Pre-processing is then performed, extracting keywords such as "sewage treatment," "intelligent monitoring," "remote management," "BIM technology," and "IoT platform."

[0045] Data processing involves calculating hash values ​​from preprocessed summaries, generating digital fingerprints, and using AI models to transform data content into semantic vectors representing its meaning. These vectors are then stored on the blockchain to form an immutable semantic traceability chain.

[0046] Data analysis, including cross-category semantic comparison, yielded similarity scores: Result A and Result B = 0.73, Result B and Result C = 0.79, and Result C and Result A = 0.68. These results indicate that although the three results differ in registration time and description, their core technologies exhibit significant inheritance and evolution, forming a progressive chain from key material development to equipment integration and intelligent operation management. The system can further automatically identify the highest similarity between Result B and Result C, suggesting they may be continuous applications or functional extensions of the same technological system. The association between Result A and Result B indicates that the latter integrates equipment-level applications based on the former's research and development achievements. Result C, on the other hand, superimposes a 5G remote intelligent monitoring module onto a BIM equipment solution, representing a higher-level system-level derivative innovation.

[0047] Based on the semantic fingerprint, a multi-dimensional evolutionary graph of achievements is constructed, with achievements as nodes and semantic similarity edges, temporal evolution edges, and derivative relationship edges as edges. This outlines a technological evolution path of "basic method (achievement A) → first application (achievement B) → suspected derivative / repeated application (achievement C)". The weight of similarity edges is determined by semantic vector similarity, and derivative edges are inferred by the system based on semantic relevance and temporal order. The temporal chain shows the connection between each achievement in terms of time and technology. The system ultimately generates a visualized view of the "intelligent evolutionary chain of wastewater treatment". Nodes are automatically arranged according to time and similarity, and the chain path is as follows: achievement A (key material research and development) → achievement B (BIM equipment integration) → achievement C (intelligent remote control). A timestamp and semantic hash identifier are attached to each node on the blockchain to achieve traceability of the achievement evolution process and data immutability.

[0048] Finally, it should be noted that the formula substitution and result display of the above embodiments demonstrate that the scoring method of the present invention can perform quantitative calculations under clearly defined parameter conditions, and the results are consistent and repeatable, meeting the requirement of "sufficient disclosure" in Article 26 of the Implementing Regulations of the Patent Law. Furthermore, the applicant has already conducted internal project planning and experimentation, and it is expected that it can be applied to existing systems in the near future. The above descriptions are merely preferred embodiments of the present invention and are not intended to limit the scope of protection of the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some technical features without departing from the spirit and principles of the present invention. Any modifications, equivalent substitutions, or improvements made within the principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for intelligent registration of scientific research achievements across categories, characterized in that: The method comprises the following steps: a. Data collection: collect the achievement texts of patents, software copyrights and completed government projects, and preprocess them; b. Data processing and storage: calculate the hash value of the preprocessed data, generate digital fingerprints and semantic vectors, and store them in the blockchain; c. Data analysis: perform registration number discrimination, fingerprint comparison and cross-category semantic comparison to determine the similarity between the data and the achievements in the database; d. Building a knowledge graph: build an achievement graph with achievements as nodes and similarity / derivation / chronology as edges, and label the version chain and chronology chain; e. Multi-dimensional scoring: based on the knowledge graph in step d and the data stored in the blockchain in step b, score and time-sort the output, including freshness / innovation score (S_new(i)) and representativeness score (S_rep(i)), and can be extended to completeness score (S_comp), impact score (S_impact) and credibility score (S_cred).

2. The cross-category scientific research achievement intelligent registration method according to claim 1, characterized in that: The step a is provided with intelligent reporting assistance; the intelligent reporting assistance includes automatic prediction of categories, labeling of potential duplication and inheritance, automatic extraction of project numbers / correlation documents and field completion, and incremental on-chain storage when triggered. 3.The cross-category scientific research achievement intelligent registration method according to claim 2, characterized in that: The intelligent reporting assistance is provided with a comprehensive visualization interface; the comprehensive visualization interface outputs clustering, sorting and multi-dimensional dashboard views for computer display and retrieval, and provides storage chain and evolution path visualization to realize traceability and auditing.

4. The cross-category scientific research achievement intelligent registration method of claim 1, wherein: The freshness / innovation score (S_new(i)) in step e is calculated based on the combination of time decay factor, similarity suppression term and ranking decay term, and the overall calculation formula is as follows: f_time(i)=exp(-λΔt_i), λ=ln2 / τ_{1 / 2} p_sim(i)=1-β*(max(similarity(i)))^γ p_rank(i)=1 / (1+α(rank-1))^η S_new(i)=clip(f_time(i)*p_sim(i)*p_rank(i),0,1) In the formula, f_time(i) is the time decay factor, λ is the decay coefficient, Δt_i is the time difference between the current time and the publication time of achievement i; p_sim(i) is the originality factor, max(similarity(i)) represents the maximum semantic similarity between achievement i and its previous achievements; p_rank(i) is the ranking decay factor, rank represents the quality ranking of achievement i in its similar cluster; clip(x,0,1) is a clipping function that takes the endpoint value when the result exceeds the range [0,1]; the parameters β, γ, α, η are adjustment parameters with a value range of (0,1), used to control the influence degree of each factor.

5. The cross-category scientific research achievement intelligent registration method according to claim 4, characterized in that: The representativeness score (S_rep(i)) in step e is calculated based on the centrality index or vector density measure in the graph, which is used to measure the representative degree and coverage ability of an achievement in a similar cluster, and the calculation formula is as follows: S_rep(i)=norm(C(i)) In the formula, C(i) is the centrality index of achievement i in the knowledge graph, which can be average similarity, degree centrality, betweenness centrality or eigenvector centrality; norm(x) is a normalization function for mapping the result to the interval [0, 1].

6. The cross-category scientific research achievement intelligent registration method according to claim 5, characterized in that: The comprehensive evaluation score S_total of the time sequence ranking output in step e is calculated by using a weighted linear comprehensive method, and the specific formula is as follows: S_total=w_new*S_new+w_rep*S_rep+w_comp*S_comp+w_impact*S_impact+w_cred*S_cred In the formula, w_new, w_rep, w_comp, w_impact and w_cred are weights, which are set by relevant field experts and are used to reflect the importance of different dimensions in the final evaluation; all weights satisfy the constraint condition w_new+w_rep+w_comp+w_impact+w_cred=1; the score result S_total is normalized to [0, 1], and is output in time sequence to generate a final ranking result.

7. The cross-category scientific research achievement intelligent registration method according to claim 1, characterized in that: The blockchain storage content in step b includes a registration number or project number, an abstract hash, a semantic fingerprint, a model version and a timestamp, which are used to ensure that the data is not tamperable and is traceable throughout the link.

8. The cross-category scientific research achievement intelligent registration method according to claim 1, characterized in that: The cross-category semantic comparison in step c calculates the cosine similarity between the semantic fingerprints of the achievements based on a vector space model, and the calculation formula is as follows: Sim(A,B)=(VA·VB) / (‖VA‖‖VB‖) In the formula, VA and VB represent the semantic vectors corresponding to achievements A and B respectively, the symbol "·" represents vector dot product, ‖VA‖ and ‖VB‖ represent the lengths of the corresponding semantic vectors, the similarity value ranges from 0 to 1, and the larger the value is, the closer the semantic contents of the two achievements are.

Images