A pattern design method and system

Abstract

The application provides a pattern design method and system, relates to the technical field of intelligent image design, and improves the core relevance of design elements and user demand through deep semantic analysis and a dynamic weight optimization mechanism, and overcomes inherent defects of traditional methods in semantic understanding and weight distribution. The application can deeply capture the implicit association of cultural concepts, effectively solve the semantic fragmentation and missing of unpopular elements caused by traditional word frequency statistics, and ensure the comprehensiveness of keyword tags and the self-consistency of cultural logic. By fusing the co-occurrence law of historical cases and user behavior feedback, a dynamic weight system considering subjective and objective evaluation is constructed, so that the weight distribution is more in line with the real preference of users, and the accuracy and interpretability of the design scheme are enhanced. Meanwhile, the element priority can be dynamically adjusted by real-time tracking of user operation, the cost of manual intervention is reduced, and the design iteration efficiency is improved.

Description

Technical Field

[0001] This invention relates to the field of intelligent image design technology, specifically a pattern design method and system. Background Technology

[0002] Existing pattern design methods face multiple technical bottlenecks when combining contextual text with visual elements. For example, when the background image contains text, traditional text extraction techniques struggle to effectively distinguish between functional text (such as time and names) and core design keywords, leading to a discrepancy between pattern elements and user intent. When users manually draw patterns, line smoothness depends on individual skill and lacks intelligent optimization mechanisms. This is especially problematic in cultural designs, where the roughness of hand-drawn lines can easily damage the aesthetic expression of traditional patterns. Furthermore, adjusting background image parameters typically requires specialized software, making it difficult for ordinary users to achieve precise control over tone and texture through intuitive interaction. The compositing of patterns and backgrounds often involves simple overlay, lacking proper light and shadow blending and stylistic consistency, resulting in a stiff final product. These issues are particularly pronounced in scenarios requiring rapid response and emphasizing visual harmony, such as cultural and creative product design and personalized customization, hindering design efficiency and quality.

[0003] In the prior art, CN112990206A discloses a pattern design method, terminal, and computer-readable storage medium. This method involves acquiring a background scene image; performing text recognition on the image; extracting keywords if text elements are identified, determining corresponding pattern elements, and generating a target pattern; if no text elements are identified, acquiring the input design pattern, processing the lines of the design pattern to obtain a smoothed design image; determining multiple feature parameters of the background scene image, displaying multiple feature parameters and a slider corresponding to each feature parameter for user operation; acquiring the user-selected feature parameters and the slider operation, processing the background scene image to obtain an adjusted background scene image; and generating the target pattern based on the adjusted background scene image and the smoothed design image. While this method can quickly generate images according to user needs, it still suffers from superficial text processing, reliance on word frequency extraction leading to missing semantic connections and overlooked elements, static weight allocation failing to track user behavior preferences and relying on manual adjustments, resulting in low efficiency. Furthermore, parameter control lacks historical data support and requires manual slider operation, indicating a low level of intelligence.

[0004] The information disclosed in the background section is only intended to enhance the understanding of the background of this disclosure, and therefore may include information that does not constitute prior art known to those skilled in the art. Summary of the Invention

[0005] The purpose of this invention is to provide a pattern design method and system to solve the problems mentioned in the background art.

[0006] To achieve the above objectives, the present invention provides the following technical solution:

[0007] A pattern design method, the specific steps of which include:

[0008] S1: Integrate user needs into input text, and use a text analysis model to perform word frequency statistics and synonym expansion on the input text to generate keyword tags;

[0009] S2: Build a historical thesaurus based on historical success cases, use the historical thesaurus to generate the element frequency of keyword tags, and calculate the initial weight of the keyword tags;

[0010] S3: Use the element co-occurrence method to correct the initial weights, generate optimized weights, and then calculate the element integral values ​​based on the user operation records to generate the final weights;

[0011] S4: Filter the template library based on the final weights and output the optimal design scheme.

[0012] Preferably, the text analysis model uses the BERT language analysis model, and the logic for generating keyword tags is as follows:

[0013] The input text is cleaned and standardized by word roots. The input text is split into multiple independent words using a word segmentation tool. Stop words, punctuation marks, and special characters are removed, and the word forms are standardized.

[0014] The pre-trained BERT language analysis model is used to process the cleaned words and extract the word vectors from their BERT hidden layers.

