A patent technology solution and an application scenario adaptation method and system

Abstract

This application relates to the field of patent application information analysis technology, and discloses a method and system for matching patent technology solutions with application scenarios. This method extracts the technical solution theme tags and application scenario theme tags corresponding to each patent technology document from the same data source, generating a dual-label dataset. This achieves a simultaneous characterization of the technology supply-side structure and the scenario demand-side distribution, providing a two-dimensional data input for supply-demand matching analysis and overcoming the limitation of existing single-theme models that struggle to distinguish between the semantics of technology supply and scenario demand. Furthermore, based on a target matrix, it identifies combinations of technical solution theme tags and application scenario theme tags with different potential values ​​to determine the degree of fit between the technical solution and the application scenario. This efficiently identifies the degree of fit between the technical solution and the application scenario, thereby efficiently identifying potentially valuable innovation opportunities.

Description

Technical Field

[0001] This invention relates to the field of information analysis technology, and discloses a patented technical solution and a method and system for adapting to application scenarios. Background Technology

[0002] With the accelerating pace of technological innovation, patent documents, as an important carrier for systematically recording the technological development process, are widely used in fields such as technology intelligence analysis, technology trend identification, and innovation decision support. In existing technologies, statistical analysis of information such as the quantity of patent documents, their classification numbers, or citation relationships is typically used to determine the development trend or competitive landscape of a particular technological field.

[0003] In recent years, with the development of natural language processing technology, some methods have begun to attempt semantic analysis of patent documents. Existing research typically uses semantic modeling of the title, abstract, or claim information of patent documents to identify technical themes, hot topics, or trends in technological evolution. While these methods have some value in revealing the structure of technological development, they still have shortcomings from the perspective of innovation decision-making and practical application. For example, existing technologies neglect the semantics of application scenario requirements, resulting in a lack of clear mapping between technical semantics and application scenario semantics. In existing patent semantic analysis research, some works have begun to focus on the direction of technological application or application field labels, but the technical semantics and application scenario semantics are usually still loosely related or described in parallel. Technical themes and application scenarios are often extracted separately as information at different levels or dimensions, lacking a unified structured mapping mechanism to characterize the correspondence, coupling strength, and combination features between the two. If a clear mapping relationship between technical themes and application scenarios cannot be established, it will lead to an inefficient identification of the suitability of patent technology solutions for application scenarios, and consequently, an inefficient identification of potentially valuable innovation opportunities.

[0004] It is evident that existing methods suffer from low efficiency in identifying the compatibility between technical solutions and application scenarios. Summary of the Invention

[0005] This invention provides a method and system for adapting patented technical solutions to application scenarios, in order to solve the problem of low efficiency in identifying the degree of adaptability between patented technical solutions and application scenarios in the prior art.

[0006] Firstly, this application provides a method for adapting a patented technical solution to an application scenario, including:

[0007] Obtain patent technology documents that are under effective protection in the target industry; Perform dual-dimensional semantic recognition on the patent technology documents to obtain the technical solution topic tags and application scenario topic tags corresponding to the patent technology documents; A dual-label dataset is generated based on the patent technology documents, the technical solution topic tags, and the application scenario topic tags. The dual-label dataset is quantitatively calculated and spatially mapped to obtain a target matrix, which is used to characterize the potential value of each patented technology solution in different application scenarios. Based on the target matrix, the degree of compatibility between the target technical solution and the target application scenario is determined. The higher the degree of compatibility between the target technical solution and the target application scenario, the higher the potential value of the target technical solution.

[0008] Secondly, this application provides a system for adapting a patented technical solution to an application scenario, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the steps of the method described in the first aspect above.

[0009] Thirdly, this application provides a computer program product stored in a storage medium, which is executed by at least one processor to implement the steps of the method as described in the first aspect.

[0010] The present invention has the following beneficial effects: The method for matching patent technology solutions with application scenarios in this application extracts the technical solution theme tags and application scenario theme tags corresponding to each patent technology document from the same data source, generating a dual-label dataset. This achieves a simultaneous characterization of the technology supply-side structure and the scenario demand-side distribution, providing a two-dimensional data input for supply-demand matching analysis and overcoming the limitation of existing single-theme models that struggle to distinguish between the semantics of technology supply and scenario demand. Furthermore, a target matrix is ​​constructed to characterize the potential value of technical solutions in different application scenarios. Based on the target matrix, combinations of technical solution theme tags and application scenario theme tags with different potential values ​​are identified to determine the degree of fit between the target technical solution and the target application scenario. This establishes a clear mapping relationship between patent technology solutions and application scenarios, moving away from the traditional model that relies on qualitative experience or simple statistics to evaluate technology value. This method can efficiently identify the degree of fit between patent technology solutions and application scenarios, thereby efficiently identifying potentially valuable innovation opportunities.

[0011] In addition to the objectives, features and advantages described above, the present invention has other objectives, features and advantages.

[0012] The present invention will now be described in further detail with reference to the figures. Attached Figure Description

[0013] The accompanying drawings, which form part of this application, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an undue limitation of the invention. In the drawings: Figure 1 This is a flowchart illustrating a preferred embodiment of the invention and a method for adapting a patented technical solution to an application scenario. Figure 2 This is a schematic diagram of the target matrix according to a preferred embodiment of the present invention. Detailed Implementation

[0014] The technical solution of the present invention will be clearly and completely described below. 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.

[0015] Unless otherwise defined, the technical or scientific terms used in this invention shall have the ordinary meaning understood by one of ordinary skill in the art to which this invention pertains. The terms "first," "second," and similar terms used in this invention do not indicate any order, quantity, or importance, but are merely used to distinguish different components. Similarly, the terms "an" or "a," and similar terms, do not indicate a quantity limitation, but rather indicate the presence of at least one.

