Intelligent Printing Layout Method and Device Based on Machine Learning

Abstract

The present invention relates to the technical field of image generation, and discloses an intelligent printing layout method and device based on machine learning. The present invention collects a set of demand descriptions of a user for a target printing layout, extracts and transforms features of the set of demand descriptions to generate a high-dimensional guiding feature distribution to characterize the layout requirements, uses a pre-trained intelligent printing layout model to perform multiple data simulations on the high-dimensional feature distribution to generate multiple original versions of the printing layout, collects optimization suggestions of the user for each original version of the layout through a user feedback channel, adjusts the high-dimensional feature distribution according to the feedback, and re-performs simulations through the intelligent model to generate a final target finished printing layout. By analyzing the description mode of the set of demand descriptions provided by the user itself, the guiding value of the set of demand descriptions is analyzed, thereby driving the intelligent printing layout model to work, and solving the problem in the prior art that the intelligent layout efficiency is low due to the inability to accurately identify the customer's expectations.

Description

Technical Field

[0001] The present invention relates to the technical field of image generation, and in particular, to an intelligent printing layout method and device based on machine learning. Background Art

[0002] Traditional printing layout design generally relies on manual experience or layout templates based on fixed rules. This method has certain limitations and cannot fully meet the changing and personalized layout needs of users. In addition, the manual layout process is not only cumbersome and time-consuming, but also the quality and effect of the design are easily restricted by the experience of designers, making it difficult to achieve a layout effect that fully meets the user's needs.

[0003] In recent years, with the rapid development of machine learning, deep learning, and natural language processing technologies, intelligent layout technology has gradually become an effective means to improve layout efficiency and accuracy. The intelligent printing layout method based on machine learning can automatically identify user needs and optimize layout design by learning a large number of historical layout samples, and then generate a high-quality printing finished product layout plan. However, when using a machine learning model for intelligent printing layout, it often fails to accurately identify the expectations for the finished product in the requirements proposed by customers, resulting in mistakes in the generation direction and leaving room for improvement in intelligent printing layout efficiency. Summary of the Invention

[0004] The purpose of the present invention is to provide an intelligent printing layout method and device based on machine learning, aiming to solve the problem of low intelligent layout efficiency caused by the inability to accurately identify customer expectations in the prior art.

[0005] The present invention is implemented as follows. In the first aspect, the present invention provides an intelligent printing layout method based on machine learning, including:

[0006] Obtain a set of requirement descriptions for the target finished product printing layout of the user; wherein, the set of requirement descriptions includes several requirement description information, and the requirement description information is used to describe the layout requirements of the user for the target finished product printing layout;

[0007] Extract the features of the description mode and perform feature conversion of the guiding value on the set of requirement descriptions to obtain a high-dimensional guiding feature distribution corresponding to the target finished product printing layout of the set of requirement descriptions;

[0008] Perform multiple data simulations on the high-dimensional guiding feature distribution according to a pre-trained intelligent printing layout model to obtain several original version printing layouts corresponding to the high-dimensional guiding feature distribution;

[0009] Construct a user feedback channel based on each of the original version printing layouts, and accept optimization suggestions from the user for each of the original version printing layouts through the user feedback channel;

[0010] Modify the high-dimensional guiding feature distribution according to the optimization suggestions, and perform data simulation on the modified high-dimensional guiding feature distribution according to a pre-trained intelligent printing layout model to obtain the target finished product printing layout.

[0011] In a second aspect, the present invention provides an intelligent printing layout device based on machine learning for implementing the intelligent printing layout method based on machine learning according to any one of the first aspects, including:

[0012] A requirement acquisition module for acquiring a set of requirement descriptions of the user for the target finished product printing layout; wherein, the set of requirement descriptions includes a number of requirement description information, and the requirement description information is used to describe the user's layout requirements for the target finished product printing layout;

[0013] A guiding analysis module for extracting the features of the description pattern and converting the features of the guiding value from the set of requirement descriptions to obtain the high-dimensional guiding feature distribution of the set of requirement descriptions corresponding to the target finished product printing layout;

