A method for identifying a breakthrough innovation theme based on scientific-technological association

Abstract

The application provides a breakthrough innovation theme identification method based on scientific-technological association, filters and judges scientific and patent paper data, and constructs scientific and technological potential data sets; a variety of text-to-vector pre-training models are used for theme identification to obtain a scientific and technological theme set; a scientific and technological breakthrough innovation measurement model is established, a plurality of preset scientific dimensions are evaluated and identified, and a scientific and technological breakthrough innovation theme is obtained; based on the Helling distance, the semantic correlation between the scientific and technological breakthrough innovation themes is calculated; and based on the Pasteur scientific quadrant theory and the high and low of semantic correlation, the identified scientific and technological breakthrough innovation themes are classified. The application fuses scientific papers and patent data, filters potential and enhances semantic theme identification. The problems of single data source, large noise interference and poor interpretability are effectively solved, and the identification accuracy and decision value are improved.

Description

Technical Field

[0001] This invention relates to the fields of data mining and technological innovation management, and in particular to a method for identifying breakthrough innovation topics based on science-technology correlation. Background Technology

[0002] Breakthrough innovation is a key force driving technological progress and industrial transformation. Timely and accurate identification of breakthrough innovation themes is of great value for science and technology policy formulation, corporate strategic planning, and the allocation of research resources. Currently, methods for identifying breakthrough innovation mainly focus on novelty, knowledge restructuring, knowledge structure, weak ties, and mutation analysis. For example, indicators such as patent citation novelty, interdisciplinary nature, and changes in knowledge network structure are used for prediction or identification. However, existing technologies still have the following shortcomings: First, the reliance on a single data source severs the intrinsic connection between scientific discovery and technological development. Existing methods largely depend on a single data source, such as scientific papers or technological patents, neglecting the continuity of the flow and integration of scientific knowledge into technological applications during the process of technological innovation. This makes it difficult to grasp the interactive relationship between scientific and technological breakthroughs as a whole.

[0003] Second, there is a lack of basic data screening for breakthrough innovation potential. Existing methods usually perform topic modeling or feature analysis directly on the entire dataset without pre-screening the input data (papers, patents) based on dimensions such as interdisciplinary nature, technological integration, and patent value. This results in the analysis results being greatly affected by noise, reducing the accuracy and reliability of the identification results.

[0004] Third, the ability to extract topic semantics is insufficient, resulting in poor interpretability of the results. Traditional topic models (such as LDA, which stands for Latent Dirichlet Allocation) have weak semantic connections during text vectorization and clustering, making it difficult to generate topic representations with domain specificity and semantic generalization. This results in ambiguous and multiple interpretations of the identified topics, making it difficult to directly serve science and technology management and decision-making.

[0005] To address the aforementioned issues, there is an urgent need in this field for a method that can integrate dual-source data from scientific papers and technological patents, screen potential data and identify topics with semantic enhancement, and classify groundbreaking innovative topics based on scientific-technological associations. Summary of the Invention

[0006] This invention provides a breakthrough innovation theme identification method based on science-technology correlation, which solves the shortcomings of existing technologies such as single data source, neglect of science-technology correlation, and insufficient theme interpretability, and realizes accurate identification and classification of breakthrough innovation themes linked by science and technology.

[0007] This invention provides a method for identifying breakthrough innovation topics based on the correlation between science and technology. It performs highly interdisciplinary screening and dataset construction on scientific paper data to obtain a scientific potential dataset; and it judges the innovation potential and constructs a dataset on patent technology data to obtain a technology potential dataset. By using various text-to-text pre-trained models, semantically enhanced topic recognition is performed based on the scientific potential dataset and the technological potential dataset to obtain a set of scientific topics and a set of technological topics. For the set of scientific topics, a scientific breakthrough innovation measurement model is established, and scientific breakthrough innovation topics are obtained by evaluating and identifying them based on multiple preset scientific dimensions; for the set of technological topics, a technological breakthrough innovation measurement model is established, and technological breakthrough innovation topics are obtained by evaluating and identifying them based on multiple preset technological dimensions. Based on Hellinger distance, the semantic relevance between the scientific breakthrough innovation topics and the technological breakthrough innovation topics is calculated; and based on Pasteur's science quadrant theory, the identified scientific breakthrough innovation topics and technological breakthrough innovation topics are classified into knowledge innovation type, strategic frontier type, and industrial transformation type according to the level of semantic relevance.

[0008] According to the present invention, a method for identifying breakthrough innovation topics based on science-technology correlation is provided, wherein the scientific paper data undergoes highly interdisciplinary screening and dataset construction to obtain a scientific potential dataset, including: For scientific paper data, based on citation networks and subject co-occurrence networks, the TDc subject crossover metric (a subject crossover metric based on a composite of true diversity and betweenness centrality) is used for screening to obtain highly interdisciplinary papers; a dataset is constructed based on the highly interdisciplinary papers to obtain a scientific potential dataset.

[0009] According to the present invention, a method for identifying breakthrough innovation topics based on science-technology correlation is provided, wherein the method involves judging the innovation potential and constructing a dataset for patent technology data to obtain a technology potential dataset, including: Based on the patent technology data, and considering the patent's own characteristics and citation status, the TDc technology fusion index (a technology fusion index based on a composite of true diversity and intermediary centrality) and the patent value index are used to screen high-potential patents. A dataset is then constructed based on these high-potential patents to obtain a technology potential dataset.

