Advanced technology maturity grade interval discriminant analysis method and device
By combining large language models and data quality scoring models, we have achieved automated technology maturity level assessment, which solves the problems of low efficiency and error-proneness of traditional methods, adapts to the assessment needs of multiple fields, and improves the accuracy and efficiency of assessment.
Patent Information
- Application Number
- CN202511744514.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-11-25
- Publication Date
- 2026-03-17
AI Technical Summary
Current technology maturity assessments mainly rely on manual judgment or statistical models, which are highly subjective, inefficient, unable to adapt to big data and rapid technological evolution, and have poor cross-domain adaptability, making it difficult to capture emerging technology terms in real time.
A large language model is used for automated collection and preprocessing of multi-source data. Combined with a data quality scoring model and manual sampling verification, the system achieves automated technology maturity level determination through triple data extraction and transition keyword matching.
It significantly improves the efficiency of processing massive amounts of data, reduces subjective bias, ensures data reliability and accuracy, adapts to multi-domain assessment needs, and reduces cross-domain adaptation costs.
Smart Images
Figure CN121681801A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of technology maturity technology, and in particular to a method and apparatus for discriminant analysis of the maturity level intervals of cutting-edge technologies. Background Technology
[0002] In simple terms, technology maturity is a standardized measure, such as the commonly used TRL 1-9 levels, to determine which stage a technology is currently in "from theoretical idea to practical use". Its significance lies in clearly defining technological progress and helping people understand "what the technology can do now, what stage of development the technology is currently in, the degree to which the technology meets the expected goals of the project or engineering, and what the next step should be". This reduces information gaps, avoids blind investment, and promotes the implementation of engineering or projects or future investment decisions more efficiently. The Critical Technology Elements (CTEs) are the core carriers of technology maturity. Their selection mainly relies on two points: first, the importance of the CTEs and whether their evolution has a significant impact on project delivery and system development progress; and second, the novelty of the CTEs and whether their technological evolution has led to significant technological risks in their application. Among these, transition keywords are highly recognizable and iconic terms in the process of technological evolution. These keywords are the core basis for measuring the process of technology transitioning from the TRL level to a higher level and provide an important reference for technology development assessment. However, the current traditional maturity assessment mainly relies on manual judgment by science and technology intelligence personnel or statistical models, which is highly subjective and inefficient. When faced with large amounts of data and rapid technological evolution, manual judgment will be inefficient and prone to errors. Summary of the Invention
[0003] In view of this, embodiments of the present invention provide a method for discriminant analysis of the maturity level interval of cutting-edge technologies, so as to eliminate or improve one or more defects existing in the prior art.
[0004] One aspect of the present invention provides a method for discriminant analysis of the maturity level interval of cutting-edge technologies, the method comprising the following steps: Obtain the pre-collected technical text of the technology to be judged and construct it into the original dataset; The original dataset is preprocessed to obtain an updated dataset. In the preprocessing step, each data item is filtered to determine whether to delete it. Each data point in the updated dataset is input into a preset data extraction model. The data extraction model extracts triplet data composed of technical direction, technical dynamics and technical indicators from each data point in the updated dataset. The data extraction model is a large language model. The transition level of the technology to be judged is determined by matching the preset transition keyword library with the updated dataset. The technical information, including the triplet data and transition level, is added to the preset structured JSON prompt template. The completed structured JSON prompt template is then input into the preset large language model, which outputs the maturity level.
[0005] By adopting the above approach, this solution overcomes the bottleneck of reliance on manual methods. Through automated collection and preprocessing of multi-source data, combined with the zero-sample information extraction capability of large models, it replaces the traditional manual judgment mode, significantly improving the efficiency of processing massive amounts of data. The introduction of a data quality scoring model and a manual sampling verification mechanism ensures data reliability from the source, reduces errors caused by subjective bias, and solves the problems of low efficiency and error-proneness of traditional methods.
[0006] In some embodiments of the present invention, the method further includes inputting reflection prompts into the large language model and receiving the maturity assessment level output by the large language model again.
[0007] In some embodiments of the present invention, the step of filtering each piece of data and determining whether to delete the data includes calculating a data quality score for each piece of data using the following formula, and determining whether to delete the data based on the data quality score; in, This represents the quality score of a single data point. For data relevance, For data integrity, For data timeliness, α, β, and γ are all preset weighting coefficients.