[0016] Please see Figure 1 This application provides a method for adapting a patented technical solution to an application scenario, including: Step S1: Obtain patent technology documents in the target industry that are under effective protection.

[0017] In this step, based on the target industry sector, patent documents in a valid protection status are selected from publicly available patent document databases. These databases can refer to free or paid websites for searching patent applications. "Valid protection status" specifically refers to the "valid" status mentioned on the publicly available website. The target industry refers to the industry to be analyzed, such as the big data industry, the chemical industry, and the engineering and construction industry; this is merely an example and not a limitation.

[0018] Step S2: Perform two-dimensional semantic recognition on the patent technology document to obtain the technical solution topic tags and application scenario topic tags corresponding to the patent technology document.

[0019] In this step, dual-dimensional semantic recognition specifically refers to the identification of the technical subject matter of the patent document and the identification of the application scenario subject matter.

[0020] Step S3: Generate a dual-label dataset based on the patent technology documents, the technical solution topic tags, and the application scenario topic tags.

[0021] Step S4: Perform quantitative calculation and spatial mapping on the dual-label dataset to obtain the target matrix. The target matrix is ​​used to characterize the potential value of each patented technology solution in different application scenarios.

[0022] In one example, the target matrix used to characterize the development potential of a technical solution in different application scenarios can be a two-dimensional spatial coordinate system.

[0023] Step S5: Based on the target matrix, determine the degree of compatibility between the target technical solution and the target application scenario. The higher the degree of compatibility between the target technical solution and the target application scenario, the higher the potential value of the target technical solution.

[0024] The degree of fit is characterized by the combined relationship between supply saturation and demand intensity.

[0025] In the specific implementation process, the search strategy adopted by this application for obtaining patent technology documents includes at least two levels of screening: The first level is industry-scope anchoring: by using the International Patent Classification (IPC) or Common Patent Classification (CPC) codes related to the target industry, the industry to which the patent technology documents belong is limited to ensure that the obtained patent technology documents are highly relevant to the target industry field. The second level is technology direction filtering: based on the industry limitation, keywords related to the target core technology field are introduced as technology filtering conditions to further screen the patent technology documents to obtain a set of patents closely related to the target technology. Through the above multi-level search strategy, patent technology documents that simultaneously possess industry relevance and technology relevance are formed. Based on the above search strategy, patent technology documents that are published within a certain time window, closely related to the target industry and core technology, and are in the effective stage can be obtained. The content of the patent technology documents specifically includes structured information such as publication number, title, abstract information, claims information, specification, and publication date.

[0026] The aforementioned method for matching patent technology solutions with application scenarios extracts the technical solution theme tags and application scenario theme tags corresponding to each patent technology document from the same data source, generating a dual-label dataset. This achieves a simultaneous characterization of the technology supply-side structure and the scenario demand-side distribution, providing a two-dimensional data input for supply-demand matching analysis and overcoming the limitations of existing single-theme models that struggle to distinguish between the semantics of technology supply and scenario demand. Furthermore, a target matrix is ​​constructed to characterize the potential value of technical solutions in different application scenarios. Based on this target matrix, combinations of technical solution theme tags and application scenario theme tags with different potential values ​​are identified to determine the degree of fit between the target technical solution and the target application scenario. This establishes a clear mapping relationship between patent technology solutions and application scenarios, moving away from the traditional model that relies on qualitative experience or simple statistics to evaluate technology value. This method can efficiently identify the degree of fit between patent technology solutions and application scenarios, thereby efficiently identifying potentially valuable innovation opportunities.

[0027] Furthermore, when acquiring patent technology documents in the target industry that are currently in a valid stage, the patent technology documents can be systematically cleaned and standardized preprocessed to remove noise, unify the format, and construct a high-quality structured patent technology document corpus that can be directly used for model training. The structured information of each patent technology document in a valid protection state is extracted and formatted uniformly. This structured information includes publication number, title, abstract, claims, background technology, legal status, and publication date.

[0028] Specifically, the process begins with structured information extraction and format standardization. Then, all extracted text undergoes unified encoding, garbled character removal, and abnormal symbol filtering, before being converted to standard plain text format to construct the basic data framework. Next, application status filtering is performed, including: based on the legal status of patent documents, removing all expired, pending, or provisionally published patent applications, retaining only valid patent documents that have been granted, specifically authorized invention patents, to ensure the legal representativeness and stability of the analyzed technical content. Then, patent documents are preprocessed, including: performing the following automated preprocessing operations on the filtered patent documents to reduce noise and improve semantic parsing accuracy: Text segmentation: using word segmentation tools and injecting a custom domain-specific dictionary to improve the accuracy of identifying technical terms and compound words. Stop word filtering: constructing a compound stop word list, adding high-frequency but semantically meaningless function words from the patent documents on top of general stop words, effectively filtering text noise. Character normalization: using regular expressions to remove numbers, punctuation, and non-standard characters to achieve text format standardization. Text vectorization representation: Transform the segmented text sequence into word vectors or sentence vectors. The model and vector dimension used can be selected according to computing resources and accuracy requirements.

[0029] This implementation effectively addresses the problems of insufficient domain relevance, high textual noise interference, and difficulty in simultaneously supporting joint analysis of patent technology solutions and application scenarios in existing technologies, providing a stable data foundation for identifying the development potential of subsequent technical solutions in different application scenarios.

[0030] It is worth noting that, on the one hand, the descriptions of technical content in patent documents are scattered and varied, making it difficult to form a unified expression of technical supply; on the other hand, the engineering application scenarios corresponding to patent documents are often implicit in explanatory texts, making them difficult to directly identify and express in a structured manner. To solve the above problems, this application transforms the content of unstructured patent documents into structured and alignable semantic units.