[0010] According to the present invention, a breakthrough innovation topic identification method based on science-technology association is provided. This method involves pre-training multiple text-to-text models and performing semantically enhanced topic identification based on the scientific potential dataset and the technological potential dataset to obtain a set of scientific topics and a set of technological topics, including: The obtained scientific paper data and patent technology data are processed by extracting the title and abstract text, and then performing stop word removal and word segmentation preprocessing to obtain processed data; a corpus is constructed based on the processed data to obtain a fused corpus. For the fused corpus, various pre-trained text steering models under the text steering framework are built; based on the fused corpus, semantic similarity calibration and context consistency evaluation are performed using the pre-trained text steering models to generate a high-dimensional document embedding space. The UMAP manifold learning algorithm (Uniform Manifold Approximation and Projection Learning Algorithm) is used to perform nonlinear dimensionality reduction based on the high-dimensional document embedding space to obtain a low-dimensional embedding vector. Based on the low-dimensional embedding vector, adaptive topic clustering is performed using the density-based HDBSCAN unsupervised clustering algorithm (a hierarchical density-based noise-applied spatial clustering unsupervised clustering algorithm) to obtain topic clusters; Based on the aforementioned topic clusters, the C-TF-IDF term weighting method (a category-based term frequency-inverse document frequency term weighting method) and KeyBERT key phrase extraction technology (a key phrase extraction technology based on bidirectional encoder representation, i.e., a key phrase extraction technology based on bidirectional encoder representation from Transformer) are combined to obtain topic feature vectors and corresponding topic labels. Based on the topic feature vectors and corresponding topic tags, a multi-dimensional evaluation system is constructed, including consistency index, diversity index, clustering quality index, and cross-domain adaptability index, to comprehensively evaluate the modeling effect, obtain a weighted comprehensive score, select the optimal topic model, and obtain the final confirmed topic clusters, topic feature vectors, and corresponding topic tags, and organize them to obtain a set of scientific topics and a set of technical topics.

[0011] According to the present invention, a method for identifying breakthrough innovation topics based on science-technology correlation is provided, wherein the construction process of the scientific breakthrough innovation measurement model includes: For the aforementioned set of scientific topics, a foundational network for a scientific breakthrough innovation measurement model is constructed based on citation networks and co-occurrence networks; Feature mapping and indicator design for a scientific breakthrough innovation measurement model based on data characteristic attributes and multiple preset scientific indicators; Multiple data feature attributes and multiple preset scientific indicators are positiveized and normalized, and the weight of each preset scientific indicator is calculated using the entropy weight method. The mutation type is determined based on the number of multiple data feature attributes and multiple preset scientific indicators. The mutation series method is used to aggregate each preset scientific indicator step by step to obtain a comprehensive evaluation value for each scientific breakthrough theme. The scientific breakthrough themes are ranked according to their comprehensive evaluation scores and then screened based on a preset scientific breakthrough threshold to obtain scientific breakthrough innovation themes.

[0012] According to the present invention, a method for identifying breakthrough innovation topics based on science-technology correlation is provided. The construction process of the technological breakthrough innovation measurement model includes: For the aforementioned set of technical topics, a foundational network for a technology breakthrough innovation measurement model is constructed based on a topic co-occurrence network. Feature mapping and indicator design of a technology breakthrough innovation measurement model based on patent data attribute characteristics and multiple preset technical indicators; Multiple patent data attribute features and multiple preset technical indicators are positiveized and normalized, and the weight of each preset technical indicator is calculated using the entropy weight method. The mutation type is determined based on the characteristics of multiple patent data attributes and the number of multiple preset technical indicators. The mutation series method is used to aggregate each preset technical indicator step by step to obtain the comprehensive evaluation value of each technological breakthrough theme. The themes of technological breakthroughs are ranked according to their comprehensive evaluation values ​​and then filtered based on preset technological breakthrough thresholds to obtain the themes of technological breakthrough innovation.

[0013] According to the present invention, a method for identifying breakthrough innovation topics based on science-technology association is provided, wherein calculating the semantic relevance between the scientific breakthrough innovation topic and the technological breakthrough innovation topic based on Hellinger distance includes: A unified vector space is established for the co-occurrence networks of scientific and technological breakthrough innovation topics and the co-occurrence networks of technological breakthrough innovation topics. The Hellinger distance is used to calculate the correlation between topic pairs to obtain the semantic correlation between the scientific and technological breakthrough innovation topics. The Hellinger distance formula is as follows: in, p i and q i Representing the distributions respectively P and Q The probability on the i-th sample. The value ranges from 0 to 1, where 0 indicates that the two distributions are exactly the same, and 1 indicates that they are completely different.

[0014] According to the present invention, a method for identifying breakthrough innovation themes based on science-technology correlation is provided. Based on Pasteur's science quadrant theory, the identified scientific and technological breakthrough innovation themes are classified into knowledge innovation type, strategic frontier type, and industrial transformation type according to the level of semantic relevance.

[0015] According to the present invention, a method for identifying breakthrough innovation themes based on science-technology correlation is provided. Based on Pasteur's science quadrant theory, the identified breakthrough innovation themes are categorized into knowledge innovation type, strategic frontier type, and industrial transformation type according to the semantic relevance between the identified scientific and technological breakthrough innovation themes. Based on the Pasteur science quadrant theory, and considering the semantic relevance between the scientific and technological breakthrough innovation themes, a high threshold and a low threshold are set. Scientific and technological breakthrough innovation themes with semantic relevance values ​​higher than the high threshold are classified as strategic frontier themes. Scientific breakthrough innovation topics whose semantic relevance values ​​are below the high threshold but above the low threshold are classified as knowledge innovation topics. Technological breakthrough innovation topics with semantic relevance values ​​below the high threshold but above the low threshold are classified as industrial transformation types.