[0008] In some embodiments of the present invention, each transition level is provided with a transition keyword lexicon. In the step of determining the transition level of the technology to be determined by matching the updated dataset with the preset transition keyword lexicon, the updated dataset is matched with the transition keyword lexicon corresponding to each transition level, and the matching degree between the updated dataset and the transition keyword lexicon corresponding to each transition level is calculated. The transition level corresponding to the updated dataset is determined based on the matching degree.
[0009] In some embodiments of the present invention, in the step of calculating the matching degree between the updated dataset and the transition keyword lexicon corresponding to each transition level, the matching degree is calculated using the following formula: ; in, This represents the keyword matching degree of the k-th TRL level in the keyword library corresponding to the currently calculated transition level. The updated dataset contains the number of matching transition keywords in the transition keyword lexicon corresponding to the currently calculated transition level. This represents the total number of transition keywords in the keyword library corresponding to the currently calculated transition level. A value ≥0.6 is considered a keyword condition that satisfies the transition level.
[0010] In some embodiments of the present invention, the method further includes calculating the weight of each key technical element in the technology to be determined based on the technical field in which the technology to be determined is located, and supplementing the calculated weight of each key technical element as technical information into a preset structured JSON prompt word template.
[0011] In some embodiments of the present invention, in the step of calculating the weight of each key technical element in the technology to be determined based on the technical field in which the technology to be determined is located, the weight of each key technical element is calculated using the following formula: ; in, Let i be the weight of the i-th key technical element. The importance of the i-th key technology element relative to the j-th key technology element is scored from 1 to 9, where 1 is equally important and 9 is extremely important. n is the total number of CTEs. The top 3 CTEs with the highest weights are designated as core CTEs and are used primarily for subsequent maturity assessment.
[0012] In some embodiments of the present invention, the step of inputting the completed structured JSON prompt template into a preset large language model further includes: based on the target field of the obtained structured prompt, using a data extraction model to extract the target data corresponding to the target field from the updated dataset, completing the completion of the structured JSON prompt template, calculating the data integrity of the completed structured JSON prompt template, and determining whether to input the completed structured JSON prompt template into the preset large language model based on the data integrity.
[0013] In some embodiments of the present invention, in the step of inputting the completed structured JSON prompt template into a preset large language model, and the large language model outputting the maturity judgment level, the large language model outputs the preliminary judgment level, the number of judgment rules that meet the preliminary judgment level, and the total number of judgment rules for the preliminary judgment level. The determination confidence level is calculated based on the number of determination criteria that meet the preliminary determination level and the total number of determination criteria for the preliminary determination level. Based on the determination confidence level, it is determined whether to use the preliminary determination level as the final maturity determination level.
[0014] A second aspect of the present invention also provides a cutting-edge technology maturity level interval discrimination analysis device, the device including a computer device, the computer device including a processor and a memory, the memory storing computer instructions, the processor being used to execute the computer instructions stored in the memory, and when the computer instructions are executed by the processor, the device implements the steps of the method described above.
[0015] A third aspect of the present invention also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the aforementioned advanced technology maturity level interval discrimination analysis method.
[0016] Additional advantages, objects, and features of the invention will be set forth in part in the description which follows, and will also become apparent in part to those skilled in the art upon studying the text, or may be learned by practice of the invention. The objects and other advantages of the invention will become apparent from the description and the accompanying drawings.
[0017] Those skilled in the art will understand that the objectives and advantages achievable with the present invention are not limited to those specifically described above, and that the above and other objectives achievable with the present invention will become clearer from the following detailed description. Attached Figure Description
[0018] The accompanying drawings, which are provided to further illustrate the invention and form part of this application, are not intended to limit the scope of the invention.
[0019] Figure 1 This is a schematic diagram of one embodiment of the method for discriminant analysis of the maturity level range of cutting-edge technologies of the present invention; Figure 2 This is a schematic diagram illustrating another implementation of the method for discriminant analysis of the maturity level range of cutting-edge technologies of the present invention; Figure 3 This is a schematic diagram of the overall process of this solution; Figure 4 This is a schematic diagram of the overall architecture of this solution. Detailed Implementation
[0020] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the embodiments and accompanying drawings. Here, the illustrative embodiments and descriptions of this invention are used to explain the invention, but are not intended to limit the invention.