[0031] In one embodiment, the step of performing bilateral semantic recognition on the patent documents to obtain the technical solution topic tag and the application scenario topic tag corresponding to each patent document includes: A technology field dictionary is constructed based on the historical technical solutions of the target industry, and an application scenario field dictionary is constructed based on the historical application scenarios of the target industry. Based on the technical field dictionary and semantic recognition model, technical topics are extracted from the abstract information of patent technical documents, and the technical solution topic tags corresponding to each patent technical document are determined according to the probability distribution of each patent technical document on each technical topic. Based on the application scenario domain dictionary and semantic recognition model, scenario themes are extracted from the abstract information, claim information and background technology information of patent technical documents, and application scenario theme tags corresponding to each patent technical document are determined according to the probability distribution of each patent technical document on each scenario theme.

[0032] In this implementation, a technical field dictionary and an application scenario domain dictionary were first constructed. The technical field dictionary was constructed by reviewing authoritative historical literature and technical materials related to the target technical field, extracting core technical terms, and thus forming the technical field dictionary. The application scenario domain dictionary was constructed by combining historical industry research reports and engineering application cases to extract a set of scenario keywords reflecting the actual application needs and engineering problems of the target industry, thus forming the application scenario domain dictionary. Injecting these dictionaries into the word segmentation system enables refined word segmentation driven by domain knowledge. After the above multi-step cleaning and preprocessing, a clean text dataset of high-quality patent technology documents is finally obtained. This dataset is characterized by unified structure, semantic purity, and domain focus.

[0033] Then, the technical solution topic tags are generated as follows: A technical field dictionary and a semantic recognition model are used to train the abstract text set of all patent technology documents to automatically discover hidden semantic clusters representing different technical directions. The semantic recognition model can be any of the following: a latent Dirichlet assignment model, a neural topic model, or a BERTopic model.

[0034] The number of technical topics is determined as follows: The number of technical topics K can be determined by calculating indicators such as the consistency score and confusion level of the model, and by combining domain knowledge for comprehensive judgment. The number of technical topics K can be set within a certain range, generally from 5 to 20, and is not limited to a fixed value.

[0035] Then, for each technical topic output by the semantic recognition model (represented by a set of keywords with the highest probability), the constructed technical field dictionary is invoked for auxiliary interpretation and verification. Through manual or semi-automatic rules, each technical topic is summarized and labeled with a tag with a clear technical meaning, such as "digital twin", "Internet of Things", "computer vision", and "reinforcement learning" as technical solution topic tags.

[0036] Finally, the patent technology documents and technical solution subject tags are associated.

[0037] Furthermore, the text from the abstract, claims, and background information fields of each patent document is concatenated and merged to form a dedicated text for scenario analysis. The merging process can employ simple text connections or assign different weights to different fields.

[0038] Then, a semantic recognition model is used to train the above-mentioned fused scene text set independently in order to discover the hidden semantic clusters that represent different application domains.

[0039] The number of scene themes is determined as follows: The number of scene topics M can be flexibly determined based on the semantic recognition model evaluation indicators and actual analysis needs, and is generally adjusted between 5 and 15.

[0040] Generate application scenario theme tags as follows: The identified scene semantics are aggregated to form several application scenario themes, and an application scenario domain dictionary is invoked for auxiliary interpretation. Through semantic matching and manual verification, each theme is labeled with a clear application scenario theme tag, such as "safety management," "smart construction site," and "structural health monitoring."

[0041] Through the above steps, each patent document is simultaneously assigned a technical solution topic tag and an application scenario topic tag. Organizing all patent documents and their corresponding dual tags generates a structured "patent document - technical solution topic tag - application scenario topic tag" associated dataset. This dataset clearly defines the technical and application attributes of each patent document, directly supporting the subsequent construction of a two-dimensional co-occurrence frequency matrix, calculation of association strength, and drawing of the target matrix. Through the aforementioned bidirectional parallel and independent semantic recognition process, this application achieves efficient separation and structured extraction of technical supply and scenario demand information from mixed text, solving the semantic confusion problem under a single analytical perspective.

[0042] In this implementation, a dual-dictionary-guided semi-supervised topic modeling method was designed and applied. By constructing independent technical field dictionaries and application scenario domain dictionaries, topic modeling and manual verification were performed on patent abstract information (focusing on technology) and fused text (abstract information, claims and background technical information, focusing on scenario) respectively, thereby completing the independent extraction and annotation of technical and scenario topics with high quality.

[0043] Specifically, the quantitative calculation and spatial mapping of the dual-label dataset to obtain the target matrix includes: Based on the dual-label dataset, the topic labels of each technical solution are paired with the topic labels of each application scenario to obtain N combinations of technical solution topic labels and application scenario topic labels; N is an integer greater than or equal to 2; Determine the number of patent documents corresponding to the combination of technical solution topic tags and application scenario topic tags, and obtain a two-dimensional co-occurrence frequency matrix based on the number of patent documents, with technical solution topic tags and application scenario topic tags as rows and columns, respectively. Based on the row and column statistics of the two-dimensional co-occurrence frequency matrix, core indicators are calculated, including supply saturation to characterize the degree of technology supply and demand intensity to characterize the degree of scenario demand. Normalize the supply saturation and demand intensity to obtain normalized supply saturation and normalized demand intensity; A two-dimensional spatial coordinate system is constructed with the normalized supply saturation and the normalized demand intensity as the horizontal and vertical axes, respectively. This two-dimensional spatial coordinate system is used as a target matrix to characterize the potential value of the patented technology solution in different application scenarios. The two-dimensional spatial coordinate system includes four quadrants, each with a different potential value.

[0044] The step of determining the compatibility between the target technical solution and the target application scenario based on the target matrix includes: The target matrix is ​​marked in a set manner according to different potential values, the set manner including color or shape; The combination of the theme tags of each technical solution and the theme tags of the application scenario is mapped to the target matrix; Based on the positional relationship of the combination of technical solution topic tags and application scenario topic tags in the four quadrants of the target matrix, the matching degree of the combination of technical solution topic tags and application scenario topic tags with different potential values ​​is identified. The degree of compatibility between the target technical solution and the target application scenario is determined based on the combination matching degree.