[0016] This invention also provides a breakthrough innovation topic identification device based on science-technology association, comprising: The data filtering module is used to perform highly interdisciplinary filtering and dataset construction on scientific paper data to obtain a scientific potential dataset; and to judge the innovation potential and construct a dataset on patent technology data to obtain a technology potential dataset. The topic recognition module is used to perform semantically enhanced topic recognition based on the scientific potential dataset and the technological potential dataset by pre-training models with multiple text shift metrics, thereby obtaining a set of scientific topics and a set of technological topics. The measurement and evaluation module is used to evaluate and identify scientific breakthrough innovation themes based on multiple preset scientific dimensions by establishing a scientific breakthrough innovation measurement model for the set of scientific themes; and to evaluate and identify technological breakthrough innovation themes based on multiple preset technological dimensions by establishing a technological breakthrough innovation measurement model for the set of technological themes. The association classification module calculates the semantic relevance between the scientific breakthrough innovation topics and the technological breakthrough innovation topics based on Herringer distance; and based on Pasteur's science quadrant theory, it classifies the identified scientific breakthrough innovation topics and technological breakthrough innovation topics into knowledge innovation type, strategic frontier type, and industrial transformation type according to the level of semantic relevance.

[0017] The present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the breakthrough innovation topic identification method based on science-technology association as described above.

[0018] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the breakthrough innovation topic identification method based on science-technology association as described above.

[0019] The present invention also provides a computer program product, including a computer program that, when executed by a processor, implements the breakthrough innovation topic identification method based on science-technology association as described above.

[0020] This invention provides a method for identifying breakthrough innovation topics based on science-technology correlation. The method involves: first, performing highly interdisciplinary screening and dataset construction on scientific paper data to obtain a scientific potential dataset; second, judging the innovation potential and constructing a dataset on patent technology data to obtain a technological potential dataset; third, using multiple pre-trained text-to-text models, performing semantically enhanced topic identification based on the scientific and technological potential datasets to obtain a set of scientific topics and a set of technological topics; fourth, establishing a scientific breakthrough innovation measurement model for the scientific topic set, evaluating and identifying it based on multiple preset scientific dimensions to obtain scientific breakthrough innovation topics; and fifth, establishing a technological breakthrough innovation measurement model for the technological topic set, evaluating and identifying it based on multiple preset technological dimensions to obtain technological breakthrough innovation topics; sixth, calculating the semantic relevance between the scientific and technological breakthrough innovation topics based on Heringer distance; and finally, based on Pasteur's science quadrant theory, classifying the identified scientific and technological breakthrough innovation topics into knowledge innovation type, strategic frontier type, and industrial transformation type according to the level of semantic relevance. By integrating scientific papers and patent data, and through potential screening and semantically enhanced topic identification, a dual-source science-technology measurement model is constructed. Hellinger distance is used to calculate semantic associations, and the models are categorized into knowledge innovation, strategic frontier, and industrial transformation types based on the Pasteur quadrant. This effectively addresses the problems of single data sources, high noise interference, and poor interpretability, improving identification accuracy and decision-making value. Attached Figure Description

[0021] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0022] Figure 1 This is a flowchart illustrating a groundbreaking innovative topic identification method based on science-technology association provided by the present invention.

[0023] Figure 2 It corresponds to Figure 1 A framework diagram of the process flow.

[0024] Figure 3 This is a schematic diagram of the structure of a groundbreaking innovative topic identification device based on science-technology association provided by the present invention.

[0025] Figure 4 This is a schematic diagram of the structure of the electronic device provided by the present invention. Detailed Implementation

[0026] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0027] Terminology Explanation: Hellinger distance: Hellinger distance.

[0028] TDc Interdisciplinary Measurement Index: An interdisciplinary measurement index based on a combination of true diversity and intermediation centrality.

[0029] TDc Technology Integration Index: A technology integration index based on a composite of true diversity and intermediation centrality.

[0030] UMAP manifold learning algorithm: Uniform manifold approximation and projection learning algorithm.

[0031] HDBSCAN Unsupervised Clustering Algorithm: A hierarchical density-based spatial clustering algorithm for noisy applications.

[0032] C-TF-IDF term weighting method: a category-based term frequency-inverse document frequency term weighting method.

[0033] KeyBERT key phrase extraction technology: a key phrase extraction technology based on bidirectional encoder representation, that is, a key phrase extraction technology based on bidirectional encoder representation from Transformer.

[0034] The following is combined Figures 1 to 4 This invention is described.

[0035] Figure 1 This is one of the flowcharts illustrating the groundbreaking innovation topic identification method based on science-technology correlation provided by this invention. Figure 2 It corresponds to Figure 1 A framework diagram of a process flow chart. Figure 2 and Figure 1 They can be referenced and correspond to each other.

[0036] like Figure 1 As shown, the method includes the following: Step 101: Perform interdisciplinary screening and dataset construction on scientific paper data to obtain a scientific potential dataset; perform innovation potential assessment and dataset construction on patent technology data to obtain a technology potential dataset.