[0021] It should also be noted that, in order to avoid obscuring the invention with unnecessary details, only the structures and / or processing steps closely related to the solution according to the invention are shown in the accompanying drawings, while other details that are not closely related to the invention are omitted.
[0022] The purpose of this solution is to provide a discriminant analysis method for the maturity level intervals of cutting-edge technologies. This addresses the shortcomings of the traditional maturity assessment methods mentioned in the background, which mainly rely on manual assessment by scientific and technological intelligence personnel or statistical models. These methods are highly subjective and inefficient. Furthermore, manual assessment is prone to errors, especially when dealing with large amounts of data and rapid technological evolution. Secondly, the identification of transition keywords is limited by static technology roadmaps, making it impossible to capture emerging technology terms in real time. The element classification systems and assessment standards are not universal across different fields, requiring redevelopment for model porting, which is costly and time-consuming, making it difficult to adapt to the assessment needs of multiple fields or emerging technologies. Thirdly, maturity assessment models rely excessively on domain-specific customization, resulting in high algorithm porting costs and weak cross-scenario adaptability, such as the difficulty in universally applying element classification systems for materials and digital technologies.
[0023] like Figure 1 As shown, this invention proposes a discriminant analysis method for determining the maturity level interval of cutting-edge technologies. The steps of the method include: Step S100: Obtain the pre-collected technical text of the technology to be judged and construct it into the original dataset; like Figure 3 As shown, in some embodiments of the present invention, text is collected in batches from a determined multi-source data collection scope, specifically including the collection of think tank reports, technical papers, industry news and official technical statements, covering the entire lifecycle of technology from "theory-research-testing-deployment" to form an original dataset.
[0024] Step S200: Preprocess the original dataset to obtain an updated dataset. In the preprocessing step, each data item is filtered to determine whether to delete it. In the specific implementation process, the preprocessing steps of the original dataset also include performing data cleaning to eliminate noise and redundant information. First, duplicate text is identified and deleted using a text fingerprint algorithm, such as repeated reports of the same technological achievement on multiple platforms. Then, text unrelated to the target technology is filtered out to remove invalid information. Finally, texts of different formats are converted into a unified plain text format, while standardizing the units of technical indicators and date formats.
[0025] Step S300: Input each piece of data in the updated dataset into a preset data extraction model. The data extraction model extracts the triplet data composed of technical direction, technical dynamics and technical indicators from each piece of data in the updated dataset. The data extraction model is a large language model. like Figure 4 As shown, in the specific implementation process, both the data extraction model and the large language model can adopt existing commercial or open-source large models. Specifically, they can be open-source generative large models, such as Qwen 3.0 or Deepseek. Specifically, the problem to be processed can be constructed as text, and the text can be input into the generative large model to obtain the output of the generative large model.
[0026] In the specific implementation process, this plan clarifies the extraction dimensions of the three parties' core information units: first, technical direction: the core field of the target technology; second, technical dynamics: a description of the current progress of the technology; and third, technical indicators: quantifiable technical performance parameters. The extraction method is based on the zero-shot information extraction capability of large models. The input prompt is "Extract the technology triples from the following text: Technology Direction T, Technology Dynamics D, Technology Indicator I, in the format..." If no corresponding information is found, fill in "None" to output structured triplet data in batches; In some embodiments of the present invention, the extracted triples are manually sampled for verification, with a sampling ratio of ≥10%. If the accuracy is <85%, a small number of samples are used to fine-tune the results in supplementary domain examples until the accuracy is ≥90%.
[0027] Step S400: Based on the preset transition keyword library and the updated dataset, determine the transition level of the technology to be judged; The existing technology transition keyword recognition is limited by static technology roadmaps and cannot capture emerging technology terms in real time. The element classification systems and judgment standards of different fields are not universal. Models need to be redeveloped when ported, which is costly and time-consuming. It is difficult to adapt to the evaluation needs of multiple fields or emerging technologies. Third, the maturity judgment model relies too much on domain-customized development, which has defects such as high algorithm porting costs and weak cross-scenario adaptability. For example, the element classification system of materials and digital technology is difficult to be universal.
[0028] Step S500: The technical information including the triplet data and transition level is added to the preset structured JSON prompt template. The completed structured JSON prompt template is input into the preset large language model, and the large language model outputs the maturity judgment level.