[0014] A preliminary simulation module for performing multiple data simulations on the high-dimensional guiding feature distribution according to a pre-trained intelligent printing layout model to obtain a number of original version printing layouts corresponding to the high-dimensional guiding feature distribution;

[0015] An optimization feedback module for constructing a user feedback channel based on each of the original version printing layouts and receiving the user's optimization suggestions for each of the original version printing layouts through the user feedback channel;

[0016] A layout correction module for modifying the high-dimensional guiding feature distribution according to the optimization suggestions, and performing data simulation on the modified high-dimensional guiding feature distribution according to a pre-trained intelligent printing layout model to obtain the target finished product printing layout.

[0017] The present invention provides an intelligent printing layout method based on machine learning, which has the following beneficial effects:

[0018] The present invention collects a set of user demand descriptions for a target printed layout, extracts and transforms features from the set of demand descriptions to generate a high-dimensional guiding feature distribution to characterize the layout requirements, uses a pre-trained intelligent printed layout model to perform multiple data simulations on the high-dimensional feature distribution to generate multiple original versions of the printed layout, collects optimization suggestions from users for each original version of the layout through a user feedback channel, adjusts the high-dimensional feature distribution according to the feedback, and re-performs simulations through the intelligent model to generate the final target finished printed layout. By analyzing the description pattern of the set of demand descriptions provided by the user, the guiding value of the set of demand descriptions is analyzed, thereby driving the intelligent printed layout model to work, and solving the problem in the prior art that the intelligent layout efficiency is low due to the inability to accurately identify the customer's expectations. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 is a schematic diagram of the steps of an intelligent printed layout method based on machine learning provided by an embodiment of the present invention;

[0020] Figure 2 is a schematic diagram of the structure of an intelligent printed layout device based on machine learning provided by an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0021] In order to make the objectives, technical solutions and advantages of the present invention clearer and more understandable, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0022] The implementation of the present invention will be described in detail below with reference to specific embodiments.

[0023] Referring to Figure 1 、 Figure 2 shown, a preferred embodiment of the present invention is provided.

[0024] In a first aspect, the present invention provides an intelligent printed layout method based on machine learning, including:

[0025] S1: Obtain a set of user demand descriptions for a target finished printed layout; wherein, the set of demand descriptions includes a number of demand description information, and the demand description information is used to describe the user's layout requirements for the target finished printed layout;

[0026] S2: Extract features of the description pattern and perform guiding value feature transformation on the set of demand descriptions to obtain a high-dimensional guiding feature distribution corresponding to the set of demand descriptions for the target finished printed layout;

[0027] S3: Perform multiple data simulations on the high-dimensional guiding feature distribution according to a pre-trained intelligent printing layout model to obtain several original version printing layouts corresponding to the high-dimensional guiding feature distribution;

[0028] S4: Construct a user feedback channel based on each of the original version printing layouts, and accept optimization suggestions from users for each of the original version printing layouts through the user feedback channel;

[0029] S5: Correct the high-dimensional guiding feature distribution according to the optimization suggestions, and perform data simulation on the corrected high-dimensional guiding feature distribution according to a pre-trained intelligent printing layout model to obtain a target finished product printing layout.

[0030] Specifically, in step S1 of the embodiment provided by the present invention, a set of requirement descriptions of the user for the target finished product printing layout is obtained. The target finished product printing layout is the printing layout that the user wants to obtain. That is, the user provides corresponding requirement description information according to the expectation of the final product of the printing layout that the user wants to obtain. These requirement description information describe the layout requirements of the user for the target finished product printing layout, and multiple requirement description information together form a set of requirement descriptions.

[0031] More specifically, the requirement description information can be text or image. The expectations of the user conveyed by different types of requirement description information are different. Therefore, different parsing methods need to be used to obtain the expectations of the target finished product printing layout that the user actually wants to express.

[0032] It should be noted that the technical solution of the present invention uses a machine learning algorithm model pre-trained with a large amount of data to generate a corresponding printing layout for the expectations proposed by the user. The generation effect of the printing layout depends on two points. One is the training degree of the machine learning algorithm model, and the other is the accuracy of the user expectations input into the model.