[0037] Optionally, for scientific paper data, based on citation networks and subject co-occurrence networks, the TDc interdisciplinary measure index is used for screening to obtain highly interdisciplinary papers; a dataset is constructed based on the highly interdisciplinary papers to obtain a scientific potential dataset.

[0038] The TDc interdisciplinary measurement index combines True-Diversity and Middleness Centrality indicators, and uses a threshold for filtering. The formula is as follows: Where, p i ,p j Let represent the proportions of subject i and subject j respectively, and let dij represent the similarity between subject i and subject j. Let α+β=1, and let Centrality∈(0,1) represent the intermediary centrality of a subject in the entire subject co-occurrence network, which can be calculated using Python.

[0039] The calculation of TDc interdisciplinary indicators adopts the following steps: Step 1: Determine the subject classification of the paper. Use the JCR classification system in the Web of Science platform as the basis for subject identification of the paper. It divides the subject categories by journal.

[0040] Step 2: Establish a paper-journal-subject mapping table, obtain journal information for papers and references, and a mapping table of JCR journal abbreviations and full names, and establish the mapping table.

[0041] Step 3: Construct a subject co-occurrence network and obtain the parameters required for the TDc index, such as centrality and similarity.

[0042] Step 4: Calculate the indicators.

[0043] Optionally, for patent technology data, based on the patent's own characteristics and citation status, the TDc technology integration index and patent value index are used for screening to obtain high-potential patents; a dataset is constructed based on the high-potential patents to obtain a technology potential dataset.

[0044] The calculation method of the TDc technology integration index is the same as that of the TDc discipline cross-disciplinary index, which represents the cross-integration characteristics of patent technologies; the patent value index reflects that patent data with high value has greater potential for breakthrough innovation; after normalizing the TDc technology integration index and the patent value index, the system model is carried out by objectively assigning weights using Critic, and high-potential patents are screened to form a technology potential dataset.

[0045] Patent Value: High-potential patents are characterized by high value. They are selected using high-value patent indicators from the Innography patent data platform as prior input data for measuring the technological convergence of patents.

[0046] Technological convergence: The intersection of technologies is an important source of groundbreaking innovation, including the following steps: Step 1: Determine the patent technology category: Selecting the patent IPC classification can represent the technical points of the patent, and the hierarchical structure reflects the semantic relationship between concepts, which is authoritative and universal.

[0047] Step 2: Construct a patent co-occurrence network: Based on the characteristics of IPC classification numbers, we will select the IPC-4-digit classification standard as the basic semantic unit, construct a classification number co-occurrence network and an IPC subclass similarity matrix, and calculate the similarity between classification numbers.

[0048] Step 3, TDc index calculation: Using the IPC classification number co-occurrence network, the calculation parameters of the TDc index are obtained, and the selection of patents with high technological integration is implemented through Python programming.

[0049] Step 102: Using multiple pre-trained text-to-text models, semantically enhanced topic recognition is performed based on the scientific potential dataset and the technological potential dataset to obtain a set of scientific topics and a set of technological topics.

[0050] Optionally, step 102 includes steps A1 to A6: Step A1: Extract the title and abstract text from the acquired scientific paper data and patent technology data, and perform stop word removal and word segmentation preprocessing to obtain processed data; construct a corpus based on the processed data to obtain a fused corpus.

[0051] The technical solution of this application exemplarily utilizes the Web of Science platform to obtain papers and uses PatSnap to obtain patents.

[0052] Step A2: For the fused corpus, build various pre-trained text steering models under the text steering framework; use the pre-trained text steering models to perform semantic similarity calibration and context consistency evaluation based on the fused corpus to generate a high-dimensional document embedding space.

[0053] The text-to-quantity pre-trained model in this application exemplarily employs the all-MiniLM-L6-v2 pre-trained model under the BERTopic framework, a variant of the isomorphic Sentence-BERT model, the SCIbert domain enhancement model, and the SPECTER relation modeling model.

[0054] Step A3: Using the UMAP manifold learning algorithm, perform nonlinear dimensionality reduction based on the high-dimensional document embedding space to obtain a low-dimensional embedding vector.

[0055] Step A4: Based on the low-dimensional embedding vector, adaptive topic clustering is performed using the density-based HDBSCAN unsupervised clustering algorithm to obtain topic clusters.

[0056] This application exemplarily demonstrates how to effectively handle density unevenness and noise interference in text data by dynamically adjusting the minimum cluster size (MinPts) and the distance threshold (eps).

[0057] Step A5: Based on the topic clusters, the C-TF-IDF term weighting method and KeyBERT key phrase extraction technology are combined to obtain topic feature vectors and corresponding topic tags.

[0058] This application constructs topic feature vectors and corresponding topic tags that combine domain specificity and semantic generality. This retains the representational ability of high-frequency terms while using cosine similarity to filter a subset of core terms, thus avoiding the impact of "long-tail term noise" on topic focus.

[0059] Step A6: Based on the topic feature vector and the corresponding topic label, the modeling effect is comprehensively evaluated by constructing a multi-dimensional evaluation system of consistency index, diversity index, clustering quality index and cross-domain adaptability index. A weighted comprehensive score is obtained, the optimal topic model is selected, and the final confirmed topic clusters, topic feature vectors and corresponding topic labels are obtained. The scientific topic set and the technical topic set are then organized.

