An artificial intelligence-based professional technology maturity assessment method

Abstract

The present invention discloses an artificial intelligence-based professional technology maturity assessment method, which includes the following steps: building a dynamic knowledge base by crawling technical literature, patents and industry reports, performing cross-technical field semantic embedding in combination with input technical text, and generating semantic representation; extracting multi-scale features based on the dynamic knowledge base and the semantic representation, and generating a comprehensive feature set; building a technology dependency graph using the comprehensive feature set and the dynamic knowledge base, and planning reasoning paths through reinforcement learning to assess the innovation, feasibility and commercial value of professional technologies and generate assessment scores; calculating comprehensive scores using cross-technical field weights with adaptive threshold adjustment, and mapping them to maturity levels; and optimizing assessment results by integrating user feedback through online learning. The method of the present invention overcomes the problems of lag in static knowledge bases, ambiguity in cross-technical fields and insufficient reasoning paths, and improves the real-time, precision and professionalism of the assessment.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence and natural language processing technology, and in particular to a professional technology maturity assessment method based on artificial intelligence. Background Art

[0002] In recent years, the rapid development of artificial intelligence (AI) has driven innovation in technology maturity assessment methods. Traditional technology maturity assessments, such as the Technology Readiness Level (TRL) framework, are widely used in sectors such as aerospace, energy, and defense. These primarily rely on experts to assess the maturity of a technology from concept to commercialization. However, with advances in big data and natural language processing, automated AI-based assessment methods are gaining popularity. These methods primarily utilize pre-trained language models to analyze literature, patents, and technical reports, extracting key technical features and integrating them with knowledge bases for reasoning, thereby achieving quantitative technology evaluation. These methods significantly outperform traditional manual methods in terms of efficiency and coverage, and are widely used in areas such as technology competition, R&D management, and patent examination.

[0003] However, current automated AI evaluation methods still have several shortcomings. First, existing methods often rely on static knowledge bases, making it difficult to integrate the latest technical literature and patent data, resulting in evaluation results lagging behind the technological frontier. Second, in the evaluation of cross-disciplinary technologies, existing methods inadequately handle the semantic ambiguity of multidisciplinary terminology, making it difficult to accurately capture the characteristics of emerging technologies, thus affecting the accuracy and universality of the evaluation. Third, the reasoning logic of existing methods lacks dynamic analysis of technology evolution paths, making it difficult to fully reflect the differences in the maturity of technical solutions. These shortcomings directly limit the professionalism and reference value of the evaluation results.

[0004] CN115600893A discloses an intelligent evaluation method and system for navigation safety management at a large shipping hub. The solution primarily constructs an evaluation index system for the current state of navigation safety management. By introducing a safety management maturity model, the established evaluation indicators are weighted based on membership analysis and reliability and validity testing. The correlation between each indicator and the current state of navigation hub safety management is weighed. An expert evaluation module is constructed, including self-evaluation, mutual evaluation, and professional evaluation. The optimal weighting of these three evaluations is calculated. The determined evaluation index system, indicator weights, and evaluation indicator levels are then quantitatively calculated. Finally, visualization tools such as radar charts and bar graphs are used to display the evaluation results, and recommendations and directions for improvement are provided based on the evaluation results. While this solution mentions the grading of maturity evaluation, it does not address its relationship with the dynamic knowledge base, failing to address the issue of evaluation results lagging behind the technological frontier.

[0005] CN117993375A discloses a method and system for cross-domain application of patented technologies. This solution is primarily based on patent documents and employs text mining and deep learning methods to extract technical elements and generate technical feature summaries. Statistical analysis and pattern recognition methods are then used to assess domain adaptability. Three-dimensional modeling and computational fluid dynamics methods are used to construct virtual prototypes. Simulation software and iterative design methods are then used to optimize technical solutions. Cross-domain application analysis is then conducted through knowledge graphs and association rule mining. While this solution provides in-depth analysis of market demand and technical compatibility through text analysis and simulation software, it lacks dynamic analysis of technology evolution paths, making it difficult to fully reflect differences in the maturity of technical solutions. Summary of the Invention

[0006] The purpose of this section is to summarize some aspects of the embodiments of the present invention and briefly introduce some preferred embodiments. Some simplifications or omissions may be made in this section and the abstract and title of this application to avoid obscuring the purpose of this section, the abstract and the title of the invention, and such simplifications or omissions should not be used to limit the scope of the present invention.