[0033] Specifically, in step S2 of the embodiment provided by the present invention, in order to improve the accuracy of the user expectations input into the model, feature extraction of the description mode and feature transformation of the guiding value are performed on the set of requirement descriptions to obtain the high-dimensional guiding feature distribution of the set of requirement descriptions corresponding to the target finished product printing layout.

[0034] It can be understood that the purpose of this step is to perform feature extraction of the description mode on the set of requirement descriptions proposed by the user, and based on the extracted description mode, judge the analysis method for performing the guiding value analysis on the set of requirement descriptions itself. Then, use the obtained analysis method to perform the guiding value analysis on the set of requirement descriptions, and obtain the expectations of the user for the target finished product printing layout applied to the model generation direction of the pre-trained intelligent printing layout model, that is, the high-dimensional guiding feature distribution.

[0035] More specifically, the description pattern is the existence pattern of each requirement description information in the set of requirement descriptions provided by the user, including the proportion of the information types of each requirement description information and the emotional color of the wording used in each requirement description information. The guiding value is the model production direction applicable to the intelligent printing layout model actually reflected by the requirement description information provided by the user, that is, the guiding feature that guides the intelligent printing layout model to perform intelligent generation work.

[0036] More specifically, based on the proportion of the information types of each requirement description information, the conversion method for converting the requirement description information into the model generation direction of the intelligent printing layout model can be deduced. When the proportion of text and images in the requirement description set is different, the conversion methods for the text-type requirement description information and the image-type requirement description information are also different. This is because the final presentation method of the target finished product printing layout is the image method. Therefore, there are differences in the degree of fit between the text-type requirement description information and the image-type requirement description information and the target finished product printing layout. When the proportion of the text-type requirement description information and the image-type requirement description information is different, the specific degree of fit will also be affected. Therefore, it is necessary to adjust the conversion method according to the proportion.

[0037] More specifically, based on the emotional color of the wording used in each requirement description information, the conversion method for converting the requirement description information into the model generation direction of the intelligent printing layout model can be deduced. Each requirement description information itself contains an emotional color, and its emotional color can guide a side of the user's final presentation state of the target finished product printing layout. The requirement description information can be directed to multiple generation directions of the intelligent printing layout model. By selecting and determining the multiple generation directions that the requirement description information can be directed to based on the side obtained from the emotional color, the model generation direction that can accurately reflect the user's actual expectations and can be applied to the intelligent printing layout model can be obtained, that is, the high-dimensional guiding feature distribution.

[0038] Specifically, in step S3 of the embodiment provided by the present invention, multiple data simulations are performed on the high-dimensional guiding feature distribution according to the pre-trained intelligent printing layout model to obtain several original version print layouts corresponding to the high-dimensional guiding feature distribution. The intelligent printing layout model is a machine learning algorithm model pre-trained through a large amount of training data. Through the generation of the intelligent printing layout model, the original version print layout of the target finished product printing layout that meets the expectations can be quickly and efficiently obtained.

[0039] More specifically, the training data of the intelligent printing layout model is a large number of layout samples and the corresponding high-dimensional guiding feature distribution. Therefore, the intelligent printing layout model can produce several original version print layouts corresponding to the high-dimensional guiding feature distribution.

[0040] More specifically, the high-dimensional guiding feature distribution is the working guiding direction of the user's expectation for the printed layout of the target finished product implemented in the intelligent printing layout model, while the original version of the printed layout is the working feedback of the intelligent printing layout model corresponding to the high-dimensional guiding feature distribution. It should be noted that there are two types of several original versions of the printed layout. The first type is consistent in the general direction but deviated in details, and the second type is a different understanding of the high-dimensional guiding feature distribution in the general direction, that is, inconsistent in the general direction.