[0060] This application measures the logical coherence of topic terms using a consistency index and quantifies the semantic distinguishability between topics within the topic quality dimension. In terms of model generalization ability, it evaluates cluster compactness and separation using a clustering quality index and introduces a cross-domain adaptability index to test the model's semantic preservation ability for heterogeneous texts. Based on the weighted comprehensive scores of the above indicators, the optimal topic model is selected according to requirements, resulting in the final confirmed topic clusters, topic feature vectors, and corresponding topic labels. These are then organized into a set of scientific topics and a set of technical topics, providing a foundation for subsequent coupled network construction.

[0061] Step 103: For the set of scientific topics, a scientific breakthrough innovation measurement model is established, and evaluation and identification are performed based on multiple preset scientific dimensions to obtain scientific breakthrough innovation topics; for the set of technological topics, a technological breakthrough innovation measurement model is established, and evaluation and identification are performed based on multiple preset technological dimensions to obtain technological breakthrough innovation topics.

[0062] Optionally, the construction process of the scientific breakthrough innovation measurement model includes steps B1 to B5: Step B1: For the set of scientific topics, construct the basic network of the scientific breakthrough innovation measurement model based on citation network and co-occurrence network.

[0063] Step B2: Based on data feature attributes and multiple preset scientific indicators, perform feature mapping and indicator design for a scientific breakthrough innovation measurement model.

[0064] This application aims to utilize indicators such as publication time, Kleinberg detection algorithm, topic strength, citation strength, and betweenness centrality to perform feature mapping and indicator design for novelty, disruptiveness, and scientific impact. The indicator system for measuring scientific breakthrough innovation is shown in Table 1.

[0065] Table 1

[0066] The specific indicators are as follows: Novelty: refers to the freshness of a paper's topic, defined as the average age of papers and references on a particular topic.

[0067] Sudden change: Kleinberg proposed a burst detection algorithm to identify burst words within a certain time period based on word frequency change rate.

[0068] Scientific impact: Measured comprehensively from three aspects: topic intensity, citation intensity, and betweenness centrality.

[0069] The intensity of a scientific topic is represented by the percentage of research papers published on that topic, which directly reflects the research popularity of that topic.

[0070] Citation intensity is represented by the average citation frequency of all papers under a topic, reflecting the influence of a research topic among peers.

[0071] Betweenness centrality is the proportion of the number of shortest paths through a particular topic node in a topic network to the total number of shortest paths. It reflects the influence of a node in the network; the larger the value, the higher the influence of the topic.

[0072] Step B3: Perform positive transformation and normalization on multiple data feature attributes and multiple preset scientific indicators, and use the entropy weight method to calculate the weight of each preset scientific indicator.

[0073] Step B4: Determine the mutation type based on the number of multiple data feature attributes and multiple preset scientific indicators, and use the mutation series method to aggregate each preset scientific indicator step by step to obtain the comprehensive evaluation value of each scientific breakthrough theme.

[0074] Step B5: Sort the scientific breakthrough topics according to the comprehensive evaluation value, and filter them based on the preset scientific breakthrough threshold to obtain scientific breakthrough innovation topics.

[0075] Optionally, the construction process of the technological breakthrough innovation measurement model includes steps C1 to C5: Step C1: For the set of technical topics, construct the basic network of the technical breakthrough innovation measurement model based on the topic co-occurrence network.

[0076] Step C2: Based on the attribute characteristics of patent data and multiple preset technical indicators, perform feature mapping and indicator design for the technological breakthrough innovation measurement model.

[0077] The patent data attribute features of this application include, for example, the patent title and abstract, and the preset technical indicators include, for example, mutability, novelty, impact, and economic value. The indicator system for measuring technological breakthrough innovation is shown in Table 2.

[0078] Table 2

[0079] The specific indicators are as follows: Novelty: Novelty based on word age reflects the newness of the patent content. It is defined as the reciprocal of the average word age of the technical subject. The word age is expressed by the patent application date.

[0080] Sudden emergence: a key characteristic of breakthrough innovation, is the rapid emergence of patent subject matter, i.e., a leap. These leaps and breaks are the key points of subject matter mutation. This paper proposes a burst detection algorithm based on Kleinberg's theory, identifying burst words within a specific time period based on word frequency change rates.

[0081] Middleness centrality: The technological influence based on network structure will be represented by patent family size, number of patent reference patents, and middleness centrality, with patent family size being a representative factor.

[0082] Economic value: This can be used to measure the commercial economic value of a patent in the market. It can be judged from the scope of protection and the scope of business. The scope of protection is the number of patent claims, which means that the patent covers a wider range of technical details and application fields. The scope of business is the number of countries in which the patent is deployed. The more claims there are, the more commercial value and commercialization potential the patent has.

[0083] Step C3: Perform positive transformation and normalization on multiple patent data attribute features and multiple preset technical indicators, and use the entropy weight method to calculate the weight of each preset technical indicator.

[0084] Step C4: Determine the mutation type based on the characteristics of multiple patent data attributes and the number of multiple preset technical indicators, and use the mutation series method to aggregate each preset technical indicator step by step to obtain the comprehensive evaluation value of each technological breakthrough theme.

[0085] Step C5: Sort the technological breakthrough topics according to the comprehensive evaluation value, and filter them based on the preset technological breakthrough threshold to obtain the technological breakthrough innovation topics.

[0086] Step 104: Based on Hellinger distance, calculate the semantic relevance between the scientific breakthrough innovation theme and the technological breakthrough innovation theme; and based on Pasteur's science quadrant theory, classify the identified scientific breakthrough innovation theme and technological breakthrough innovation theme into knowledge innovation type, strategic frontier type and industrial transformation type according to the level of semantic relevance.