[0007] In view of the above existing problems, the present invention is proposed. Therefore, the present invention provides an artificial intelligence-based professional technology maturity assessment method to solve the problems raised in the background technology.

[0008] To solve the above technical problems, the present invention provides the following technical solution: a professional technology maturity assessment method based on artificial intelligence, comprising:

[0009] Acquire technical literature, patents, and industry reports to build a dynamic knowledge base, and then embed cross-technical domain semantics into the dynamic knowledge base by inputting technical text to generate semantic representations.

[0010] extracting multi-scale features from input technical text based on the dynamic knowledge base and semantic representation to generate a comprehensive feature set;

[0011] Utilizing the comprehensive feature set and the dynamic knowledge base to construct a technology dependency graph, generating an evaluation score for the professional technology, calculating a comprehensive score based on the evaluation score using cross-technology-area weights adjusted with adaptive thresholds, mapping the comprehensive score to a professional technology maturity level, and generating a professional maturity evaluation result;

[0012] By incorporating user feedback through online learning, the assessment results are optimized and the professionalism of the professional maturity assessment results is improved.

[0013] As a preferred solution of the professional technology maturity assessment method based on artificial intelligence described in the present invention, the process of acquiring technical literature, patents, and industry reports to build a dynamic knowledge base includes:

[0014] Use crawler tools to obtain technical literature, patents, and industry reports from open databases to build a dynamic knowledge base;

[0015] By extracting the keywords of the input technical text and each crawled data and applying time weighting, we can filter out the technical content associated with the keywords of the input technical text.

[0016] The screened technical content is merged with the knowledge base to update the knowledge base.

[0017] As a preferred solution of the professional technology maturity assessment method based on artificial intelligence of the present invention, wherein: semantic embedding across technical fields by inputting text content and combining it with the dynamic knowledge base to generate semantic representation includes:

[0018] The Transformer model of multi-task learning is used to generate semantic representations that are adapted across technical fields.

[0019] As a preferred embodiment of the professional technology maturity assessment method based on artificial intelligence of the present invention, wherein: according to the dynamic knowledge base and semantic representation, multi-scale features are extracted from the input technical text to generate a comprehensive feature set, including:

[0020] Perform sentence segmentation, word segmentation and dependency syntax analysis on the input technical text to generate syntactic structure representation;

[0021] According to the syntactic structure representation, global technical goals and sub-technical goals are extracted respectively, and a feature vector representing the overall function of the technology and a modular sub-technical feature set are generated;

[0022] The feature vectors of the overall functionality of the technology and the feature sets of the sub-technologies are integrated to generate a comprehensive feature set.

[0023] As a preferred embodiment of the professional technology maturity assessment method based on artificial intelligence of the present invention, wherein: using the comprehensive feature set and the dynamic knowledge base to construct a technology dependency graph and generate an assessment score for the professional technology, the following steps are included:

[0024] The nodes in the technology dependency graph are represented as sub-technologies, and the edges are represented as the dependency relationships between sub-technologies.

[0025] By planning the reasoning path of the technology dependency graph through reinforcement learning, the innovation, feasibility and commercial value of professional technology are evaluated to generate an evaluation score for the professional technology.

[0026] As a preferred solution of the professional technology maturity assessment method based on artificial intelligence described in the present invention, the innovation, feasibility and commercial value of professional technology are assessed, including:

[0027] Evaluate the innovativeness of specialized technologies by comparing the semantic similarity between sub-technologies and technical content in the dynamic knowledge base;

[0028] Evaluate the feasibility of specialized technologies by matching currently available resources in the dynamic knowledge base with the required resources in the sub-technology feature set and considering the adaptability of the sub-technology environment;

[0029] Evaluate the commercial value of professional technology by estimating the potential market size and the degree of matching between application scenarios.