[0015] A case library is built based on historical cases. The cleaned words are compared with the case library to perform word frequency analysis and ranking, with the most frequent words ranked first. Each word is identified as a keyword and categorized into keyword tags.

[0016] Preferably, when performing word frequency analysis and ranking, the word frequency is calculated as follows:

[0017]

[0018] In the formula , They represent the first The word frequency and number of occurrences of each word in the input text. Represents a word index, and , This indicates the total number of words in the input text. This represents the total number of historical cases in the case library. This indicates that the case library contains the first... The number of historical cases for each word;

[0019] The cleaned words are sorted by word frequency, with the most frequent words listed first. One, and word frequency The words are marked as keywords, in the formula This indicates the preset word frequency threshold.

[0020] Preferably, if the word frequency is satisfied Insufficient vocabulary One, using cosine similarity to find values ​​with greater than 1 in the BERT vector space. The synonyms are sorted according to their similarity, and then merged with the original words for output. This represents the similarity threshold, and .

[0021] Preferably, the logic for generating the initial weight of the keyword tag is as follows:

[0022] Collect each user's rating of historical cases. ,score Using a ten-point scale, subscript Indicates an index of historical cases, and , rating Historical cases are designated as historical success cases;

[0023] The words with keyword tags from historical success stories are compiled into a historical thesaurus. Based on this historical thesaurus, the initial weight of each word currently tagged with a keyword is calculated. The calculation method is as follows:

[0024]

[0025] In the formula This indicates the first corresponding to the input text. The initial weight of a word with keyword tags. Indicates the first The first word with keyword tags in the 1st The number of times it appears in a historical case Indicates the first Total number of words in each historical case , Represents the model parameters, where Both are greater than 0.

[0026] Preferably, the initial weights are corrected using the element co-occurrence method, and the calculation method for generating optimized weights is as follows:

[0027]

[0028]

[0029] In the formula This indicates the optimization weights. Indicates the first The first word with keyword tags in the 1st The number of times it appears in a historical case Indicates the first The first word with keyword tags and the first The first word with keyword tags in the 1st The number of times they appear simultaneously in a historical case Indicates the co-occurrence enhancement coefficient. Represents the volatility coefficient, and , This indicates the co-occurrence reinforcement coefficient in the historical lexicon. The total number of times, in the formula Indicates the co-occurrence threshold, and .

[0030] The preferred method for calculating and updating the element integral value based on the user operation record is as follows:

[0031]

[0032] In the formula , They represent the first Tianhe Di The timing, corresponding to the current input text, is... The element integral value of each word, The index representing the number of days to update. The integral variable is represented by the following method for taking its value:

[0033]

[0034] In the formula This indicates the user's view on the current input text, specifically the first character. The number of times each word was modified;

[0035] The final weights are calculated as follows:

[0036]

[0037] In the formula This indicates the final weight.

[0038] Preferably, the element integral value is updated once a day, and when the integral change rate is greater than 2, that is:

[0039]

[0040] The final weight calculation method becomes:

[0041]

[0042] In the formula This indicates the preset number of observation days.

[0043] Preferably, the matching score for each template is calculated based on the final weight, and the calculation method is as follows:

[0044]

[0045] In the formula Indicates the first The matching score of each template. Indicates the first The word in the first The number of times it appears in a template. Indicates the first The total number of elements in the keyword tags contained in each template. This represents the preset complexity attenuation coefficient;

[0046] Finally, the template with the highest matching score is output as the optimal design solution.

[0047] A pattern design system, wherein the pattern design system employs the above-mentioned pattern design method, specifically includes:

[0048] The text analysis module has a built-in text analysis model for performing word frequency statistics and synonym expansion on the input text to generate keyword tags.

[0049] A word frequency analysis module is used to construct a historical word library and generate the element frequencies of keyword tags using the historical word library;

[0050] The weight calculation module is used to calculate the initial weight of the keyword tag, and to correct the initial weight to obtain the final weight.

[0051] The template filtering module is used to calculate the matching score of each template and output the template with the highest matching score as the optimal design scheme.