[0087] Optionally, calculating the semantic relevance between the scientific breakthrough innovation topic and the technological breakthrough innovation topic based on Hellinger distance includes: A unified vector space is established for the co-occurrence networks of scientific and technological breakthrough innovation topics and the co-occurrence networks of technological breakthrough innovation topics. The Hellinger distance is used to calculate the correlation between the topics to obtain the semantic correlation between the scientific and technological breakthrough innovation topics. The Hellinger distance formula is as follows: in, p i and q i Representing the distributions respectively P and Q The probability on the i-th sample. The value ranges from 0 to 1, where 0 indicates that the two distributions are exactly the same, and 1 indicates that they are completely different.

[0088] Optionally, based on Pasteur's scientific quadrant theory, the identified scientific and technological breakthrough innovation themes are categorized according to their semantic relevance into knowledge innovation, strategic frontier, and industrial transformation types, forming a breakthrough innovation quadrant coordinate map; including steps D1 to D3: Step D1: Based on the Pasteur science quadrant theory, and considering the semantic relevance between the scientific breakthrough innovation topics and the technological breakthrough innovation topics, a high threshold and a low threshold are set. Scientific breakthrough innovation topics and technological breakthrough innovation topics with semantic relevance values ​​higher than the high threshold are classified as strategic frontier topics. Step D2: Scientific breakthrough innovation topics with semantic relevance values ​​below the high threshold but above the low threshold are classified as knowledge innovation topics; Step D3: Classify technological breakthrough innovation topics whose semantic relevance values ​​are lower than the high threshold but higher than the low threshold as industrial transformation types.

[0089] This invention provides a method for identifying breakthrough innovation topics based on science-technology correlation. The method involves: first, performing highly interdisciplinary screening and dataset construction on scientific paper data to obtain a scientific potential dataset; second, judging the innovation potential and constructing a dataset on patent technology data to obtain a technological potential dataset; third, using multiple pre-trained text-to-text models, performing semantically enhanced topic identification based on the scientific and technological potential datasets to obtain a set of scientific topics and a set of technological topics; fourth, establishing a scientific breakthrough innovation measurement model for the scientific topic set, evaluating and identifying it based on multiple preset scientific dimensions to obtain scientific breakthrough innovation topics; and fifth, establishing a technological breakthrough innovation measurement model for the technological topic set, evaluating and identifying it based on multiple preset technological dimensions to obtain technological breakthrough innovation topics; sixth, calculating the semantic relevance between the scientific and technological breakthrough innovation topics based on Heringer distance; and finally, based on Pasteur's science quadrant theory, classifying the identified scientific and technological breakthrough innovation topics into knowledge innovation type, strategic frontier type, and industrial transformation type according to the level of semantic relevance. By integrating scientific papers and patent data, and through potential screening and semantically enhanced topic identification, a dual-source science-technology measurement model is constructed. Hellinger distance is used to calculate semantic associations, and the models are categorized into knowledge innovation, strategic frontier, and industrial transformation types based on the Pasteur quadrant. This effectively addresses the problems of single data sources, high noise interference, and poor interpretability, improving identification accuracy and decision-making value.

[0090] The following describes a breakthrough innovation theme identification device based on science-technology association provided by the present invention. The breakthrough innovation theme identification device based on science-technology association described below and the breakthrough innovation theme identification method based on science-technology association described above can be referred to and correspond to each other.

[0091] Figure 3 This is a flowchart illustrating a groundbreaking innovative topic identification device based on science-technology correlation provided by the present invention, such as... Figure 3 As shown, the device includes the following: The data filtering module is used to perform highly interdisciplinary filtering and dataset construction on scientific paper data to obtain a scientific potential dataset; and to judge the innovation potential and construct a dataset on patent technology data to obtain a technology potential dataset. The topic recognition module is used to perform semantically enhanced topic recognition based on the scientific potential dataset and the technological potential dataset by pre-training models with multiple text shift metrics, thereby obtaining a set of scientific topics and a set of technological topics. The measurement and evaluation module is used to evaluate and identify scientific breakthrough innovation themes based on multiple preset scientific dimensions by establishing a scientific breakthrough innovation measurement model for the set of scientific themes; and to evaluate and identify technological breakthrough innovation themes based on multiple preset technological dimensions by establishing a technological breakthrough innovation measurement model for the set of technological themes. The association classification module calculates the semantic relevance between the scientific breakthrough innovation topics and the technological breakthrough innovation topics based on Herringer distance; and based on Pasteur's science quadrant theory, it classifies the identified scientific breakthrough innovation topics and technological breakthrough innovation topics into knowledge innovation type, strategic frontier type, and industrial transformation type according to the level of semantic relevance.