[0029] By adopting the above approach, this solution overcomes the bottleneck of reliance on manual methods. Through automated collection and preprocessing of multi-source data, combined with the zero-sample information extraction capability of large models, it replaces the traditional manual judgment mode, significantly improving the efficiency of processing massive amounts of data. The introduction of a data quality scoring model and a manual sampling verification mechanism ensures data reliability from the source, reduces errors caused by subjective bias, and solves the problems of low efficiency and error-proneness of traditional methods.
[0030] like Figure 2 As shown, in some embodiments of the present invention, the method further includes step S600, inputting reflection prompts into the large language model and receiving the maturity level assessment output by the large language model again.
[0031] In some embodiments of the present invention, reflection prompts are injected for result verification, and reflection instructions are input into the large model. These reflection instructions include: 1. Is any key dynamic information from the Content module in the JSON, such as experimental test data, missing? 2. Do the core CTE technical indicators fully meet all the criteria for this level? 3. Is there any calculation bias in the keyword matching degree of the transition? If there are any issues, please readjust the level according to the TRL criteria - the correspondence table of transition keywords; Furthermore, establish rules for adjusting the level range. If the level after reflection is consistent with the initial level of the previous stage, and If the score is ≥0.9, it is determined to be a single level. If the score after reflection differs from the initial score of the previous stage by 1 level, and 0.6 ≤ If the score is <0.9, it is determined to be a grade range. If the grade after reflection differs from the initial grade of the previous stage by ≥2 levels, or If the result is less than 0.6, return to the previous step and re-execute the initial reasoning until one of the two conditions mentioned above is met.
[0032] Output the final grade range report. The report should include the final grade range, the reason for the judgment (i.e., the core CTE satisfaction status + transition keyword matching degree + dynamic information support, and technical shortcomings analysis).
[0033] In the specific implementation process, the core data of the final report and the preceding stages are visualized to form an interactive decision-making tool. Based on the requirements of the interactive visualization engine, the Streamlit low-code framework of Python is adopted, and data processing Pandas, chart generation Plotly, large model interaction library OpenAI, and function development module LangChain are integrated to ensure the system is lightweight. The core visualization modules are designed as follows: First, the data tracing module displays the distribution of high-quality datasets to be analyzed, such as a pie chart showing: technical papers 35%, think tank reports 25%, and news updates 40%. Next, the core information module presents the technology triples and core CTE list in a structured manner, using a bar chart to show the CTE weight distribution. Then, the level evolution module uses a timeline to show the maturity changes of technology from historical stages to the current stage. Finally, the results report module clearly outputs the final level range, the reasons for the judgment, and the analysis of technical shortcomings in text, and uses a Gantt chart to compare the maturity differences between the target technology and other technologies in the same field. The system performance is optimized and output by using the Streamlit caching mechanism (session module) to cache the triple extraction and CTE weight calculation results, ensuring that the system response time is less than 3 seconds. Finally, the results are output as a visualization through a web interface, supporting user interaction.
[0034] In some embodiments of the present invention, the step of filtering each piece of data and determining whether to delete the data includes calculating a data quality score for each piece of data using the following formula, and determining whether to delete the data based on the data quality score; in, This represents the quality score of a single data point. For data relevance, For data integrity, For data timeliness, α, β, and γ are all preset weighting coefficients.
[0035] In its implementation, this solution quantitatively assesses data quality and filters high-value data, introducing a data quality scoring model to ensure that data entering subsequent stages meets the requirements of "relevance, completeness, and timeliness," based on the data quality scoring formula: R, with a value of 0-1, is used to determine the correlation between the text and the target technology based on large-scale semantic clustering technology. The higher the correlation, the closer R is to 1. C represents data completeness, with a value of 0-1, to assess whether the text contains key information for maturity assessment. The more complete the information, the closer C is to 1. T represents data timeliness, with a value of 0-1, calculated inversely based on the interval between the data release time and the current analysis time. The shorter the interval, the closer T is to 1. α, β, and γ are weight coefficients, each with a value of 1 / 3. They can be adjusted according to the specific technical field requirements. For example, when assessing emerging technologies, the weight of γ can be increased. Finally, only data with Q≥60 are retained to proceed to the next stage.
[0036] In some embodiments of the present invention, each transition level is provided with a transition keyword lexicon. In the step of determining the transition level of the technology to be determined by matching the updated dataset with the preset transition keyword lexicon, the updated dataset is matched with the transition keyword lexicon corresponding to each transition level, and the matching degree between the updated dataset and the transition keyword lexicon corresponding to each transition level is calculated. The transition level corresponding to the updated dataset is determined based on the matching degree.