[0030] As a preferred embodiment of the professional technology maturity assessment method based on artificial intelligence of the present invention, a comprehensive score is calculated based on the assessment score using cross-technical field weights with adaptive threshold adjustment, and the comprehensive score is mapped to a professional technology maturity level to generate a professional maturity assessment result, including:

[0031] Based on the evaluation scores, weights across technical fields are determined through reinforcement learning to calculate a comprehensive score;

[0032] Adaptive threshold adjustments for weights across technology areas using tiered accuracy based on historical technology assessment data;

[0033] The comprehensive score after threshold adjustment is mapped to the professional and technical maturity level to obtain the professional maturity assessment result.

[0034] As a preferred solution of the professional technology maturity assessment method based on artificial intelligence described in the present invention, the professional technology maturity levels include the start-up period, bubble period, trough period, climbing period and mature period.

[0035] As a preferred embodiment of the professional technology maturity assessment method based on artificial intelligence of the present invention, wherein: optimizing the assessment results by incorporating user feedback through online learning includes:

[0036] Online learning adjusts the weights across technical fields based on the differences between user feedback and professional maturity assessment results.

[0037] Compared with the prior art, the invention has the following beneficial effects:

[0038] 1. By crawling technical literature, patents, and industry reports to build a dynamic knowledge base, and combining time-weighted updates to update the knowledge base content, this method can promptly reflect the latest technological trends, overcome the shortcomings of static knowledge bases that lead to delayed evaluation results, and ensure the real-time nature of the evaluation results;

[0039] 2. The Transformer model with multi-task learning is used for cross-domain semantic embedding. Combined with multi-scale feature extraction, it generates global feature vectors and sub-technology feature sets. This effectively solves the problems of interdisciplinary technical terminology ambiguity and insufficient feature capture, providing reliable support for emerging technology assessment.

[0040] 3. By reinforcing the reasoning path of the learning planning technology dependency graph, dynamic evaluation of innovation, feasibility, and commercial value is achieved. By combining online learning models with user feedback optimization, not only does it improve the professionalism and reasoning depth of the evaluation results, but it can also dynamically adjust the weights based on user feedback, enhancing the reference value of the evaluation results and user experience. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] To more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for describing the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. Those skilled in the art can also derive other drawings based on these drawings without inventive effort. Among them:

[0042] Figure 1 This is an overall flow chart of an artificial intelligence-based professional technology maturity assessment method according to an embodiment of the present invention. DETAILED DESCRIPTION

[0043] To make the above-mentioned objects, features, and advantages of the present invention more clearly understood, the following detailed description of the specific embodiments of the present invention is given in conjunction with the accompanying drawings. It is obvious that the described embodiments are only part of the embodiments of the present invention, but not all of them. Based on the embodiments of the present invention, all other embodiments obtained by ordinary persons in this field without creative work should fall within the scope of protection of the present invention.

[0044] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways different from those described herein. Those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0045] Secondly, the term "one embodiment" or "embodiment" herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in various places throughout this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive of other embodiments.

[0046] The present invention is described in detail with reference to schematic diagrams. For ease of illustration, cross-sectional views of device structures may be partially enlarged and not to scale when describing embodiments of the present invention. Furthermore, the schematic diagrams are merely illustrative and should not limit the scope of the present invention. Furthermore, in actual production, the three-dimensional dimensions of length, width, and depth should be included.

[0047] In the description of the present invention, it should be noted that the terms "upper, lower, inner, and outer" and other references to orientations or positional relationships are based on the orientations or positional relationships shown in the accompanying drawings and are intended solely to facilitate and simplify the description of the present invention. They are not intended to indicate or imply that the devices or components referred to must have, be constructed, or operate in a specific orientation, and therefore should not be construed as limitations on the present invention. Furthermore, the terms "first, second, or third" are used for descriptive purposes only and should not be construed as indicating or implying relative importance.

[0048] In this disclosure, unless otherwise specified or limited, the terms "mounted," "connected," and "connected" should be interpreted broadly. For example, they may refer to fixed, removable, or integral connections. They may also refer to mechanical, electrical, or direct connections, indirect connections through an intermediary, or internal communication between two components. Those skilled in the art will understand the specific meanings of these terms in this disclosure.