[0052] Compared with the prior art, the beneficial effects of the present invention are:

[0053] This invention significantly enhances the core relevance between design elements and user needs through deep semantic analysis and a dynamic weight optimization mechanism, overcoming the inherent shortcomings of traditional methods in semantic understanding and weight allocation. The semantic expansion technology based on the BERT model can deeply capture the implicit connections between cultural concepts, effectively solving the semantic fragmentation and omission of less common elements caused by traditional word frequency statistics, ensuring the comprehensiveness of keyword tags and the consistency of cultural logic. By integrating the co-occurrence patterns of historical cases and user behavior feedback, a dynamic weight system that considers both subjective and objective evaluations is constructed, making weight allocation more aligned with users' actual preferences and enhancing the accuracy and interpretability of the design solution. Simultaneously, the adaptive integration mechanism dynamically adjusts element priorities by tracking user operations in real time, reducing the cost of manual intervention and improving design iteration efficiency. In cultural integration design scenarios, this solution can efficiently balance the expressive conflicts between traditional elements and modern aesthetics, outputting logically rigorous and visually harmonious personalized solutions, significantly improving user satisfaction and the first-time approval rate. Attached Figure Description

[0054] Figure 1 This is a schematic diagram of the overall method flow of the present invention;

[0055] Figure 2 This is a schematic diagram of the module structure of the present invention. Detailed Implementation

[0056] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to specific embodiments.

[0057] It should be noted that, unless otherwise defined, the technical or scientific terms used in this invention should have the ordinary meaning understood by one of ordinary skill in the art to which this invention pertains. The terms "first," "second," and similar terms used in this invention do not indicate any order, quantity, or importance, but are merely used to distinguish different components. Terms such as "comprising" or "including" mean that the element or object preceding the word encompasses the elements or objects listed following the word and their equivalents, without excluding other elements or objects. Terms such as "connected" or "linked" are not limited to physical or mechanical connections, but can include electrical connections, whether direct or indirect. Terms such as "upper," "lower," "left," and "right" are used only to indicate relative positional relationships; when the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0058] Example:

[0059] Please see Figures 1-2 The present invention provides a technical solution:

[0060] A pattern design method, the specific steps of which include:

[0061] S1: Integrate user needs into input text, and use a text analysis model to perform word frequency statistics and synonym expansion on the input text to generate keyword tags.

[0062] The text analysis model uses the BERT language analysis model, and the logic for generating keyword tags is as follows:

[0063] The input text is cleaned and standardized by word roots. The input text is split into multiple independent words using a word segmentation tool. Stop words, punctuation marks, and special characters are removed, and the word forms are standardized.

[0064] The pre-trained BERT language analysis model is used to process the cleaned words and extract the word vectors from their BERT hidden layers.

[0065] A case library is built based on historical cases. The cleaned words are compared with the case library to perform word frequency analysis and ranking, with the most frequent words ranked first. Each word is identified as a keyword and categorized into keyword tags.

[0066] Specifically, in practical use, the NLTK segmentation tool can be used to split the words into individual words, removing stop words and non-textual symbols. Next, in the standardization stage, plural forms are unified to singular, and verb tenses are converted to their prototypes to ensure consistency in word form. Finally, a 768-dimensional hidden layer word vector is extracted using a pre-trained BERT model to establish a semantic feature space.

[0067] When performing word frequency analysis and ranking, the word frequency is calculated as follows:

[0068]

[0069] In the formula , They represent the first The word frequency and number of occurrences of each word in the input text. Represents a word index, and , This indicates the total number of words in the input text. This represents the total number of historical cases in the case library. This indicates that the case library contains the first... The number of historical cases for each word;

[0070] The cleaned words are sorted by word frequency, with the most frequent words listed first. One, and word frequency The words are marked as keywords, in the formula This indicates the preset word frequency threshold.

[0071] If word frequency is satisfied Insufficient vocabulary One, using cosine similarity to find values ​​with greater than 1 in the BERT vector space. The synonyms are sorted according to their similarity, and then merged with the original words for output. This represents the similarity threshold, and .

[0072] In this embodiment, assuming the user input text is "Chinese style jewelry design, which should reflect a combination of traditional culture and modern simplicity," the following settings are configured. After the input text is cleaned, it is sorted by word frequency analysis based on the historical case library (assuming it contains 1000 successful cases). The word frequency of each word is calculated to obtain a set of keywords {"Guofeng", "Jewelry", "Design", "Tradition", "Culture", "Modern", "Simple"}. At this time, there are less than 10 keywords. The cosine similarity is calculated using BERT word vectors to expand synonyms such as "blue and white porcelain", "entwined branches", and "cloud pattern". After merging, a keyword tag containing 10 elements is generated.

