A patent analysis and evaluation method based on artificial intelligence

Through patent analysis and evaluation methods based on artificial intelligence, combined with natural language processing and patent age factors, the problem that traditional methods are difficult to evaluate the value and influence of new patents is solved, and the accuracy of keyword extraction and patent impact assessment is improved. , ensure that the evaluation results are more in line with human understanding and market needs.

CN118657634BActive Publication Date: 2024-12-06QINGDAO HUAHUIZE INTELLECTUAL PROPERTY AGENCY CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411147495.0
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-08-21
Publication Date
2024-12-06
Estimated Expiration
2044-08-21

AI Technical Summary

Technical Problem

Traditional patent value evaluation methods are difficult to immediately and accurately judge the value and influence of new patents, especially due to the lack of sufficient historical data or citation records, it is difficult for existing systems to immediately reflect the real influence changes of patents.

Method used

Using patent analysis and evaluation methods based on artificial intelligence, the natural language processing is carried out by collecting multi-source data for cleaning and preprocessing, including topic modeling, keyword extraction and semantic similarity calculation, and part-of-speech weighting is introduced in keyword extraction. , word frequency weighting, position weighting and context similarity factors are optimized, and factors affecting patent age are introduced influencing the analysis to form a patent value evaluation report.

Benefits of technology

Improve the accuracy and relevance of keyword extraction, and can take into account the attributes of vocabulary in text more comprehensively, ensure that the evaluation results are more in line with human understanding and reading habits, and through the introduction of patent age factors, newer patents can more accurately reflect Its potential influence solves the problem that the value and influence of new patents are difficult to judge.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118657634B_ABST
    Figure CN118657634B_ABST
Patent Text Reader

Abstract

The present invention relates to the technical field of patent analysis and evaluation, and specifically, to a patent analysis and evaluation method based on artificial intelligence. It includes the following steps: collecting multi-source data of patents, cleaning and text preprocessing the multi-source data; performing natural language processing on the cleaned and preprocessed data; conducting value assessment on the patent through the results of natural language processing, and the value assessment includes impact analysis, life cycle judgment and future value prediction; integrating the results of impact analysis, life cycle judgment and future value prediction to form a patent value assessment report. This method can not only improve the accuracy and relevance of keyword extraction, but also introduce the influencing factor of patent age in the impact analysis for optimization, solving the problem that it is difficult to accurately judge the value of a new patent immediately after it is submitted or published.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of patent analysis and evaluation, and in particular to a patent analysis and evaluation method based on artificial intelligence. Background Art

[0002] With the development of emerging technologies such as artificial intelligence, big data, and cloud computing, the creation, protection, and commercialization of patents are also constantly evolving, which places higher demands on analytical methods, which need to be able to adapt to technological trends and provide forward-looking insights. The true value and influence of a patent is a dynamic process, which is affected by multiple factors such as technological development, market demand, and legal environment. Therefore, a patent value assessment method is needed to reflect the changes in its true influence. Traditional patent keyword extraction methods may only be based on simple statistical methods such as word frequency or TF-IDF, ignoring the context and grammatical structure of the vocabulary in the text, and when a new patent is submitted or published, due to the lack of sufficient historical data or citation records, it is difficult for the existing evaluation system to immediately and accurately judge the value and influence of the patent. Therefore, a patent analysis and evaluation method based on artificial intelligence is designed. Summary of the invention

[0003] The purpose of the present invention is to provide a patent analysis and evaluation method based on artificial intelligence to solve the problem raised in the above background technology that it is difficult to immediately and accurately judge the value and influence of a new patent after it is submitted or published.

[0004] To achieve the above objectives, the present invention provides a patent analysis and evaluation method based on artificial intelligence, comprising the following steps:

[0005] S1. Collect multi-source data of patents, clean and pre-process the multi-source data;

[0006] S2. Perform natural language processing on the cleaned and preprocessed data. Natural language processing includes topic modeling, keyword extraction, and semantic similarity calculation. In keyword extraction, factors such as part of speech weighting, word frequency weighting, position weighting, and context similarity are introduced for optimization.

[0007] S3. Use the results of natural language processing to evaluate the value of the patent, including impact analysis, life cycle judgment and future value prediction; and introduce the influencing factor of patent age into the impact analysis for optimization;

[0008] S4. Integrate the results of impact analysis, life cycle judgment and future value prediction to form a patent value assessment report, and invite experts to review and analyze the report.