[0049] Example 1

[0050] Reference Figure 1 , which is the first embodiment of the present invention, provides a professional technology maturity assessment method based on artificial intelligence, including:

[0051] S1. Acquire technical literature, patents, and industry reports to build a dynamic knowledge base. By inputting technical text and combining it with the dynamic knowledge base, perform semantic embedding across technical fields and generate semantic representations.

[0052] Furthermore, crawler tools are used to obtain technical literature, patents, and industry reports from open databases to build a dynamic knowledge base;

[0053] Specifically, we use the Scrapy framework combined with Selenium to crawl open (free) databases (such as the Chinese Academy of Sciences Open Data Platform, CNIPA, and the National Bureau of Statistics of China);

[0054] It should be explained that Selenium is a third-party library for Python that operates the browser through the interface provided to the outside world, allowing the browser to complete automated operations;

[0055] Specifically, before crawling data, you need to set the time range for crawling data; give priority to crawling data within the past month (based on the current date), and gradually expand the months to ensure the timeliness of the crawled data; for example, taking the current date of April 27, 2025 as the benchmark, crawl the data within the past month of that date, that is, March 27, 2025 to April 27, 2025. Then, as the number of months iterates, assuming that the current date is May 27, 2025, then the data within the past month of that date will become April 27, 2025 to May 27, 2025;

[0056] It should be noted that in actual applications, due to the speed of data updates and the huge amount of crawled data, crawling data gradually on a monthly basis can not only ensure the timeliness of the data, but also ensure that the data has sufficient time to be stored;

[0057] Furthermore, we used a natural language processing tool (TF-IDF) to simultaneously extract keywords from both the input technical text and the crawled data. We then assigned a time weight to each crawled data item (recent data had a higher weight, while older data had a lower weight). Using this time weighting, we screened out technical content associated with the keywords in the input technical text.

[0058] Specifically, based on the difference between the release date of the crawled data and the current date, an exponential decay function is used to assign a time weight to each crawled data:

[0059]

[0060] in, Indicates crawling data Time weight; Indicates crawling data The difference between the release time and the current date (unit: day);

[0061] Specifically, the keywords of the input technical text and crawled data are converted into semantic representations, and their relevance is calculated by cosine similarity:

[0062]

[0063] in, Indicates the correlation between technical text and crawled data keywords, K represents the set of input technical text keywords, Represented as a collection of crawled data keywords; It is represented as the semantic vector of the technical text keyword k; Represents the crawled data keywords Semantic vector of Expressed as the cosine similarity of two keyword vectors; Represented as the size of the technical text and crawled data keyword sets respectively;

[0064] Specifically, the comprehensive score of each crawled data is calculated by combining the relevance of technical text and crawled data keywords with the time weight of crawled data. :

[0065]

[0066] It should be noted that by sorting the comprehensive score of each crawled data in descending order, it is possible to filter out crawled data with similar days and that match the technical text keywords, thereby obtaining technical content associated with the input technical text keywords;

[0067] Furthermore, the screened technical content is merged with the knowledge base to update the knowledge base;

[0068] It should be noted that the merge and update operations are intended to optimize the query performance of the dynamic knowledge base and remove duplicates (eliminate redundant or low-relevant content). For example, if the knowledge base already contains a technical content on "deep learning", the newly crawled similar technical content will update the metadata of the existing record (such as publication date, number of citations) instead of adding it again.

[0069] Furthermore, a multi-task learning Transformer model is used to generate semantic representations that are adaptable across technical fields;

[0070] Specifically, we use a pre-trained Transformer model and set the following model tasks: Task 1: Identify technical terms; Task 2: Determine the technical field to which the input text belongs; Task 3: Generate semantic embeddings and optimize term representation;

[0071] Specifically, the Transformer model first receives the input technical text through the input layer, uses a word segmenter (such as BERT's WordPiece) to split the text into sub-word units, and defines special token tags ([CLS] and [SEP], where [CLS] is used to aggregate the semantic information of the entire input technical text, and [SEP] is used to mark the end of the input technical text or separate different text fragments) to convert it into a token sequence (consisting of multiple tokens); then each token is mapped to an ID (Token ID) in the vocabulary, and a positional encoding is added to each token to capture word order information; the Token ID in the input layer is converted into an embedding vector, and is input into the Transformer encoder to capture the contextual relationship between tokens and generate context-related hidden representations; the input layer simultaneously performs terminology recognition and domain classification based on the hidden representation generated by the Transformer encoder;