[0045] In this implementation, the generated dual-label dataset undergoes quantitative calculation and spatial mapping to identify technology-scenario combinations with different development trends, such as innovation potential areas with "high demand and low supply" and mature advantage areas with "high demand and high supply." Specifically, the steps include: (1) Data preparation The generated structured dual-label dataset of "patent technology documents - technical solution topic tags - application scenario topic tags" is used as the core input. Based on the above structured dual-label dataset, the statistics of each "technical solution topic tag" are calculated. ")" and "Application Scenario Theme Tags ( The number of patent documents corresponding to each "" is paired to form a two-dimensional co-occurrence frequency matrix with m rows (number of technical solution topic tags) × n columns (number of application scenario topic tags). The value of each cell in the matrix... This means that it is simultaneously labeled as a technology and scene The number of patent technology documents.

[0046] (2) The core indicators are calculated as follows: Calculate supply saturation ( ) Supply saturation ( This is used to measure the concentration of patent literature supply or market penetration saturation of a technology in a specific scenario. In the scene Supply saturation It is calculated as the proportion of its patent technology documents to the total number of patent technology documents in that scenario. The formula is expressed as: ; in, Technical solution topic tags In application scenarios, thematic tags The number of patents under Application scenario theme tags The total number of patents under (i.e., the number of patents in the co-occurrence frequency matrix) The sum of all values ​​in the column. The higher the value, the better the scene. In China, technology The more concentrated the supply, the more mature the application of the technology in this scenario or the more intense the competition.

[0047] Calculate demand intensity ( ) Calculate demand intensity ( This is used to measure the breadth of demand coverage for a technology across different application scenarios. (Regarding technology...) Demand intensity The formula for calculating the number of application scenario topic tags covered by this technology out of the total number of application scenario topic tags is as follows: ; in, Technical solution topic tags The number of application scenario topic tags involved, that is, the number of tags in the co-occurrence frequency matrix. The number of columns in the row where the number of patent technology documents is greater than zero. The total number of topic tags for all application scenarios is n, which is the number of columns in the matrix. The higher the value, the better the technology. The wider its application scope, the greater the demand in different scenarios, and the stronger its cross-scenario migration potential and market adaptability.

[0048] (3) Perform data standardization as follows The min-max normalization method is used to calculate the original... and Standardization is performed, linearly transforming all values ​​of each indicator to the [0,1] interval. After standardization, each technology-scenario combination corresponds to a pair of standardized coordinate values. , ),in Represents the normalized supply saturation. This represents the intensity of demand for normalization.

[0049] Furthermore, the two-dimensional coordinate space is used as the target matrix to characterize the development potential of the technical solution in different application scenarios, as follows: With normalized supply saturation The horizontal axis (X-axis) represents the normalized demand intensity. Establish a two-dimensional coordinate system with the vertical axis (Y-axis). Treat each combination of technical solution hashtag and application scenario hashtag as a data point, and determine its coordinates (...). , (Draw it in this coordinate system.)

[0050] Specifically, to distinguish different types of combinations, a threshold needs to be set. This is done using all data points... The median of the values The median of the values ​​is used as the dividing line. Therefore, the coordinate system is divided into four quadrants: First Quadrant (High Potential Zone): high, Low. The combination of technical solution tags and application scenario tags within this quadrant is in high demand but currently insufficient in supply, representing a potential innovation vacuum or blue ocean market with high R&D and investment value.

[0051] Second Quadrant (Star Zone): high, High. The demand for and supply of combinations of technical solution tags and application scenario tags within this quadrant are extensive and sufficient, indicating that technical solutions and scenarios are well matched, forming a relatively mature and active market, which is the core area of ​​current industrial development.

[0052] Third Quadrant (Weak Area): Low, Low. The demand and supply of the combination of technical solution tags and application scenario tags in this quadrant are both weak, and the technology may be in its infancy, exploration stage, or be a marginal application with limited potential for short-term breakthroughs.

[0053] Fourth Quadrant (Mature Zone): Low, High. In this quadrant, the demand for the combination of technical solution tags and application scenario tags is relatively narrow, but the supply is concentrated. This may indicate that the technical solution is already highly mature or even saturated in this specific scenario, with limited room for growth, and it is necessary to seek expansion into new scenarios.

[0054] In this way, a target matrix based on two indicators (supply saturation O and demand intensity S) was constructed. "Supply saturation" is defined as the concentration of patent technology literature for a certain technology in a certain scenario, and "demand intensity" is defined as the breadth of scenarios covered by a certain technology. These two core quantitative indicators objectively divide the combination of technical solution theme tags and application scenario theme tags into four quadrants: "high potential zone", "star zone", "mature zone" and "weak zone", thereby achieving accurate, visual positioning and priority ranking of innovation opportunities.

[0055] In one embodiment, technology lifecycle prediction can also be performed based on the target matrix, including: The number of patent documents labeled with the same technical solution theme tag each year is counted, and for each technical solution theme tag, the cumulative number of patents from the starting year to each year is calculated to obtain time series data; The logistic growth model is used to fit the cumulative patent growth trajectory of each technical solution's topic tag, satisfying the following relationship:

[0056] in, Indicates the cumulative number of patents. This represents the potential saturation value of the growth curve. It is a constant related to the initial value. It is a growth rate parameter. For time; The model parameters corresponding to each technical solution's topic label are estimated by fitting the time series data of each technical solution's topic label using the nonlinear least squares method. , , ; Based on the growth trajectory and model parameters obtained from the fitting, the typical life cycle stage of each technical solution's topic tag is quantitatively determined.