[0037] In some embodiments of the present invention, in the step of calculating the matching degree between the updated dataset and the transition keyword lexicon corresponding to each transition level, the matching degree is calculated using the following formula: ; in, This represents the keyword matching degree of the k-th TRL level in the keyword library corresponding to the currently calculated transition level. The updated dataset contains the number of matching transition keywords in the transition keyword lexicon corresponding to the currently calculated transition level. This represents the total number of transition keywords in the transition keyword library corresponding to the currently calculated transition level.
[0038] In the specific implementation process, words related to level transitions are extracted, and the specific Technology Maturity Level (TRL) criteria are structured according to the 1-9 level definition. They are classified into theoretical stage TRL1-2 → laboratory stage TRL3-4 → prototype stage TRL5-6 → experimental verification stage TRL7-8 → deployment stage TRL9. The core judgment details of each level are sorted out. For example, TRL4 represents the completion of component-level verification in a laboratory environment and the achievement of key performance indicators. Then, based on the fact that transition keywords are the core basis for level advancement, transition keyword matching is performed. A signature keyword is matched for each TRL level advancement, and the keyword matching degree formula is calculated: , A value ≥0.6 is considered a keyword condition that satisfies the transition level.
[0039] In some embodiments of the present invention, the method further includes calculating the weight of each key technical element in the technology to be determined based on the technical field in which the technology to be determined is located, and supplementing the calculated weight of each key technical element as technical information into a preset structured JSON prompt word template.
[0040] In some embodiments of the present invention, in the step of calculating the weight of each key technical element in the technology to be determined based on the technical field in which the technology to be determined is located, the weight of each key technical element is calculated using the following formula: ; in, Let i be the weight of the i-th key technical element. The importance score of the i-th key technical element relative to the j-th key technical element is given, where n is the total number of key technical elements.
[0041] In the specific implementation process, the key technology elements (CTEs) are selected according to the dual criteria of importance and novelty. Domain adaptation is achieved through large-scale model context learning: input technology field examples, such as CTE examples in the new energy field: electrode materials, electrolyte conductivity, and CTE examples in the AI field: algorithm model, computing power support. Then, the large model is guided to select elements that have a significant impact on the progress of technology research and development and may cause technical risks as CTEs from the technology dynamics (D) and technology indicators (I) of the technology triplet. The weights of key technology elements (CTEs) are calculated, and the analytic hierarchy process (AHP) is introduced to quantify the importance of CTEs. The score ranges from 1 to 9, with 1 indicating equal importance and 9 indicating extreme importance. n represents the total number of CTEs. The top 3 CTEs with the highest weights are considered as core CTEs. The large model uses the triplet data of all key CTEs to determine the maturity level.
[0042] like Figure 3 As shown, in some embodiments of the present invention, the step of inputting the completed structured JSON prompt template into a preset large language model further includes: based on the target field of the obtained structured prompt, using a data extraction model to extract the target data corresponding to the target field from the updated dataset, completing the completion of the structured JSON prompt template, calculating the data integrity of the completed structured JSON prompt template, and determining whether to input the completed structured JSON prompt template into the preset large language model based on the data integrity.
[0043] In the specific implementation process, the steps for constructing a structured JSON prompt word template with dynamic slots to guide the large model to perform maturity assessment according to unified logic are as follows: The template is designed using JSON format and includes a task instruction module, a context criteria module, and a dynamic information slot module. The task instruction module, based on the TRL guidelines, determines the maturity level range of the target technology using the following information. The output must include the level range, the reasoning for the determination, and supporting evidence. The context criteria module requires filling in the TRL level determination details and transition keywords. The dynamic information slot module reserves spaces for filling in core CTEs, technology triples, and data source metadata. The name is the JSON template. For different fields such as materials and digital technologies, only the TRL details and transition keywords in the "context criteria module" need to be replaced; the overall template structure does not need to be modified, reducing cross-domain adaptation costs.
[0044] In the specific implementation process, during the step of calculating the data integrity of the completed structured JSON prompt template, the scattered technical information is integrated into a standardized JSON format. The specific steps for connecting the data and prompts are as follows: A JSON data integrity scoring formula is introduced to ensure that no critical information is missing. Where C is the integrity score, with a value of 0-1. The number of fields that are not None. The total number of fields in the JSON must be C≥0.7. Otherwise, technical triples should be extracted again or supplementary data should be collected until the requirement is met. Then, all the high-quality data after filtering are batch converted according to the above JSON structure to form a structured technical information dataset, which is stored as a JSON file for later use.