[0009] As a further improvement of the present technical solution, the text preprocessing includes word segmentation, removal of stop words and stemming.

[0010] As a further improvement of this technical solution, the topic modeling is based on Model, The model is as follows:

[0011] ;

[0012] in, for The joint probability distribution of the model; is the set of all words; A collection of all themes; is the topic distribution of all documents; The word distribution for all topics; and are the hyperparameters of the model; The word distribution for all topics The prior probability of The topic distribution of all documents The prior probability of is the total number of documents; For the The number of words in a document; For the In the document Topic assignment of words Given the topic distribution of the document The probability of the following; For the In the document the subject to which the word belongs; For the The topic distribution of the documents; For a given topic And the word distribution of all topics Under the conditions of In the document Words probability; For the In the document words; ; .

[0013] As a further improvement of this technical solution, the keyword extraction is based on , as follows:

[0014] ;

[0015] in, For Node score; is the damping coefficient; For the node The number of directly connected neighbor nodes; For Node Point to Node The weight of the edge; For Node The out-degree of For Node score; .

[0016] As a further improvement of this technical solution, factors influencing part of speech weighting, word frequency weighting, position weighting and context similarity are introduced in keyword extraction for optimization, specifically:

[0017] ;

[0018] in, For the optimized node score; For Node The frequency of words; For Node Weighting factors for parts of speech; For Node and nodes Contextual similarity between For Node The weight of the position in the text; For Node The weighted out-degree of .

[0019] As a further improvement of the present technical solution, the semantic similarity calculation is first normalized according to the cosine similarity score of the bag-of-words model and the similarity score of the deep learning model, and then obtained by weighted average of the two normalized scores.

[0020] As a further improvement of this technical solution, the impact analysis is based on The algorithm used to assess the importance of a patent in a field is as follows:

[0021] ;

[0022] in, For patent of value; is the damping factor; is the total number of patents in the network; To point to patents All patent collections of; To point to patents All patent collections of; For patent of value; For patent The number of citations; For patent The frequency of citations.

[0023] As a further improvement of the technical solution, the influencing factor of patent age is introduced into the impact analysis for optimization, specifically:

[0024] ;

[0025] in, For patent A natural growth trend of gaining citations over time; To adjust the parameters; For patent Age;

[0026] ;

[0027] in, For the optimized patent of value; For patent A natural growth trend of gaining citations over time; For patent A natural growth trend of gaining citations over time.

[0028] As a further improvement of the present technical solution, the life cycle includes an application period, a review period, an authorization period, a maintenance period and an expiration period.

[0029] As a further improvement of this technical solution, the results of the impact analysis, life cycle judgment and future value prediction are integrated as follows:

[0030] ;

[0031] in, For patent the value of is the weight factor; for The degree of influence of value factors on patent value; For patent life cycle; The impact of the life cycle on the patent value; For patent The future value of The degree of impact of future value on patent value.

[0032] Compared with the prior art, the present invention has the following beneficial effects:

[0033] 1. In this patent analysis and evaluation method based on artificial intelligence, the factors of part-of-speech weighting, word frequency weighting, position weighting and context similarity are introduced in keyword extraction for optimization, which can more comprehensively consider the various attributes of words in the text, thereby improving the accuracy and relevance of keyword extraction and making it more in line with human understanding and reading habits.

[0034] 2. In this patent analysis and evaluation method based on artificial intelligence, the influencing factor of patent age is introduced into the impact analysis for optimization. In this way, even if the citation frequency of newer patents is not high, they will be more cited due to the effect of age factor. At the same time, as the age of the patent increases, its age factor tends to a stable value, which means that its The value will rely more on actual citations, solving the problem that it is difficult to accurately judge the value and influence of a patent immediately after it is submitted or published. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] Figure 1 The figure is a flow chart of the overall method of the present invention. DETAILED DESCRIPTION

[0036] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0037] Example

[0038] See also Figure 1 As shown, a patent analysis and evaluation method based on artificial intelligence is provided, comprising the following steps:

[0039] S1. Collect multi-source data of patents, clean and pre-process the multi-source data;

[0040] Text preprocessing includes word segmentation, stop word removal, stemming, and preparing text data for natural language processing;

[0041] Word segmentation is the process of splitting text into words or phrases, which can help algorithms identify the basic building blocks of text, form the basis for constructing word frequency statistics, and thus more accurately identify topics. It can also determine the independent words in the text, facilitating the identification of which words appear frequently in the text and may constitute keywords. It enables the system to compare the lexical compositions of different texts and calculate their similarities.