[0092] This invention provides a breakthrough innovation topic identification device based on science-technology correlation. It obtains a scientific potential dataset by performing highly interdisciplinary screening and dataset construction on scientific paper data; and a technological potential dataset by judging the innovation potential of patent technology data. Using multiple pre-trained text-to-text models, it performs semantically enhanced topic identification based on the scientific and technological potential datasets, resulting in a set of scientific topics and a set of technological topics. For the scientific topic set, it establishes a scientific breakthrough innovation measurement model, evaluating and identifying it based on multiple preset scientific dimensions to obtain scientific breakthrough innovation topics. For the technological topic set, it establishes a technological breakthrough innovation measurement model, evaluating and identifying it based on multiple preset technological dimensions to obtain technological breakthrough innovation topics. Based on Heringer distance, it calculates the semantic correlation between the scientific and technological breakthrough innovation topics. Based on Pasteur's science quadrant theory, it classifies the identified scientific and technological breakthrough innovation topics into knowledge innovation type, strategic frontier type, and industrial transformation type according to the level of semantic correlation. By integrating scientific papers and patent data, and through potential screening and semantically enhanced topic identification, a dual-source science-technology measurement model is constructed. Hellinger distance is used to calculate semantic associations, and the models are categorized into knowledge innovation, strategic frontier, and industrial transformation types based on the Pasteur quadrant. This effectively addresses the problems of single data sources, high noise interference, and poor interpretability, improving identification accuracy and decision-making value.

[0093] Figure 4 An example is a schematic diagram of the physical structure of an electronic device, such as... Figure 4As shown, the electronic device may include a processor 810, a communications interface 820, a memory 830, and a communication bus 840. The processor 810, communications interface 820, and memory 830 communicate with each other via the communication bus 840. The processor 810 can call logical instructions from the memory 830 to execute a breakthrough innovation topic identification method based on science-technology correlation.

[0094] Furthermore, the logical instructions in the aforementioned memory 830 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or a 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 the present invention. 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.

[0095] On the other hand, the present invention also provides a computer program product, which includes a computer program that can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer is able to execute a breakthrough innovation topic identification method based on science-technology association provided by the above methods.

[0096] In another aspect, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, is implemented to perform a breakthrough innovation topic identification method based on science-technology association provided by the methods described above.

[0097] The device embodiments described above are merely illustrative. 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 modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0098] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0099] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for identifying groundbreaking innovation topics based on science-technology correlation, characterized in that, A scientific potential dataset is obtained by conducting interdisciplinary screening and dataset construction on scientific paper data; and a technological potential dataset is obtained by judging the innovation potential and constructing a dataset on patent technology data. By using various text-to-text pre-trained models, semantically enhanced topic recognition is performed based on the scientific potential dataset and the technological potential dataset to obtain a set of scientific topics and a set of technological topics. For the set of scientific topics, a scientific breakthrough innovation measurement model is established, and scientific breakthrough innovation topics are obtained by evaluating and identifying them based on multiple preset scientific dimensions; for the set of technological topics, a technological breakthrough innovation measurement model is established, and technological breakthrough innovation topics are obtained by evaluating and identifying them based on multiple preset technological dimensions. Based on Heringer distance, the semantic relevance between the scientific breakthrough innovation topic and the technological breakthrough innovation topic is calculated; Based on Pasteur's science quadrant theory, and according to the degree of semantic relevance, the identified scientific and technological breakthrough innovation themes are classified into knowledge innovation, strategic frontier, and industrial transformation themes.

2. The method for identifying breakthrough innovation topics based on science-technology correlation according to claim 1, characterized in that, The aforementioned interdisciplinary screening and dataset construction of scientific paper data yields a scientific potential dataset, including: For scientific paper data, based on citation networks and subject co-occurrence networks, a subject cross-disciplinary measure index based on true diversity-mediation centrality is used for screening to obtain highly interdisciplinary papers; a dataset is constructed based on the highly interdisciplinary papers to obtain a scientific potential dataset.

3. The method for identifying breakthrough innovation topics based on science-technology correlation according to claim 1, characterized in that, The process of assessing innovation potential and constructing a dataset based on patent technology data yields a technology potential dataset, including: Based on the patent technology data, and considering the patent's own characteristics and citation status, a combination of technology fusion index and patent value index based on true diversity-mediation centrality is used to screen high-potential patents. A dataset is then constructed based on these high-potential patents to obtain a technology potential dataset.

4. The method for identifying breakthrough innovation topics based on science-technology correlation according to claim 1, characterized in that, The method involves pre-training models using multiple text-to-text conversion metrics, performing semantically enhanced topic recognition based on the scientific potential dataset and the technological potential dataset, to obtain a set of scientific topics and a set of technological topics, including: The obtained scientific paper data and patent technology data are processed by extracting the title and abstract text, and then performing stop word removal and word segmentation preprocessing to obtain processed data; a corpus is constructed based on the processed data to obtain a fused corpus. For the fused corpus, various pre-trained text steering models under the text steering framework are built; based on the fused corpus, semantic similarity calibration and context consistency evaluation are performed using the pre-trained text steering models to generate a high-dimensional document embedding space. By using a uniform manifold approximation and projection learning algorithm, a nonlinear dimensionality reduction is performed based on the high-dimensional document embedding space to obtain a low-dimensional embedding vector. Based on the low-dimensional embedding vector, an adaptive topic clustering algorithm is performed using density-based and hierarchical density-based noise-applied spatial clustering to obtain topic clusters. Based on the aforementioned topic clusters, a category-based term frequency-inverse document frequency weighting method and a key phrase extraction technique based on bidirectional encoder representation are combined to obtain topic feature vectors and corresponding topic tags; Based on the topic feature vectors and corresponding topic tags, a multi-dimensional evaluation system is constructed, including consistency index, diversity index, clustering quality index, and cross-domain adaptability index, to comprehensively evaluate the modeling effect, obtain a weighted comprehensive score, select the optimal topic model, obtain the final confirmed topic clusters, topic feature vectors and corresponding topic tags, and organize them to obtain a set of scientific topics and a set of technical topics.