[0041] Specifically, in step S4 of the embodiment provided by the present invention, a user feedback channel is constructed based on each original version of the printed layout, and the user's optimization suggestions for each original version of the printed layout are received through the user feedback channel. The optimization suggestions received through the user feedback channel are used to determine the version of the several original versions of the printed layout that best meets the user's expectations, and subsequent optimizations are performed based on the user's feedback to further improve the printed layout, so as to obtain the printed layout of the target finished product.

[0042] It should be noted that the user feedback channel can have various forms, including the basic selection form, that is, for the user to interactively select the original version of the printed layout that best meets the expectations. The user feedback channel can also be a description form consistent with the requirement description set, that is, for the user to interactively provide optimization suggestions in the form of text and images, and further obtain the expectations for the printed layout of the target finished product reflected by the user's identification information through the analysis of the optimization suggestions.

[0043] Specifically, in step S5 of the embodiment provided by the present invention, the high-dimensional guiding feature distribution is corrected according to the optimization suggestions, and data simulation is performed on the corrected high-dimensional guiding feature distribution according to the pre-trained intelligent printing layout model to obtain the printed layout of the target finished product.

[0044] It can be understood that after obtaining the optimization suggestions, the high-dimensional guiding feature distribution is corrected to obtain a high-dimensional guiding feature distribution that better meets the user's expectations. Data simulation is performed on the corrected high-dimensional guiding feature distribution according to the pre-trained intelligent printing layout model. After that, the steps of obtaining the optimization suggestions and correcting the high-dimensional guiding feature distribution according to the optimization suggestions can be repeated, and this process is continuously cycled until the printed layout of the target finished product is obtained.

[0045] The present invention provides an intelligent printing layout method based on machine learning, which has the following beneficial effects:

[0046] The present invention collects a set of requirement descriptions of users for a target printing layout, extracts and transforms features of the set of requirement descriptions, generates a high-dimensional guiding feature distribution to characterize the layout requirements, uses a pre-trained intelligent printing layout model to perform multiple data simulations on the high-dimensional feature distribution, generates multiple original versions of the printing layout, collects optimization suggestions of users for each original version of the layout through a user feedback channel, adjusts the high-dimensional feature distribution according to the feedback, and re-performs simulations through the intelligent model to generate a final target finished printing layout. By analyzing the description mode of the set of requirement descriptions provided by users, the present invention analyzes the guiding value of the set of requirement descriptions, and further drives the intelligent printing layout model to work, solving the problem in the prior art that the intelligent layout efficiency is low due to the inability to accurately identify the customer's expectations.

[0047] Preferably, the steps of obtaining the set of requirement descriptions of users for the target finished printing layout include:

[0048] S11: Receive requirement description information corresponding to the target finished printing layout provided by the user through an information interaction port;

[0049] S12: Evaluate the perfection and rationality of the requirement description information provided by the user, and mark loopholes in the requirement description information according to the evaluation results to obtain a correction mark based on the requirement description information;

[0050] S13: Receive information correction processing of the user corresponding to the correction mark through the information interaction port to correct the description of the loopholes in the requirement description information and obtain the final version of the requirement description information;

[0051] S14: Combine all the final versions of the requirement description information provided by the user through the information interaction port to obtain the set of requirement descriptions.

[0052] Specifically, the system provides an information interaction port for the user to input layout requirement description information. The user inputs requirement descriptions through this port. It can be understood that the user has an overall conceptual expectation for the target finished printing layout, but cannot specifically describe the specific layout or construct all the details. Therefore, the requirement description information needs to be parsed and model-generated subsequently.

[0053] More specifically, evaluate the perfection and rationality of the requirement description information provided by the user, and mark loopholes in the requirement description information according to the evaluation results to obtain a correction mark based on the requirement description information. Use technologies such as natural language processing (NLP) and rule verification to analyze the requirement description information input by the user, evaluate the integrity and rationality of the information, mark the discovered deficiencies or contradictions, and generate a correction mark.

[0054] More specifically, receive the information correction process corresponding to the correction mark from the user through the information interaction port to correct the description of the vulnerabilities in the requirement description information, obtain the requirement description information of the final version, and feedback the evaluation result (correction mark) to the user to prompt the content that needs to be supplemented or modified. The user submits the corrected requirement description information again through the information interaction port, and the system evaluates the corrected information again to ensure that all vulnerabilities and deficiencies have been corrected.