[0045] In some embodiments of the present invention, in the step of inputting the completed structured JSON prompt template into a preset large language model, and the large language model outputting the maturity judgment level, the large language model outputs the preliminary judgment level, the number of judgment rules that meet the preliminary judgment level, and the total number of judgment rules for the preliminary judgment level. The determination confidence level is calculated based on the number of determination criteria that meet the preliminary determination level and the total number of determination criteria for the preliminary determination level. Based on the determination confidence level, it is determined whether to use the preliminary determination level as the final maturity determination level.
[0046] In the specific implementation process, relying on the logical reasoning capabilities of the large model, the specific steps to complete the preliminary reasoning of maturity level are as follows: Structured technical information is filled into the dynamic slots of the prompt word template to form the complete input text of the large model. An example of the input format is: [Prompt word template content] + [Structured technical information JSON]; A thought chain is added to the prompt words to guide the process: A preliminary judgment of the possible TRL level is made, and the judgment criteria for that level are compared one by one to check whether the theoretical / prototype / experimental / deployment dynamics of the Content module in the JSON are met. The matching degree of the transition keywords for that level is calculated. ,like If the threshold is ≥0.6 and all detailed criteria are met, then attempt to match a higher level; if not, backtrack to a lower level. The large model outputs a preliminary maturity level and supporting evidence, and simultaneously, according to the formula... To calculate the inference confidence, where This represents the confidence level of the inference, with a value between 0 and 1. The number of TRL criteria to be satisfied, The total number of detailed criteria for this preliminary level, if If the value is ≥0.8, the preliminary results proceed to the next stage. If the value is less than 0.8, return to the previous step, supplement the structured information or optimize the prompt word template, and reason again.
[0047] Using the above approach, the input text is formed by combining structured technical information with prompt word templates. The large model is guided by a thought chain to follow the logic of "preliminary judgment → item-by-item verification → level adjustment". Based on the core CTE, TRL discrimination rules and the matching degree of transition keywords, the technology maturity level is initially judged. Then, the reliability of the results is quantified by the reasoning confidence formula, and the preliminary level results that meet the confidence level are selected, providing a basis that meets the requirements for subsequent reflection, verification and final level determination.
[0048] In practice, the template can be as shown in Table 1 below: Table 1 This invention also provides a cutting-edge technology maturity level interval discrimination analysis device. The device includes a computer device, which includes a processor and a memory. The memory stores computer instructions, and the processor is used to execute the computer instructions stored in the memory. When the computer instructions are executed by the processor, the device implements the steps of the method described above.
[0049] This invention also provides a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the steps of the aforementioned method for discriminant analysis of the maturity level range of cutting-edge technologies. The computer-readable storage medium can be a tangible storage medium, such as random access memory (RAM), main memory, read-only memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, floppy disks, hard disks, removable storage disks, CD-ROMs, or any other form of storage medium known in the art.
[0050] Those skilled in the art will understand that the exemplary components, systems, and methods described in conjunction with the embodiments disclosed herein can be implemented in hardware, software, or a combination of both. Whether implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this invention. When implemented in hardware, it can be, for example, electronic circuits, application-specific integrated circuits (ASICs), appropriate firmware, plug-ins, function cards, etc. When implemented in software, the elements of this invention are programs or code segments used to perform the desired tasks. The programs or code segments can be stored in a machine-readable medium or transmitted over a transmission medium or communication link via data signals carried in a carrier wave.
[0051] It should be clarified that the present invention is not limited to the specific configurations and processes described above and shown in the figures. For the sake of brevity, detailed descriptions of known methods are omitted here. In the above embodiments, several specific steps are described and shown as examples. However, the method process of the present invention is not limited to the specific steps described and shown. Those skilled in the art can make various changes, modifications, and additions, or change the order of steps, after understanding the spirit of the present invention.
[0052] In this invention, features described and / or illustrated for one embodiment may be used in the same or similar manner in one or more other embodiments, and / or combined with or in place of features of other embodiments.
[0053] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, various modifications and variations of the embodiments of the present invention are possible. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.