[0073] In this step, compared to the traditional TF-IDF which only counts word frequency, the BERT model captures the semantic association between "Guofeng" (traditional Chinese style) and "Qinghuaci" (blue and white porcelain) through a 768-dimensional vector, resulting in a higher keyword recall rate. Moreover, by adjusting the similarity threshold, it can adaptively adjust the expansion range to solve the problem of insufficient coverage of niche demands. At the same time, it can also provide highly relevant input for subsequent weight calculations, avoiding the deviation of the solution from the core needs caused by the omission of keywords in traditional methods.

[0074] S2: Build a historical thesaurus based on historical success cases, use the historical thesaurus to generate the element frequency of keyword tags, and calculate the initial weight of the keyword tags.

[0075] The logic for generating the initial weight of keyword tags is as follows:

[0076] Collect each user's rating of historical cases. ,score Using a ten-point scale, subscript Indicates an index of historical cases, and , rating Historical cases are designated as historical success cases;

[0077] The words with keyword tags from historical success stories are compiled into a historical thesaurus. Based on this historical thesaurus, the initial weight of each word currently tagged with a keyword is calculated. The calculation method is as follows:

[0078]

[0079] In the formula This indicates the first corresponding to the input text. The initial weight of a word with a keyword tag, indicating the th occurrence count of the word with a keyword tag in the th historical case, indicating the total number of words in the , representing model parameters, where , and both are greater than 0.

[0080] The specific values of the model parameters can be determined according to expert experience or by using the grid search method to find the optimal allocation. The specific principle is prior art and will not be elaborated here. In this embodiment, it is assumed that the optimal model parameters are , , respectively. At the same time, taking "national style" numbered 1 as an example, assuming it appears 1200 times in 450 successful cases, the total word frequency of the whole library is 50000 times, and the average total score is 8.64, then:

[0081]

[0082] In this step, by integrating the objective data of word frequency (accounting for 65%, that is, the model parameter ) and the user's subjective score (accounting for 35%, that is, the model parameter ), compared with the single-index method, it can greatly improve the rationality of the weight, making it more meet the actual needs of customers. Moreover, only using the data of successful cases to exclude the interference of low-quality designs can make the overall convergence speed of the model faster, reduce its response time, and can also construct a robust weight baseline to provide a reliable starting point for subsequent dynamic optimization, avoiding the initial deviation caused by data mixing in traditional methods.

[0083] S3: Use the element co-occurrence method to correct the initial weight to generate an optimized weight, and then calculate the element integral value according to the user operation record to generate the final weight.

[0084] The calculation method of using the element co-occurrence method to correct the initial weight to generate an optimized weight is:

[0085]

[0086]

[0087] In the formula represents the optimized weight, represents the th occurrence count of the word with a keyword tag in the th historical case, Indicates the number of times that the th word with a keyword tag and the th word with a keyword tag appear simultaneously in the th historical case. represents the co-occurrence enhancement coefficient, , represents the total number of times in the historical thesaurus that meet the co-occurrence enhancement coefficient . In the formula, represents the co-occurrence threshold, and .

[0088] In this embodiment, assuming that the preset co-occurrence threshold is 0.6, then a co-occurrence enhancement coefficient greater than 0.6 can be called a strong association rule. Specifically, if the number of co-occurrence cases of the word "national style" numbered 1 and the word "blue and white porcelain" numbered 7 is 280, and the number of cases where "blue and white porcelain" appears alone is 400, then . The value of the fluctuation coefficient can also be determined or modified according to expert experience. Assuming that the fluctuation coefficient is 0.15, taking "national style" as an example, if it has strong associations with 3 words, and the co-occurrence enhancement coefficients are 0.7, 0.65, and 0.62 respectively, then the optimized weight of "national style" can be expressed as:

[0089]

[0090] The calculation method for calculating and updating the element integral value based on the user operation record is:

[0091]

[0092] In the formula, , respectively represent the th day and the th day, and the element integral value corresponding to the th word in the current input text, represents the index of the update days, represents the integral variable, and its value-taking method is: <了

[0093]

[0094] In the formula, represents the number of times the user modifies the th word in the current input text;

[0095] The calculation method of the final weight is:

[0096]

[0097] In the formula, It should be noted that there seems to be some incorrect text in the original Chinese text, such as "了0000338" which is likely a misrepresentation. This translation is based on the best understanding of the provided text. This indicates the final weight.