[0072] Specifically, term recognition is represented by conditional random field (CRF) as:

[0073]

[0074] Where x represents the input technical text and y represents the output target label sequence; Represents the parameters of the Transformer encoder; W and U represent the conversion matrix and emission matrix respectively; and Respectively represent the i-th Token label and the i-1-th Token label in the label sequence; Represents the context-dependent hidden representation generated by the Transformer encoder; Y is an enumeration; n is the total number of Token tags; represents all possible target label sequences;

[0075] It should be noted that the conditional random field is Provide the token range of the term to ensure that the semantic vector is calculated only for the term token;

[0076] Specifically, the domain classification is expressed through the fully connected layer and Softmax in the output layer as follows:

[0077]

[0078] in, Expressed as a probability distribution of technology domain categories, Represents the hidden representation of [CLS]Token; C represents the total number of technology fields; Represented as a technical field index, used to traverse all technical fields; Represented as the bias term of the fully connected layer; Represented as the weight matrix of the fully connected layer, Mapping to the technical domain;

[0079] It should be noted that the fully connected layer and Softmax in the output layer can optimize The quality of Adapt to specific technical fields;

[0080] Specifically, the Transformer model for multi-task learning generates the following semantic representation:

[0081] in, It is represented as the semantic vector of the term W, and Avg represents the average pooling operation, that is, taking the average of the hidden representations of multiple tokens of the term; Represented as the Token index corresponding to the term;

[0082] It should be noted that the Conditional Random Field works in conjunction with the fully connected layer and Softmax in the output layer to generate a semantic representation that is adaptable across technical fields. This ensures that the semantic representation is adaptable to different technical fields and resolves the problem of terminology ambiguity across technical fields (for example, in computer networks, the processing pressure of a server or network can be called a load, such as network load, while in mechanical engineering, the weight or external force borne by a structure can also be called a load, such as constant torque load).

[0083] S2, extract multi-scale features from the input technical text based on the dynamic knowledge base and semantic representation to generate a comprehensive feature set;

[0084] Furthermore, the input technical text is subjected to sentence segmentation, word segmentation and dependency syntax analysis to generate a syntactic structure representation;

[0085] Specifically, by calling the spaCy library in Python, the input technical text is segmented into multiple independent sentences based on punctuation (such as periods and commas) and grammatical rules. For example, "Federated learning is a distributed algorithm that protects data security through differential privacy and is suitable for multi-party collaboration scenarios" is segmented into sentence 1: "Federated learning is a distributed algorithm that protects data security through differential privacy and is suitable for multi-party collaboration scenarios", sentence 2: "Protecting data security through differential privacy", and sentence 3: "Suitable for multi-party collaboration scenarios";

[0086] Specifically, the Jieba library in Python is called to split the sentence into words in combination with the technical field dictionary;

[0087] Specifically, by calling the DependencyParser of the spaCy library in Python, the syntactic structure is analyzed and a dependency tree is generated to reveal the syntactic dependencies between words (such as subject-verb-object and modification relationships);

[0088] Specifically, syntactic structure Expressed as:

[0089]

[0090] Among them, DepTree represents the dependency tree, and s represents the sentence;

[0091] Furthermore, based on the syntactic structure representation, global technical goals and sub-technical goals are extracted respectively to generate a feature vector representing the overall function of the technology and a modular sub-technical feature set;

[0092] Specifically, the main sentence (usually the first sentence or the sentence with the most dependencies) is selected from the syntactic structure representation, and the centrality score of the sentence is calculated based on the dependency weights between words:

[0093]

[0094] in, represents the dependency weight, represents the i-th sentence; Represented as the centrality score for each sentence;

[0095] It should be noted that the sentence with the highest sentence centrality score is the main sentence;

[0096] Specifically, by traversing the dependency tree, we find the root node (predicate or copula), subject, and object / complement; if the root node is a verb, we extract the verb + object; if the root node is a copula, we extract the subject + complement; and we get the global objective function Expressed as:

[0097] or

[0098] in, Indicates object / complement, Indicates the subject, Represents the root node;

[0099] Specifically, Substituting the Transformer model for multi-task learning, we get:

[0100]

[0101] in, A feature vector representing the overall functionality of the technology;

[0102] Specifically, by combining the dependency tree with the technical field dictionary, we check whether there is a subordinate clause or adverbial clause in the sentence, and calculate the technical relevance of the clause to the main clause. :

[0103]

[0104] Where Dictionary represents the technical field dictionary, F is an indicator function, which is 1 when both the clause and the main clause are in the technical field dictionary, otherwise it is 0; e represents a subordinate clause or adverbial clause;

[0105] Specifically, if , then call the Transformer model of multi-task learning and get:

[0106]

[0107] in, Represents a modular set of sub-technical features;

[0108] Furthermore, the feature vectors of the overall functionality of the technology and the feature sets of the sub-technologies are integrated to generate a comprehensive feature set;

[0109] It should be noted that by generating a comprehensive feature set, syntactic relations and related terms are taken into account to ensure that the extracted technical text features are both grammatically logical and technically relevant, thus solving the problem that traditional methods cannot capture the features of input technical texts sufficiently.

[0110] S3. Build a technology dependency graph using a comprehensive feature set and a dynamic knowledge base to generate a professional technology assessment score. Based on the assessment score, calculate a comprehensive score using cross-technology weights adjusted with adaptive thresholds. Map the comprehensive score to a professional technology maturity level to generate a professional maturity assessment result.

[0111] Furthermore, the comprehensive feature set is input and the dynamic knowledge base is called to construct a technology dependency graph, in which the nodes are represented as sub-technologies and the edges are represented as the dependency relationships of the sub-technologies.

[0112] It should be noted that a sub-technology is composed of a sub-technology feature set, and the dependency relationship of a sub-technology is the dependency relationship between each sub-technology in the sub-technology feature set;

[0113] Specifically, by traversing the technology dependency graph, a set of all reasoning paths (passing through each edge and each node to form a reasoning path) in the technology dependency graph is obtained;

[0114] Furthermore, through reinforcement learning, the reasoning path of the technology dependency graph is planned to evaluate the innovation, feasibility and commercial value of the professional technology, and generate an evaluation score for the professional technology;

[0115] Specifically, reinforcement learning uses a deep Q-network to plan the inference path of the technology dependency graph. The deep Q-network is set as a 3-layer fully connected network with a hidden layer dimension of 256. The learning rate is 1e-3. The exploration rate is initially 0.9 and linearly decays to 0.1. The current inference node, comprehensive feature set, and evaluation dimensions (innovation, feasibility, and commercial value) are used as the state of the deep Q-network. The evaluation dimensions (innovation, feasibility, and commercial value) for selecting the next inference path are used as the actions of the deep Q-network. The evaluation dimensions (innovation, feasibility, and commercial value) are used as rewards.

[0116] Specifically, the innovation of professional technology is evaluated by comparing the semantic similarity between sub-technology and technical content in the dynamic knowledge base.

[0117]

[0118] in, It represents the novelty of the sub-technology, which is obtained by comparing it with the dynamic knowledge base; Indicates technical similarity; Represented as a dynamic knowledge base; Represents a single sub-technical feature in a sub-technical feature set; Represents a single technology feature in a dynamic knowledge base (referring to the technology itself, not the sub-technology);

[0119] It should be noted that an Innov score of 7 or more indicates high innovation, 4 to 6 indicates improvements to existing technologies, and less than 4 indicates a combination of existing technologies;

[0120] Specifically, the feasibility of professional technology is evaluated by matching the currently available resources in the dynamic knowledge base with the required resources in the sub-technology feature set and considering the adaptability of the sub-technology environment:

[0121]

[0122] in, Available resources represented as technologies (e.g., open source algorithms) are matched from a dynamic database; The resources required by the technology (such as computing power) are obtained through the sub-technology feature set; Indicates the environmental adaptation coefficient (0-1), such as policy compliance;

[0123] It should be noted that Feas greater than or equal to 8 is considered highly feasible, 5-7 is feasible under restricted conditions (i.e., it can be implemented but certain conditions need to be met), and less than 5 is feasible under risk conditions (i.e., it may be implemented but there are certain risks);

[0124] Specifically, the commercial value of professional technology is evaluated by estimating the potential market size and the degree of matching between application scenarios.