[0042] Stop word removal is to remove common words in the text that contribute little to semantics, such as "de", "shi", "zai", etc. It can reduce noise, enabling the model to focus more on meaningful words, improve the accuracy of topic recognition, avoid wasting resources on processing meaningless words, and also enhance the efficiency and quality of keyword recognition and improve the comparison accuracy, because stop words often do not affect the main meaning of the sentence.

[0043] Stemming converts words into their basic forms or roots, which can ensure that different inflections of the same root are regarded as the same word, reducing the sparsity of the topic model. It can also reduce word variants, making keyword statistics more concentrated, enabling the algorithm to recognize that different inflected words are actually the same word, and thus more accurately calculate text similarity.

[0044] S2. Perform natural language processing on the cleaned and preprocessed data. Natural language processing includes topic modeling, keyword extraction, and semantic similarity calculation; and introduce influencing factors such as part-of-speech weighting, word frequency weighting, position weighting, and context similarity in keyword extraction for optimization;

[0045] Through topic modeling, the main research topics and development trends in a certain technical field can be identified, and the data can be automatically classified according to its topics, which helps to quickly search for and understand patent information in specific fields; analyze the patent portfolios of competitors and identify their R & D focuses and strategies;

[0046] Extracting keywords from the data helps to quickly understand the core content of the data; thus generating keyword tags to improve the search efficiency in the database; through keyword analysis, the specific technical fields involved in the data can be determined, facilitating the analysis of the segmented market.

[0047] By calculating the semantic similarity between patents, identifying technically similar or duplicate inventions helps to avoid the risk of patent infringement; identifying patents within the same family, even if they are applied for in different countries, may have similar technical descriptions; tracking the semantic changes of patents in a specific technical field to understand the technical evolution path.

[0048] Topic modeling is based on model, The model is as follows:

[0049] ;

[0050] in, for The joint probability distribution of the model; is the set of all words in the corpus, Contains all instances of the word in all documents; A collection of all themes; is the topic distribution of all documents; The word distribution for all topics; and is a hyperparameter of the model, Control The shape of the multivariate probability distribution of determines how many topics documents tend to contain and the concentration of these topics. controls the prior on the probability distribution of words in a topic, Influence The multivariate probability distribution of determines whether the distribution of words in each topic is uniform or biased towards a few words; The word distribution for all topics The prior probability of The topic distribution of all documents The prior probability of is the total number of documents; For the The number of words in a document; For the In the document Topic assignment of words Given the topic distribution of the document The probability of the following; For the In the document the subject to which the word belongs; For the The topic distribution of the documents; For a given topic And the word distribution of all topics Under the conditions of In the document Words probability; For the In the document words; ; .

[0051] exist During the model generation process, and is used to generate the topic distribution of documents and the word probability distribution of topics, and and are randomly generated observations based on these distributions. In the inference phase, our goal is to To estimate , and 's value to understand the topic structure of the document.

[0052] Keyword extraction based on , as follows:

[0053] It is a graphical model algorithm used for text summarization and keyword extraction. It constructs a graph to represent sentences or words in the text and calculates the score of each node through an iterative process. These scores reflect the importance of the node in the entire text.

[0054] ;

[0055] in, For Node score; is the damping coefficient; For the node The number of directly connected neighbor nodes; For Node Point to Node The weight of the edge; For Node The out-degree of For Node score; ;

[0056] The formula continuously updates the score of each node through an iterative process until convergence is reached, that is, the score change is less than a certain threshold or the preset maximum number of iterations is reached;

[0057] In keyword extraction, we introduce factors such as part of speech weighting, word frequency weighting, position weighting, and context similarity to optimize the extraction. Word frequency weighting takes into account that high-frequency words may be more important than low-frequency words. Modified to a value proportional to the word frequency. Position weighting In a sentence or paragraph, the words at the beginning and end may be more informative than the words in the middle. The position factor can be introduced to adjust the weight. Part-of-speech weighting is because nouns, verbs, etc. may have more potential as keywords than prepositions, articles, etc., and different weights can be given according to the part of speech. Contextual similarity is that the weight of the edge between words can be based on their semantic similarity, such as using a word embedding model to calculate the cosine similarity of word vectors. Specifically:

[0058] ;

[0059] in, For the optimized node score; For Node The frequency of words; For Node Weighting factors for parts of speech; For Node and nodes Contextual similarity between For Node The weight of the position in the text; For Node The weighted out-degree of .