[0098] The element integral value is updated once a day, and when the rate of change of the integral is greater than 2, that is:

[0099]

[0100] The final weight calculation method becomes:

[0101]

[0102] In the formula This indicates the preset number of observation days. The number of observation days can be set according to user needs. 3 days, 7 days, and 30 days are common settings. This is to avoid extreme user operations affecting the overall stability of the system. Therefore, data smoothing is required when sudden changes occur.

[0103] In this step, by setting an adjustable co-occurrence threshold, implicit matching rules of design elements can be effectively identified, greatly improving the interpretability of the scheme logic. Moreover, by setting element integral values ​​that are updated according to user operations, the weight system can have adaptive capabilities, realizing "prioritizing the strengthening of low-weight elements", thereby reducing the number of user modifications and solving the pain point that traditional static weights cannot track changes in user preferences.

[0104] S4: Filter the template library based on the final weights and output the optimal design scheme.

[0105] The matching score for each template is calculated based on the final weights, and the calculation method is as follows:

[0106]

[0107] In the formula Indicates the first The matching score of each template. Indicates the first The word in the first The number of times it appears in a template. Indicates the first The total number of elements in the keyword tags contained in each template. This represents the preset complexity attenuation coefficient. Similarly, the complexity attenuation coefficient is determined and modified based on expert experience, and can usually be set to 0.05.

[0108] Finally, the template with the highest matching score is output as the optimal design solution.

[0109] Assuming template 1 contains the keywords "traditional Chinese style," "blue and white porcelain," and "simple," with occurrences of 2, 3, and 1 respectively, and final weights of 0.613, 0.581, and 0.527 respectively, then:

[0110]

[0111] The score is rounded to one decimal place (approximately 1.6). The top three templates with the highest scores (e.g., Score ≥ 1.5) are then used to generate design schemes via the rendering engine.

[0112] A pattern design system, employing the above-mentioned pattern design method, specifically includes:

[0113] The text analysis module has a built-in text analysis model used to perform word frequency statistics and synonym expansion on the input text, and generate keyword tags.

[0114] The word frequency analysis module is used to build a historical thesaurus and generate the element frequencies of keyword tags using the historical thesaurus.

[0115] The weight calculation module is used to calculate the initial weight of the keyword tag, and then correct the initial weight to obtain the final weight.

[0116] The template filtering module calculates the matching score for each template and outputs the template with the highest matching score as the optimal design solution.

[0117] In summary, this invention significantly enhances the core relevance between design elements and user needs through deep semantic analysis and a dynamic weight optimization mechanism, overcoming the inherent shortcomings of traditional methods in semantic understanding and weight allocation. The semantic expansion technology based on the BERT model can deeply capture the implicit connections between cultural concepts, effectively solving the semantic fragmentation and omission of less common elements caused by traditional word frequency statistics, ensuring the comprehensiveness of keyword tags and the self-consistency of cultural logic. By integrating the co-occurrence patterns of historical cases and user behavior feedback, a dynamic weight system that considers both subjective and objective evaluations is constructed, making weight allocation more aligned with users' actual preferences and enhancing the accuracy and interpretability of the design solution. Simultaneously, the adaptive integration mechanism dynamically adjusts element priorities by tracking user operations in real time, reducing the cost of manual intervention and improving design iteration efficiency. In cultural integration design scenarios, this solution can efficiently balance the expressive conflicts between traditional elements and modern aesthetics, outputting logically rigorous and visually harmonious personalized solutions, significantly improving user satisfaction and the first-time approval rate.

[0118] The above formulas are all dimensionless calculations. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters in the formulas are set by those skilled in the art according to the actual situation.

[0119] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented in software, the above embodiments can be implemented, in whole or in part, as a computer program product. Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented by electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution.

[0120] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment, depending on actual needs.

[0121] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application.