[0057] In this implementation, in a two-dimensional coordinate system, in addition to plotting scatter points, different colors or shapes can be used to distinguish quadrants, technology-scenario labels can be added, and a median baseline can be drawn to ultimately generate an intuitive OS-Matrix innovation opportunity map. By introducing a time dimension, OS-Matrix maps for different time periods can be constructed separately, and by comparing the movement trajectories of observation points, the dynamic evolution trend of technology-scenario combinations can be analyzed.

[0058] Specifically, the time-series patent data is first constructed as follows: Based on the technical solution topic tag data generated in the above steps, for each technical solution topic tag T, count the number of patents tagged with that technical solution topic tag each year. When making the statistics, the patent application year or publication year can be used as a priority, and the start and end years can be set according to data availability.

[0059] For each technical solution topic tag T, calculate the cumulative number of patents from the starting year to each year, forming a time series data with time as the horizontal axis and the cumulative number of patents as the vertical axis.

[0060] The cumulative patent growth trajectory of the technical solution topic tags is fitted using a logistic growth model. The model is in the following form:

[0061] in, Indicates the cumulative number of patents. This represents the potential saturation value of the growth curve. It is a constant related to the initial value. It is a growth rate parameter, reflecting how fast or slow the growth is. For time.

[0062] Model parameters are estimated using the nonlinear least squares method, and... The goodness of fit is tested with the residual sum of squares (RSS). Based on the model parameters, this application identifies the life cycle stages of each technology: high growth rate (b) corresponds to the growth stage, slowing growth represents the maturity stage, and stabilization or decline reflects the decline trend.

[0063] By using optimization algorithms such as nonlinear least squares, the time series data of each technical solution's topic tag are fitted to estimate the model parameters corresponding to that technology. , , The fitting process can be implemented using programming software like Python or statistical tools, and the goodness of fit of the model can be determined by the coefficients. Evaluation was conducted using metrics such as the residual sum of squares (RSS). The closer the value is to 1, the better the fit.

[0064] Based on the fitted growth curves and their characteristic parameters, the typical life cycle stage of each technical solution's topic tag is quantitatively determined. The technology life cycle is typically divided into four main stages, and the typical characteristics and criteria for each stage are as follows: Emerging stage: The accumulated number of patent technology documents is small, the number of newly added patent technology documents each year is small and fluctuates greatly, the growth curve is at the bottom of the S-shape, the growth rate begins to rise slowly but the absolute value is low.

[0065] Growth stage: The cumulative amount of patent technology documents accelerates, the number of newly added patent technology documents per year rises rapidly, the growth curve is located in the steep part of the middle of the S-shape, and the growth rate reaches near the peak. This stage can be further divided into the early growth stage and the rapid growth stage.

[0066] Maturity stage: The cumulative growth of patent technology documents slows down, and the number of newly added patent technology documents each year remains at a high level or begins to decline. The growth curve approaches or reaches its potential saturation point. The growth rate declined significantly and approached zero.

[0067] Decline phase: The number of newly added patent technology documents per year continues to decline, and the cumulative number of patent technology documents is close to or has reached saturation. The growth curve has passed its inflection point and tends to plateau, and may even decline further.

[0068] To determine which stage of the life cycle a technology is in, calculate the current cumulative number of patent documents. With predicted saturation value ratio This ratio measures the relative maturity of technological development. A ratio below 0.2 may indicate the nascent stage, 0.2-0.8 may indicate the growth stage, and a ratio above 0.8 may indicate the mature stage.

[0069] In this implementation, a method for quantitatively predicting the technology lifecycle by integrating a Logistic growth model is proposed. The cumulative historical patent literature data for each technology's thematic tags is fitted to a Logistic curve. By estimating its potential saturation value (K) and growth rate parameter (b), and combining this with the current cumulative saturation ratio, the method quantitatively determines whether the technology is in the "embryonic stage," "growth stage," or "mature stage," providing a forward-looking dynamic trend insight for technology deployment.

[0070] The following is a specific example illustrating the method for adapting the above-mentioned patented technical solution to its application scenario.

[0071] Patent technology document data collection and preprocessing includes the following processes: In this embodiment, patent technology documents in the Architecture, Engineering and Construction Industry (AEC) are used as the analysis object to construct a high-quality dataset of patent technology documents for subsequent analysis.

[0072] (1) Collection of patent technology literature data To obtain patent literature data that is both industry-relevant and technically targeted, this embodiment designed and implemented a two-stage search strategy: Data source: The commercial patent database PatSnap was selected as the data source, and the publication date range of patent literature was set from January 2008 to July 2025, with the patent literature type limited to invention patents. First-level search (industry anchoring): This was used to limit the industry scope to which the patents belong.

[0073] Second-level search (technology filtering): Based on the results of the first-level search, further filtering is performed using keywords related to artificial intelligence (AI) technologies to focus on AI-related patents.

[0074] (2) Initial dataset construction After executing the above search strategy, a total of 6,216 patent records were obtained, forming the initial dataset. Each patent record contains structured fields such as publication number, title, abstract information, claims information, full text of the specification, and publication date.

[0075] (3) Patent data validity screening and preprocessing Structured information extraction and format standardization: From 6,216 patent documents, the system extracts the text content of six core fields: "Publication Number", "Title", "Abstract Information", "Claims Information", "Background Technology Information" and "Publication Date", and performs unified encoding and format standardization.

[0076] Application status screening: Based on whether the patent technology documents are under valid protection, all expired, rejected, or only temporarily disclosed patent technology documents are manually removed, retaining only those that have been granted or are under valid examination. This step ensures the technical validity and legal certainty of the analysis sample.