[0060] By considering word frequency, high-frequency words may be given higher weights because they appear more often in the text, which helps to identify the subject words or core concepts in the text; different parts of speech play different roles in the text, for example, nouns and verbs often carry more substantive information, while prepositions, conjunctions, etc. may be relatively minor. By weighting the parts of speech, the algorithm can pay more attention to those semantically more important words; by calculating the contextual similarity between words, the association between words can be captured, which helps to identify those words that are closely related in semantics but not necessarily frequently co-occurring. For example, two professional terms may not appear together often, but they are closely related in semantics, and the improved formula can better reflect this. At the same time, in a sentence or paragraph, the words at the beginning and end often attract readers' attention more than the words in the middle, and may also contain more key information. The introduction of position weights can allow the algorithm to pay more attention to the words in these positions.

[0061] In summary, the improved The formula can more comprehensively consider the various attributes of words in the text, thereby improving the accuracy and relevance of keyword extraction and making it more in line with human understanding and reading habits.

[0062] The semantic similarity calculation is first normalized according to the cosine similarity score of the bag-of-words model and the similarity score of the deep learning model, and then the weighted average of the two normalized scores is obtained.

[0063] S3. Use the results of natural language processing to evaluate the value of the patent, including impact analysis, life cycle judgment and future value prediction; and introduce the influencing factor of patent age into the impact analysis for optimization;

[0064] Impact analysis based on Algorithms to assess the importance of patents in the field, The algorithm was originally used to rank web pages, but can also be used to rank patent citation networks. The basic idea is that the importance of a patent depends not only on the number of patents that directly cite it, but also on the importance of the patents that cite it, as follows:

[0065] ;

[0066] in, For patent of value; is the damping factor; is the total number of patents in the network; To point to patents All patent collections of; To point to patents All patent collections of; For patent of value; For patent The number of citations; For patent The number of citations;

[0067] In the above formula, patent of The value consists of two parts, one is the global average Value , the other part is patent The weighted sum of the importance of the citations. In this weighted sum, the patent Patent The contribution depends on of value, The number of citations, and How many patents have cited it? With higher value and citation frequency, and is cited by many patents at the same time, then it has a great impact on the patent The contribution is even greater.

[0068] This formula is calculated iteratively until The value converges, thus obtaining the stability of each patent in the patent network. The value helps to assess the influence and importance of patents in the field of technology.

[0069] The impact analysis is optimized by introducing the influencing factor of patent age, because newer patents may not have accumulated many citations yet, but this does not mean that they are not important. Specifically:

[0070] ;

[0071] in, For patent A natural growth trend of gaining citations over time; To adjust the parameters; For patent Age, which is the difference in years between the patent application date and the current date;

[0072] ;

[0073] in, For the optimized patent of value; For patent A natural growth trend of gaining citations over time; For patent A natural growth trend of gaining citations over time.

[0074] Thus, even if the citation frequency of newer patents is not high, they will still be cited because of the age factor. At the same time, as the age of the patent increases, its age factor tends to a stable value, which means that its The value will depend more on the actual reference situation.

[0075] The life cycle includes the application period, examination period, authorization period, maintenance period and expiration period;

[0076] S4. Integrate the results of impact analysis, life cycle judgment and future value prediction to form a patent value assessment report, and invite experts to review and analyze the report.

[0077] The results of impact analysis, life cycle judgment and future value forecast are integrated as follows:

[0078] ;

[0079] in, For patent the value of is the weight factor; for The degree of influence of value factors on patent value; For patent life cycle; The impact of the life cycle on the patent value; For patent The future value of The degree of impact of future value on patent value.