Claims

1. A pattern design method, characterized in that, The specific steps include: S1: Integrate user needs into input text, and use a text analysis model to perform word frequency statistics and synonym expansion on the input text to generate keyword tags; S2: Build a historical thesaurus based on historical success cases, use the historical thesaurus to generate the element frequency of keyword tags, and calculate the initial weight of the keyword tags; S3: Use the element co-occurrence method to correct the initial weights, generate optimized weights, and then calculate the element integral values ​​based on the user operation records to generate the final weights; The initial weights are corrected using the element co-occurrence method, and the optimized weights are calculated as follows: ； ； In the formula This indicates the optimization weights. This indicates the first corresponding to the input text. The initial weight of a word with keyword tags. Indicates the first The first word with keyword tags in the 1st The number of times it appears in a historical case Indicates the first The first word with keyword tags and the first The first word with keyword tags in the 1st The number of times they appear simultaneously in a historical case Indicates the co-occurrence enhancement coefficient. Represents a word index, and , This indicates the total number of words in the input text. This represents the total number of historical cases in the case library. Represents the volatility coefficient, and , This indicates the co-occurrence reinforcement coefficient in the historical lexicon. The total number of times, in the formula Indicates the co-occurrence threshold, and ; The calculation method for calculating and updating the element integral value based on the user operation record is as follows: ； In the formula , They represent the first Tianhe Di The timing, corresponding to the current input text, is... The element integral value of each word, The index representing the number of days to update. The integral variable is represented by the following method for taking its value: ； In the formula This indicates the user's view on the current input text, specifically the first character. The number of times each word was modified; The final weights are calculated as follows: ； In the formula Indicates the final weight; S4: Filter the template library based on the final weights and output the optimal design scheme.

2. The pattern design method according to claim 1, characterized in that: The text analysis model uses the BERT language analysis model, and the logic for generating keyword tags is as follows: The input text is cleaned and standardized by word roots. The input text is split into multiple independent words using a word segmentation tool. Stop words, punctuation marks, and special characters are removed, and the word forms are standardized. The pre-trained BERT language analysis model is used to process the cleaned words and extract the word vectors from the BERT hidden layer. A case library is built based on historical cases. The cleaned words are compared with the case library to perform word frequency analysis and ranking, with the most frequent words ranked first. Each word is identified as a keyword, and keyword tags are added to these words. It is a positive integer.

3. The pattern design method according to claim 2, characterized in that: When performing word frequency analysis and ranking, the word frequency is calculated as follows: ； In the formula , They represent the first The word frequency and number of occurrences of each word in the input text. This indicates that the case library contains the first... The number of historical cases for each word; The cleaned words are sorted by word frequency, with the most frequent words listed first. One, and word frequency The words are marked as keywords, in the formula This indicates the preset word frequency threshold.

4. The pattern design method according to claim 3, characterized in that: If word frequency is satisfied Insufficient vocabulary One, using cosine similarity to find values ​​with greater than 1 in the BERT vector space. The synonyms are sorted according to their similarity, and then merged with the original words for output. This represents the similarity threshold, and .

5. The pattern design method according to claim 3, characterized in that: The logic for generating the initial weights of the keyword tags is as follows: Collect each user's rating of historical cases. ,score Using a ten-point scale, subscript Indicates an index of historical cases, and , rating Historical cases are designated as historical success cases; The words with keyword tags from historical success stories are compiled into a historical thesaurus. Based on this historical thesaurus, the initial weight of each word currently tagged with a keyword is calculated. The calculation method is as follows: ； In the formula Indicates the first The first word with keyword tags in the 1st The number of times it appears in a historical case Indicates the first Total number of words in historical cases , Represents the model parameters, where Both are greater than 0.

6. The pattern design method according to claim 5, characterized in that: The element integral value is updated once a day, and when the integral change rate is greater than 2, that is: ； The final weight calculation method becomes: ； In the formula This indicates the preset number of observation days.

7. A pattern design method according to claim 6, characterized in that: The matching score for each template is calculated based on the final weights, and the calculation method is as follows: ； In the formula Indicates the first The matching score of each template. Indicates the first The word in the first The number of times it appears in a template. Indicates the first The total number of elements in the keyword tags contained in each template. This represents the preset complexity attenuation coefficient; Finally, the template with the highest matching score is output as the optimal design solution.

8. A pattern design system, characterized in that: The pattern design system employs the pattern design method as described in any one of claims 1-7, specifically including: The text analysis module has a built-in text analysis model for performing word frequency statistics and synonym expansion on the input text to generate keyword tags. A word frequency analysis module is used to construct a historical word library and generate the element frequencies of keyword tags using the historical word library; The weight calculation module is used to calculate the initial weight of the keyword tag, and to correct the initial weight to obtain the final weight. The template filtering module is used to calculate the matching score of each template and output the template with the highest matching score as the optimal design scheme.

Images