[0077] Automated text preprocessing performs the following operations on the filtered patent technology documents: Word segmentation: The "jieba" word segmentation tool was used for Chinese word segmentation. To improve the accuracy of domain terminology segmentation, a custom professional dictionary for the construction field (containing terms such as "digital twin", "slope monitoring", and "prefabricated construction") was loaded.

[0078] Noise reduction: A composite stop word list was constructed, which, based on general Chinese stop words, added words that frequently appear in patent documents but have no practical analytical value, such as "this invention", "described", and "example".

[0079] Cleaning: Use regular expressions to remove pure numeric sequences, special punctuation marks, and other non-text characters from the text, achieving further text purification.

[0080] (4) Domain dictionary construction This implementation case constructs two domain dictionaries.

[0081] Technical Field Dictionary: By systematically reviewing 68 authoritative academic papers and technical reports related to AI in the AEC field, 144 core technical terms such as "reinforcement learning," "convolutional neural network," "robotics," and "generative adversarial network" were manually extracted to form the technical field dictionary.

[0082] Application Scenario Domain Dictionary: By analyzing industry white papers, technical solutions, and typical engineering cases, 381 scenario keywords such as "safety management", "structural health monitoring", "smart construction site", "green building", and "quality management" have been extracted to form the application scenario domain dictionary.

[0083] The two dictionaries mentioned above are injected into the word segmentation system through the jieba.load_userdict() function to achieve refined word segmentation based on domain knowledge.

[0084] (5) Final dataset After the complete collection, screening, cleaning, and enhancement process described above, a clean dataset containing 5,366 high-quality patent texts was finally obtained. This dataset has the following characteristics: 1) Each patent document is associated with complete structured field information; 2) The text noise is extremely low and the semantics are pure; 3) Through domain dictionary injection, key technologies and scenario terms are accurately identified and preserved.

[0085] The semantic recognition on the technology supply and application scenario side includes the following processes: Based on the 5,366 clean patent technology documents obtained above, the technology supply theme and application scenario theme tags were separated and extracted from the mixed text in a structured manner, and each patent technology document was simultaneously labeled with dual tags.

[0086] (1) Semantic recognition of technology supply side ① Selection of technical feature text: From the preprocessed text of each patent document, the "Abstract Information" field is selected as the main input for technical semantic recognition. The abstract information is chosen because it usually concisely summarizes the core technical solution and innovation of the invention.

[0087] ② Technical Topic Model Training: The Latent Dirichlet Allocation (LDA) topic model was used to train the model on all 5,366 patent abstract texts. The model training parameters were set as follows (all adjustable parameters): the number of topics k was selected as 10 after multiple adjustments; the number of iterations was set to 30; hyperparameters α=0.1 and η=0.01 to promote topic concentration; Gibbs sampling was used for inference.

[0088] ③ Extraction and Labeling of Technical Solution Topics: After model training, 10 technical topics are output, each represented by a set of keywords with the highest probability. Combining the "Technical Field Dictionary" constructed in Example 1, each topic is manually semantically summarized and named. The final 10 technical solution topic tags are: Image Processing, Reinforcement Learning, Digital Twin, UAV, Robotics, Machine Learning, Attention Mechanism, IoT, Feature Extraction, and Neural Networks.

[0089] ④ Association of Patent Technical Documents with Technical Solution Theme Tags: Based on the probability distribution of the abstract text of each patent technical document across 10 technical solution theme tags, it is assigned to one or more themes with the highest probability. For example, a patent about using cameras and deep learning algorithms to monitor the wearing of safety helmets by construction workers has the highest probability of its abstract text on the themes of "Image Processing" and "Neural Networks," so it is tagged with both of these technical tags.

[0090] (2) Semantic recognition on the application scenario side ① Scene Feature Text Fusion: In order to fully capture scene information, the texts of the three fields of "abstract information", "claim information" and "background technology information" of each patent technical document are spliced ​​together to form a long text of scene analysis corresponding to each patent technical document.

[0091] ② Scene Topic Model Training: The LDA model was used to independently train the 5,366 scene texts obtained from the fusion process. The number of application scene topic tags was determined to be 10 after evaluation. Other training parameters were set similarly to those of the technical model but were adjusted independently.

[0092] ③ Application Scenario Topic Tag Extraction and Labeling: The model outputs 10 scenario topics and their keywords. These are then manually verified and named using the constructed "Application Scenario Domain Dictionary." The final 10 application scenario topic tags are: Human-Machine Interaction, Safety Management, Smart Construction Site, Smart Building, Intelligent Equipment Detection, Intelligent Equipment, Intelligent Early Warning, Structural Health Monitoring, Green Buildings, and Quality Control.

[0093] ④ Association of Patent Technology Documents with Application Scenarios: Based on the probability distribution of the scenario-integrated text of each patent technology document across 10 application scenario theme tags, it is associated with one or more of the most relevant application scenario theme tags. For example, for a patent on safety helmet monitoring, if the scenario text is prominently displayed on the themes of "Safety Management" and "Smart Construction Site," then both scenario tags will be used to label it.

[0094] (3) Generation of dual-label datasets After the aforementioned bidirectional independent identification, each patent document was assigned at least one technical solution topic tag and at least one application scenario topic tag. All results were integrated to generate a structured dataset linking "patent - technical solution topic tag - application scenario topic tag". This dataset forms the basis for all subsequent quantitative analyses.

[0095] The construction of the target matrix includes the following process: (1) Construct a two-dimensional co-occurrence frequency matrix The dataset generated in Example 2 is used as input. This dataset contains 10 technical solution topic tags and 10 application scenario topic tags. The number of patents corresponding to each pair of "technical solution topic tag (T)" and "application scenario topic tag (S)" (100 combinations in total) is counted, and a 10x10 two-dimensional co-occurrence frequency matrix is ​​constructed. The value of each cell in the matrix... This refers to the number of patents that are simultaneously labeled as technology (T) and scenario (S).