Claims
1. A method of frontier technology maturity level interval discrimination analysis, characterized by, The steps of the method comprise: Obtaining technical text pre-collected by a to-be-judged technology, and constructing as an original data set; Preprocessing the original data set to obtain an updated data set, and in the preprocessing step, screening each piece of data to determine whether to delete the data; Each piece of data in the updated data set is input into a preset data extraction model, the data extraction model extracts triple data composed of technical direction, technical dynamics and technical indicators from each piece of data in the updated data set, and the data extraction model is a large language model; Based on the matching between the preset transition keyword library and the updated data set, the transition level of the to-be-judged technology is determined; The technical information including the triple data and the transition level is supplemented to the preset structured JSON prompt word template, and the completed structured JSON prompt word template is input into the preset large language model, and the large language model outputs the maturity judgment level.
2. The frontier technology maturity level interval discriminant analysis method of claim 1, wherein, The steps of the method further comprise inputting a reflection prompt word into the large language model, and receiving the maturity judgment level output by the large language model again.
3. The frontier technology maturity level interval discrimination analysis method according to claim 1 or 2, characterized in that, The step of screening each piece of data to determine whether to delete the data comprises calculating the data quality score of each piece of data using the following formula, and determining whether to delete the data based on the data quality score; wherein, represents a quality score of a single piece of data, is a data relevance, is a data integrity, is a data timeliness, and α, β and γ are all preset weight coefficients.
4. The frontier technology maturity level interval discriminant analysis method of claim 1 or 2, wherein, Each of the transition levels is provided with a transition keyword library, and in the step of determining the transition level of the to-be-judged technology based on the matching between the preset transition keyword library and the updated data set, the updated data set is matched with the transition keyword library corresponding to each transition level, and the matching degree of the updated data set and the transition keyword library corresponding to each transition level is calculated, and the transition level corresponding to the updated data set is determined based on the matching degree.
5. The frontier technology maturity level interval discriminant analysis method of claim 4, wherein, In the step of calculating the matching degree of the updated data set and the transition keyword library corresponding to each transition level, the matching degree is calculated using the following formula: ； wherein, is the matching degree of the keyword in the kth TRL level in the transition keyword library corresponding to the transition level currently calculated, is the number of matched transition keywords in the transition keyword library corresponding to the transition level currently calculated for the updated data set, is the total number of level transition keywords in the transition keyword library corresponding to the transition level currently calculated.
6. The frontier technology maturity level interval discriminant analysis method of claim 1, wherein, The steps of the method further comprise calculating the weight of each key technical element in the to-be-judged technology based on the technical field in which the to-be-judged technology is located, and supplementing the calculated weight of each key technical element as technical information to the preset structured JSON prompt word template.
7. The frontier technology maturity level interval discriminant analysis method of claim 6, wherein, In the step of calculating the weight of each key technical element in the to-be-judged technology based on the technical field in which the to-be-judged technology is located, the weight of each key technical element is calculated using the following formula: ； wherein, is the weight of the i-th key technology element, is the importance score of the i-th key technology element relative to the j-th key technology element, and n is the total number of key technology elements.
8. The frontier technology maturity level interval discriminant analysis method of claim 1, wherein, The step of inputting the completed structured JSON prompt word template into the preset large language model further comprises extracting target data corresponding to a target field from the updated data set based on the target field of the obtained structured prompt word using the data extraction model, completing the filling of the structured JSON prompt word template, calculating the data completeness of the completed structured JSON prompt word template, and determining whether to input the completed structured JSON prompt word template into the preset large language model based on the data completeness.
9. The frontier technology maturity level interval discriminant analysis method of claim 1, wherein, In the step of inputting the completed structured JSON prompt word template into a preset large language model, the large language model outputs a maturity judgment level, the large language model outputs a preliminary judgment level, a number of judgment details meeting the preliminary judgment level, and a total number of judgment details of the preliminary judgment level; A judgment confidence is calculated based on the number of judgment details meeting the preliminary judgment level and the total number of judgment details of the preliminary judgment level, and whether the preliminary judgment level is determined as the final maturity judgment level based on the judgment confidence.
10. A leading-edge technology maturity level interval discriminant analysis device characterized by comprising: The device comprises a computer device comprising a processor and a memory, the memory storing computer instructions, and the processor is configured to execute the computer instructions stored in the memory, and the device implements the steps of the method as claimed in any one of claims 1-9 when the computer instructions are executed by the processor.