[0080] This model expresses that the three factors jointly determine the patent value, and the influence of these factors can be expressed by the index , and By adjusting these indexes, the model can reflect the advantages of different patents in different aspects, thus giving a more objective evaluation of the patent value. , , and This needs to be determined through statistical analysis, market testing or expert evaluation to ensure the accuracy and reliability of the model.

[0081] The above shows and describes the basic principles, main features and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited by the above embodiments, and the above embodiments and descriptions are only preferred examples of the present invention, and are not intended to limit the present invention. Without departing from the spirit and scope of the present invention, the present invention may have various changes and improvements, and these changes and improvements all fall within the scope of the present invention to be protected.

Claims

1. A patent analysis and evaluation method based on artificial intelligence, It is characterized in that The steps include: S1. Collect multi-source data of patents, clean and pre-process the multi-source data; S2. Perform natural language processing on the cleaned and preprocessed data. Natural language processing includes topic modeling, keyword extraction, and semantic similarity calculation. In keyword extraction, factors such as part of speech weighting, word frequency weighting, position weighting, and context similarity are introduced for optimization. S3. Use the results of natural language processing to evaluate the value of the patent, including impact analysis, life cycle judgment and future value prediction; and introduce the influencing factor of patent age into the impact analysis for optimization; S4. Integrate the results of impact analysis, life cycle judgment and future value prediction to form a patent value assessment report, and invite experts to review and analyze the report; The impact analysis is based on The algorithm used to assess the importance of a patent in a field is as follows: ; in, For patent of value; is the damping factor; is the total number of patents in the network; To point to patents All patent collections of; To point to patents All patent collections of; For patent of value; For patent The number of citations; For patent The number of citations; The influencing factors of patent age are introduced into the impact analysis for optimization, specifically: ; in, For patent A natural growth trend of gaining citations over time; To adjust the parameters; For patent Age; ; in, For the optimized patent of value; For patent A natural growth trend of gaining citations over time; For patent A natural growth trend of gaining citations over time.

2. According to the patent analysis and evaluation method based on artificial intelligence according to claim 1, Features: The text preprocessing includes word segmentation, stop word removal and stemming.

3. According to the patent analysis and evaluation method based on artificial intelligence according to claim 2, Features: The topic modeling is based on Model, The model is as follows: ; in, for The joint probability distribution of the model; is the set of all words; A collection of all themes; is the topic distribution of all documents; The word distribution for all topics; and are the hyperparameters of the model; The word distribution for all topics The prior probability of The topic distribution of all documents The prior probability of is the total number of documents; For the The number of words in a document; For the In the document Topic assignment of words Given the topic distribution of the document The probability of the following; For the In the document the subject to which the word belongs; For the The topic distribution of the documents; For a given topic And the word distribution of all topics Under the conditions of In the document Words probability; For the In the document words; ; .

4. According to the patent analysis and evaluation method based on artificial intelligence according to claim 3, Features: The keyword extraction is based on , as follows: ; in, For Node score; is the damping coefficient; For the node The number of directly connected neighbor nodes; For Node Point to Node The weight of the edge; For Node The out-degree of For Node score; .

5. According to the patent analysis and evaluation method based on artificial intelligence according to claim 4, Features: In keyword extraction, factors such as part of speech weighting, word frequency weighting, position weighting, and context similarity are introduced for optimization, specifically: ; in, For the optimized node score; For Node The frequency of words; For Node Weighting factors for parts of speech; For Node and nodes Contextual similarity between For Node The weight of the position in the text; For Node The weighted out-degree of .

6. According to the patent analysis and evaluation method based on artificial intelligence according to claim 5, Features: The semantic similarity calculation is first normalized according to the cosine similarity score of the bag-of-words model and the similarity score of the deep learning model, and then obtained by weighted average of the two normalized scores.

7. According to the patent analysis and evaluation method based on artificial intelligence according to claim 6, Features: The life cycle includes application period, examination period, authorization period, maintenance period and expiration period.

8. According to the patent analysis and evaluation method based on artificial intelligence according to claim 7, Features: The results of the impact analysis, life cycle judgment and future value forecast are integrated as follows: ; in, For patent the value of is the weight factor; for The degree of influence of value factors on patent value; For patent life cycle; The impact of the life cycle on the patent value; For patent The future value of The degree of impact of future value on patent value.