[0096] Some statistical results are as follows: The number of patents co-occurring in the technologies "Internet of Things (IoT)" and the scenarios "Safety Management" .

[0097] The number of patent documents showing the co-occurrence of the technology "Digital Twin" and the scenario "Smart Construction Site" .

[0098] The number of patent documents that co-occur with the technology "Reinforcement Learning" and the scenario "Green Buildings" .

[0099] (2) Calculation of core indicators For each of the 100 technology-scenario combinations in the co-occurrence matrix, calculate their supply saturation (O) and demand intensity (S).

[0100] ① Calculate supply saturation (O): This measures the concentration of patent supply for a particular technology in a specific scenario. The formula is as follows:

[0101] in, For technology In the scene The number of patents under For the combination (IoT, Safety Management), the total number of patents related to the "Safety Management" scenario. .

[0102] but .

[0103] ② Calculate Demand Intensity (S): This measures the breadth of demand coverage for a technology across different scenarios.

[0104] in, Representation technology The number of application scenario topic tags covered .

[0105] Regarding the technology "feature extraction," analyzing its corresponding rows in the matrix revealed that it had a patent count greater than 0 in 7 out of 10 scenarios. .

[0106] ③ Data standardization: Use the min-max normalization method to normalize all 100... Value and 10 The values ​​are linearly transformed to the [0,1] interval to obtain the normalized supply saturation. and demand intensity .

[0107] (3) Matrix construction and quadrant division ① Establish a coordinate system: using normalized supply saturation The horizontal axis (X-axis) represents the normalized demand intensity. For technology A unified ordinate (Y-axis, note the differences between techniques) (Maintaining the same status across all scenario combinations), the combination of technical solution theme tags and application scenario theme tags is plotted as data points in a two-dimensional coordinate system.

[0108] ② Divide the space into four quadrants: The High Potential zone (high demand – low supply) includes 23 items, such as Reinforcement Learning – Safety Management and Feature Extraction – Safety Management. The strong demand but insufficient supply indicates that it is still in the initial stage in the fields of construction safety and risk identification.

[0109] The Star Zone (High Demand – High Supply) comprises 25 items, including combinations such as Digital Twin – Safety Management and IoT – Smart Building. With ample patent accumulation and mature applications, it represents the core growth area for AI in the construction field.

[0110] The Weak Zone (low demand – low supply) comprises nine items, including combinations such as Image Processing and Green Buildings. Both supply and demand are weak, reflecting that this sector has not yet formed an effective market driving force.

[0111] The Mature zone (low demand – high supply) contains 14 items, including combinations such as IoT – Quality Control and Digital Twin – Smart Building. The technology supply is sufficient, but demand growth is slowing down, and further growth needs to be achieved through scenario innovation.

[0112] (4) The results are visualized and interpreted as follows: like Figure 2As shown, the target matrix (OS-Matrix) of this invention is generated. The combination of high-potential areas (top left) is identified as the innovation direction that should be given priority and resources should be invested in in the future; the combination of star areas (top right) is the pillar of current industrial development and needs to be further deepened; the combination of weak areas (bottom left) and mature areas (bottom right) provides risk warnings and transformation references for technology layout.

[0113] Technology lifecycle prediction includes the following processes: Based on the technical solution topic tags and their corresponding patent data identified in Example 2, this embodiment quantitatively evaluates the relative maturity of various AI technologies in the AEC field and predicts their future growth trajectory and saturation point.

[0114] (1) Time series patent data Based on the data linking "patent technology documents and technical solution topic tags", we first count the number of patents marked with technical solution topic tag T each year according to the publication year of the patent technology documents (2008 to 2025). Then, for each technical solution topic tag T, we calculate the cumulative number of patents from the starting year (2008) to each year. We construct a time series data for each technical solution topic tag, with the horizontal axis representing the year (2008-2025) and the vertical axis representing the corresponding cumulative number of patents. Finally, we obtain 10 time series curves representing different technology growth trajectories, which serve as the basis for model fitting.

[0115] (2) Growth curve model fitting A Logistic growth model is used to fit each time series to reveal its inherent growth pattern and predict future trends. The model formula is: (6) in, Indicates the cumulative number of patents. This represents the potential saturation value of the growth curve. It is a constant related to the initial value. It is a growth rate parameter, reflecting how fast or slow the growth is. For time.

[0116] The fitting was performed using the nonlinear least squares method. The key parameter estimation results and stage discrimination are as follows: ① Internet of Things (IoT): Fit saturation value K≈12,572. Current cumulative growth has slowed significantly, and the goodness of fit is... High. Considering that the current cumulative number of patents is nearing saturation, it is determined that it has entered the mature stage and is a fundamental supporting technology for the industry.

[0117] ② Digital Twin: Fit saturation value K≈10,268. Its growth curve characteristics are similar to IoT, with a current maturity of approximately 29.2%, placing it in the early stages of maturity.

[0118] ③ Machine Learning: Fit saturation value K≈9,013. Maintaining a high growth rate (large b value), the current maturity is about 35.3%, which is in the later stage of high-speed growth, showing strong growth momentum and cross-scenario transfer potential.

[0119] ④ Robotics and Image Processing: The current maturity levels of both are approximately 22.8% and 19.0% respectively, both in the growth stage, with rapid patent growth.

[0120] ⑤ Neural Networks and Feature Extraction: The growth curve has stabilized, and the growth rate parameter b is relatively small, indicating that the method system has become relatively stable.

[0121] ⑥ Unmanned Aerial Vehicles (UAVs), Attention Mechanisms, and Reinforcement Learning: These three technologies show very high growth rate parameters (b-values) and steep growth curves. Their current maturity levels are relatively low (approximately 12.1%, 15.2%, and 13.2% respectively), but they are in a period of rapid growth and are considered to be in a growth phase, representing potential future innovation hotspots.