5. The breakthrough innovation topic identification method based on science-technology correlation according to claim 1, characterized in that, The construction process of the scientific breakthrough innovation measurement model includes: For the aforementioned set of scientific topics, a foundational network for a scientific breakthrough innovation measurement model is constructed based on citation networks and co-occurrence networks; Feature mapping and indicator design for a scientific breakthrough innovation measurement model based on data characteristic attributes and multiple preset scientific indicators; Multiple data feature attributes and multiple preset scientific indicators are positiveized and normalized, and the weight of each preset scientific indicator is calculated using the entropy weight method. The mutation type is determined based on the number of multiple data feature attributes and multiple preset scientific indicators. The mutation series method is used to aggregate each preset scientific indicator step by step to obtain a comprehensive evaluation value for each scientific breakthrough theme. The scientific breakthrough themes are ranked according to their comprehensive evaluation scores and then screened based on a preset scientific breakthrough threshold to obtain scientific breakthrough innovation themes.

6. The method for identifying breakthrough innovation topics based on science-technology correlation according to claim 1, characterized in that, The construction process of the technological breakthrough innovation measurement model includes: For the aforementioned set of technical topics, a foundational network for a technology breakthrough innovation measurement model is constructed based on a topic co-occurrence network. Feature mapping and indicator design of a technology breakthrough innovation measurement model based on patent data attribute characteristics and multiple preset technical indicators; Multiple patent data attribute features and multiple preset technical indicators are positiveized and normalized, and the weight of each preset technical indicator is calculated using the entropy weight method. The mutation type is determined based on the characteristics of multiple patent data attributes and the number of multiple preset technical indicators. The mutation series method is used to aggregate each preset technical indicator step by step to obtain the comprehensive evaluation value of each technological breakthrough theme. The themes of technological breakthroughs are ranked according to their comprehensive evaluation values ​​and then filtered based on preset technological breakthrough thresholds to obtain the themes of technological breakthrough innovation.

7. The method for identifying breakthrough innovation topics based on science-technology correlation according to claim 1, characterized in that, The calculation of the semantic relevance between the scientific breakthrough innovation topic and the technological breakthrough innovation topic based on Heringer distance includes: A unified vector space is established for the co-occurrence networks of scientific and technological breakthrough innovation topics and the co-occurrence networks of technological breakthrough innovation topics. Herringer distance is used to calculate the correlation between topics to obtain the semantic correlation between the scientific and technological breakthrough innovation topics. The Herringer distance formula is as follows: in, p i and q i Representing the distributions respectively P and Q The probability on the i-th sample. The value ranges from 0 to 1, where 0 indicates that the two distributions are exactly the same, and 1 indicates that they are completely different.

8. The method for identifying breakthrough innovation topics based on science-technology correlation according to claim 1, characterized in that, Based on Pasteur's science quadrant theory, and according to the degree of semantic relevance between the identified scientific and technological breakthrough innovation themes, the identified scientific and technological breakthrough innovation themes are classified into knowledge innovation type, strategic frontier type, and industrial transformation type.

9. The method for identifying breakthrough innovation topics based on science-technology correlation according to claim 1, characterized in that, Based on Pasteur's science quadrant theory, and according to the semantic relevance between the identified scientific and technological breakthrough innovation themes, the themes are categorized into three types: knowledge innovation, strategic frontier, and industrial transformation. Based on the Pasteur science quadrant theory, and considering the semantic relevance between the scientific and technological breakthrough innovation themes, a high threshold and a low threshold are set. Scientific and technological breakthrough innovation themes with semantic relevance values ​​higher than the high threshold are classified as strategic frontier themes. Scientific breakthrough innovation topics whose semantic relevance values ​​are below the high threshold but above the low threshold are classified as knowledge innovation topics. Technological breakthrough innovation topics with semantic relevance values ​​below the high threshold but above the low threshold are classified as industrial transformation types.

10. A breakthrough innovation theme identification device based on science-technology correlation, characterized in that, include: The data filtering module is used to perform highly interdisciplinary filtering and dataset construction on scientific paper data to obtain a scientific potential dataset. Based on patent technology data, we assess innovation potential and construct a dataset to obtain a technology potential dataset. The topic recognition module is used to perform semantically enhanced topic recognition based on the scientific potential dataset and the technological potential dataset by pre-training models with multiple text shift metrics, thereby obtaining a set of scientific topics and a set of technological topics. The measurement and evaluation module is used to evaluate and identify scientific breakthrough innovation themes based on multiple preset scientific dimensions by establishing a scientific breakthrough innovation measurement model for the set of scientific themes; and to evaluate and identify technological breakthrough innovation themes based on multiple preset technological dimensions by establishing a technological breakthrough innovation measurement model for the set of technological themes. The association classification module calculates the semantic relevance between the scientific breakthrough innovation topics and the technological breakthrough innovation topics based on Herringer distance; and based on Pasteur's science quadrant theory, it classifies the identified scientific breakthrough innovation topics and technological breakthrough innovation topics into knowledge innovation type, strategic frontier type, and industrial transformation type according to the level of semantic relevance.

11. An electronic device comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the computer program, it implements a breakthrough innovation topic identification method based on science-technology association as described in any one of claims 1 to 9.

12. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements a breakthrough innovation topic identification method based on science-technology association as described in any one of claims 1 to 9.

13. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by the processor, it implements a breakthrough innovation topic identification method based on science-technology association as described in any one of claims 1 to 9.

Images