[0125]

[0126] in, represents the potential market size, obtained from the dynamic knowledge base; Indicates the application scenario matching degree, obtained by scenario embedded calculation;

[0127] It should be noted that Comm greater than or equal to 7 indicates high commercial value, 4 to 6 indicates moderate commercial value, and less than 4 indicates low commercial value;

[0128] Furthermore, based on the evaluation scores, weights across technical fields are determined through reinforcement learning to calculate a comprehensive score;

[0129] Specifically, the comprehensive score Maturity is expressed as:

[0130]

[0131] in, 、 、 Represent the weights of Innov, Feas, and Comm, respectively. Their initial weights are 0.4, 0.3, and 0.3, respectively, and are determined by the rewards in reinforcement learning:

[0132]

[0133]

[0134] in, Expressed as an increase or decrease in action;

[0135] Furthermore, adaptive threshold adjustments are made to weights across technology areas using hierarchical accuracy based on historical technology assessment data;

[0136] Specifically, each time the weight is adjusted, it needs to be normalized:

[0137]

[0138] Specifically, based on the graded accuracy of historical technical evaluation data, two action changes are subtracted to ensure the stability of the reward R:

[0139]

[0140] Where H_Accuracy represents the graded accuracy of historical technical assessment data;

[0141] Furthermore, the comprehensive score after threshold adjustment is mapped to the professional and technical maturity level to obtain the professional maturity assessment result;

[0142] Specifically, the mapped professional and technical maturity levels are:

[0143] Start-up stage: 0~2; indicates that the technical concept is initially formed and lacks practical verification;

[0144] Bubble period: 2~4; indicates that the technology has attracted much attention, but its practicality has yet to be proven;

[0145] Low period: 4~6; indicates that the technology is not yet perfect and needs further research and development;

[0146] Climbing stage: 6~8; indicates that the technology has been initially verified and has commercial potential;

[0147] Maturity: 8-10; indicates that the technology is stable and can be applied on a large scale;

[0148] It should be noted that by adaptively adjusting the threshold and mapping the comprehensive score to the professional technology maturity level, it is not only possible to effectively identify the maturity of the technical solution, but also to determine whether the technical solution has reference value;

[0149] S4. Optimize assessment results by integrating user feedback through online learning to improve the professionalism of professional maturity assessment results;

[0150] Furthermore, through the online learning model, a loss function of user rating and model prediction mean square error is established to adjust the weights across technical fields;

[0151] Specifically, the loss function is expressed as:

[0152]

[0153] in, Indicates the comprehensive score obtained previously. represents the comprehensive score obtained through user feedback; M represents the number of user feedback; L represents the loss value;

[0154] Specifically, adjust the weights across technical fields:

[0155]

[0156] Specifically, when the loss value is 0, the cross-technical field weights are not adjusted; Represents an action, Expressed as the gradient of descent of the loss value; represents the adjusted new cross-technology field weights, Indicates the cross-technical field weights that currently need to be adjusted;

[0157] It should be noted that the dynamic adjustment of weights through user feedback optimization enhances the reference value of the evaluation results and user experience.

[0158] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present application may take the form of a fully hardware embodiment, a fully software embodiment, or an embodiment combining software and hardware. Furthermore, the present application may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code. The solutions in the embodiments of the present application may be implemented using various computer languages, such as the object-oriented programming language Java and the interpreted scripting language JavaScript.

[0159] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 a process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0160] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 a process or multiple processes and / or boxes Figure 1 The function specified in one or more boxes.

[0161] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing the instructions executed on the computer or other programmable device for implementing the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0162] Although the preferred embodiments of the present application have been described, those skilled in the art may make additional changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the present application.

[0163] Obviously, those skilled in the art may make various changes and modifications to this application without departing from the spirit and scope of this application. Thus, if these modifications and variations of this application fall within the scope of the claims of this application and their equivalents, this application is intended to include these modifications and variations.