[0122] (3) Visualization of life cycle stages Based on the model fitting parameters, especially the saturation value K and the growth rate b, and the ratio of the current cumulative number of patent technology documents to K, a comprehensive judgment is made on the topic tags of all technical solutions. The technology lifecycle exhibits significant differentiation: Stage Trends: AI technology in the construction industry was in the introduction phase from 2010 to 2020; it entered the growth phase rapidly from 2020 to 2030; and it is expected that most technologies will reach maturity after 2030. Currently, we are in a critical stage of transitioning from the growth phase to the maturity phase.

[0123] Technological Differences: Different technologies mature at significantly different paces. Technologies such as the Internet of Things (IoT) and digital twins have matured first; machine learning and robotics are in a rapid growth phase; while emerging algorithms such as reinforcement learning and attention mechanisms are on the verge of explosive growth.

[0124] This application also provides a system for adapting a patented technical solution to an application scenario, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the steps of the aforementioned method for adapting a patented technical solution to an application scenario. This system can implement various embodiments of the method for adapting a patented technical solution to an application scenario and achieve the same beneficial effects, which will not be elaborated upon here.

[0125] This application also provides a computer program product stored in a storage medium. The program product is executed by at least one processor to implement the steps of the adaptation method for the patented technical solution and application scenario described above. This computer program product can implement various embodiments of the adaptation method for the patented technical solution and application scenario, and can achieve the same beneficial effects, which will not be elaborated here.

[0126] The above description is merely a specific embodiment of this disclosure, but the scope of protection of this disclosure is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this disclosure should be included within the scope of protection of this disclosure. Therefore, the scope of protection of this disclosure should be determined by the scope of the claims.

Claims

1. A method for adapting a patented technical solution to an application scenario, characterized in that, include: Obtain patent technology documents that are under effective protection in the target industry; Perform dual-dimensional semantic recognition on the patent technology documents to obtain the technical solution topic tags and application scenario topic tags corresponding to the patent technology documents; A dual-label dataset is generated based on the patent technology documents, the technical solution topic tags, and the application scenario topic tags. The dual-label dataset is quantitatively calculated and spatially mapped to obtain a target matrix, which is used to characterize the potential value of each patented technology solution in different application scenarios. Based on the target matrix, the degree of compatibility between the target technical solution and the target application scenario is determined. The higher the degree of compatibility between the target technical solution and the target application scenario, the higher the potential value of the target technical solution.

2. The method according to claim 1, characterized in that, The content of the patent document includes abstract information, claims information, and background technology information; The step of performing dual-dimensional semantic recognition on the patent technology document to obtain the technical solution topic tags and application scenario topic tags corresponding to the patent technology document includes: A technology field dictionary is constructed based on the historical technical solutions of the target industry, and an application scenario field dictionary is constructed based on the historical application scenarios of the target industry. Based on the technical field dictionary and semantic recognition model, technical topics are extracted from the abstract information of patent technical documents, and the technical solution topic tags corresponding to each patent technical document are determined according to the probability distribution of each patent technical document on each technical topic. Based on the application scenario domain dictionary and semantic recognition model, scenario themes are extracted from the abstract information, claim information and background technology information of patent technical documents, and application scenario theme tags corresponding to each patent technical document are determined according to the probability distribution of each patent technical document on each scenario theme.

3. The method according to claim 1, characterized in that, The quantitative calculation and spatial mapping of the dual-label dataset to obtain the target matrix includes: Based on the dual-label dataset, the topic labels of each technical solution are paired with the topic labels of each application scenario to obtain N combinations of topic labels of technical solutions and application scenarios; N is an integer greater than or equal to 2. Determine the number of patent documents corresponding to the combination of technical solution topic tags and application scenario topic tags, and obtain a two-dimensional co-occurrence frequency matrix based on the number of patent documents, with technical solution topic tags and application scenario topic tags as rows and columns, respectively. Based on the row and column statistics of the two-dimensional co-occurrence frequency matrix, core indicators are calculated, including supply saturation to characterize the degree of technology supply and demand intensity to characterize the degree of scenario demand. Normalize the supply saturation and demand intensity to obtain normalized supply saturation and normalized demand intensity; A two-dimensional spatial coordinate system is constructed with the normalized supply saturation and the normalized demand intensity as the horizontal and vertical axes, respectively. This two-dimensional spatial coordinate system is used as a target matrix to characterize the potential value of the patented technology solution in different application scenarios. The two-dimensional spatial coordinate system includes four quadrants, each with a different potential value.

4. The method according to claim 3, characterized in that, The step of determining the compatibility between the target technical solution and the target application scenario based on the target matrix includes: The target matrix is ​​marked in a set manner according to different potential values, the set manner including color or shape; The combination of the theme tags of each technical solution and the theme tags of the application scenario is mapped to the target matrix; Based on the positional relationship of the combination of technical solution topic tags and application scenario topic tags in the four quadrants of the target matrix, the matching degree of the combination of technical solution topic tags and application scenario topic tags with different potential values ​​is identified. The degree of compatibility between the target technical solution and the target application scenario is determined based on the combination matching degree.

5. The method according to claim 3, characterized in that, The formula for calculating the supply saturation is as follows: ； in, To achieve supply saturation, Technical solution topic tags In application scenarios, thematic tags The number of patents under Application scenario theme tags The total number of patents under [the relevant authority].

6. The method according to claim 3, characterized in that, The formula for calculating the demand intensity is as follows: ； in, For demand intensity, Technical solution topic tags The number of application scenario topic tags involved. This represents the total number of topic tags for all application scenarios.

7. A system for adapting a patented technical solution to an application scenario, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the method described in any one of claims 1-6.

8. A computer program product, characterized in that, The program product is stored in a storage medium and is executed by at least one processor to perform the steps of the method as described in any one of claims 1-6 above.

Images