Claims

1. A professional technology maturity assessment method based on artificial intelligence, characterized in that: include: Acquire technical literature, patents, and industry reports to build a dynamic knowledge base, and then embed cross-technical domain semantics into the dynamic knowledge base by inputting technical text to generate semantic representations. extracting multi-scale features from input technical text based on the dynamic knowledge base and semantic representation to generate a comprehensive feature set; The method extracts multi-scale features from the input technical text based on the dynamic knowledge base and semantic representation to generate a comprehensive feature set, including: Perform sentence segmentation, word segmentation and dependency syntax analysis on the input technical text to generate syntactic structure representation; According to the syntactic structure representation, global technical goals and sub-technical goals are extracted respectively, and a feature vector representing the overall function of the technology and a modular sub-technical feature set are generated; Fusing the feature vectors of the overall functionality of the technology and the feature sets of the sub-technologies to generate a comprehensive feature set; Utilizing the comprehensive feature set and the dynamic knowledge base to construct a technology dependency graph, generating an evaluation score for the professional technology, calculating a comprehensive score based on the evaluation score using cross-technology-area weights adjusted with adaptive thresholds, mapping the comprehensive score to a professional technology maturity level, and generating a professional maturity evaluation result; The method of constructing a technology dependency graph using the comprehensive feature set and the dynamic knowledge base to generate an evaluation score for professional technology includes: The nodes in the technology dependency graph are represented as sub-technologies, and the edges are represented as the dependency relationships between sub-technologies. By planning the reasoning path of the technology dependency graph through reinforcement learning, the innovation, feasibility and commercial value of professional technologies are evaluated to generate an evaluation score for the professional technologies; The innovation, feasibility and commercial value assessment of professional technologies shall include: Evaluate the innovativeness of specialized technologies by comparing the semantic similarity between sub-technologies and technical content in the dynamic knowledge base; Evaluate the feasibility of specialized technologies by matching currently available resources in the dynamic knowledge base with the required resources in the sub-technology feature set and considering the adaptability of the sub-technology environment; Evaluate the commercial value of professional technologies by estimating the potential market size and the degree of matching between application scenarios; By incorporating user feedback through online learning, the assessment results are optimized and the professionalism of the professional maturity assessment results is improved.

2. The method for evaluating professional technical maturity based on artificial intelligence according to claim 1, wherein: Access technical literature, patents, and industry reports to build a dynamic knowledge base, including: Use crawler tools to obtain technical literature, patents, and industry reports from open databases to build a dynamic knowledge base; By extracting the keywords of the input technical text and each crawled data and applying time weighting, we can filter out the technical content associated with the keywords of the input technical text. The screened technical content is merged with the knowledge base to update the knowledge base.

3. The method for evaluating professional technical maturity based on artificial intelligence according to claim 2, wherein: The input text content is combined with the dynamic knowledge base to perform semantic embedding across technical fields to generate semantic representations, including: The Transformer model of multi-task learning is used to generate semantic representations that are adapted across technical fields.

4. The method for evaluating professional technical maturity based on artificial intelligence according to claim 1, wherein: Based on the assessment scores, a comprehensive score is calculated using cross-technical field weights with adaptive threshold adjustment. The comprehensive score is mapped to a professional technology maturity level to generate a professional maturity assessment result, including: Based on the evaluation scores, weights across technical fields are determined through reinforcement learning to calculate a comprehensive score; Adaptive threshold adjustments for weights across technology areas using tiered accuracy based on historical technology assessment data; The comprehensive score after threshold adjustment is mapped to the professional and technical maturity level to obtain the professional maturity assessment result.

5. The method for evaluating professional technical maturity based on artificial intelligence according to claim 4, wherein: The professional and technical maturity levels include the start-up period, bubble period, trough period, climbing period and mature period.

6. The method for evaluating professional technical maturity based on artificial intelligence according to claim 4, wherein: Optimizing the assessment results by incorporating user feedback through online learning, including: Online learning adjusts the weights across technical fields based on the differences between user feedback and professional maturity assessment results.

Images