[0055] More specifically, integrate the requirement description information of the final version submitted by the user multiple times to generate a complete requirement description set, which contains all the user requirement details, provide a comprehensive requirement description set, and provide a complete and accurate reference for the subsequent layout design to ensure that all user requirements are recorded and considered by the system, reducing the possibility of omission or misunderstanding.

[0056] Preferably, the steps of extracting the feature of the description mode and converting the feature of the guiding value for the requirement description set to obtain the high-dimensional guiding feature distribution corresponding to the target finished product printing layout for the requirement description set include:

[0057] S21: Extract the feature of the description mode for the requirement description set to obtain the description mode feature of the requirement description set;

[0058] S22: Convert the feature of the description mode feature of the requirement description set into the feature of the guiding value to obtain the high-dimensional guiding feature distribution corresponding to the target finished product printing layout for the requirement description set.

[0059] Specifically, extract the feature of the description mode for each requirement description information in the requirement description set, including the proportion of the information type of each requirement description information in the requirement description set and the emotional color of each requirement description information, and generate the description mode feature of the requirement description set based on the proportion of the information type of each requirement description information in the requirement description set and the emotional color of each requirement description information.

[0060] More specifically, the requirement description set is the expectation of the user for the target finished product printing layout, and the description mode feature is the combination of requirement descriptions, that is, the form in which the user puts forward the expectation. Based on this form, the way to interpret the requirement description set, that is, the expectation put forward by the user, can be obtained. Based on this interpretation method, the requirement description set can be interpreted to obtain the high-dimensional guiding feature distribution of the guiding value of the feedback requirement description set.

[0061] More specifically, in the technical solution of the present invention, a machine learning model is applied to generate the target finished product printing layout. Therefore, it is necessary to input the guiding feature value into the machine learning model to provide the generation direction for the model, and this generation direction is the high-dimensional guiding feature distribution obtained in the above steps.

[0062] Preferably, the steps of extracting the feature of the description pattern from the requirement description set to obtain the description pattern feature of the requirement description set include:

[0063] S211: Analyze the information characteristics of each requirement description information in the requirement description set in multiple dimensions to obtain the type characteristics and emotional characteristics of each requirement description information in the requirement description set; wherein, the type characteristics include text type and image type;

[0064] S212: Conduct an overview analysis of the type distribution of the requirement description set according to the type characteristics of each requirement description information in the requirement description set to obtain the description set type distribution of the requirement description set;

[0065] S213: Conduct an analysis of the overall emotional tendency of the requirement description set according to the emotional characteristics of each requirement description information in the requirement description set to obtain the description set emotional tendency of the requirement description set;

[0066] S214: The description set type distribution and the description set emotional tendency together constitute the description pattern feature of the requirement description set.

[0067] Specifically, analyze each requirement description in the requirement description set and extract different information dimensions: type characteristics: According to the content of the description, judge the type of each requirement description information, and distinguish it into text type and image type; emotional characteristics: Analyze the emotional tendency in the requirement description, judge whether the description content is positive, negative or neutral, and the degree of conformity with various design styles.

[0068] More specifically, use an emotion analysis model (such as an emotion dictionary-based model, a deep learning model, etc.) to classify the emotion in each requirement description, and give the emotional tendency of each description. The multi-dimensional information characteristic analysis makes the understanding of the requirement description more comprehensive, not limited to the literal meaning, but also involving deeper information such as emotion and category. Emotion analysis enables the system to understand the emotional direction of the user's requirements, which has guiding significance for the selection of design layout styles (such as simplicity, luxury, creativity, etc.).

[0069] More specifically, according to the type characteristics (text type and image type) of each requirement description information, the entire set of requirement descriptions is statistically analyzed, and the proportion of each type of requirement description in the set is analyzed to generate a type distribution map. For example, to determine the ratio of text type and image type in the requirement description set, frequency analysis or classification algorithms (such as K-means clustering, etc.) can be used to identify the distribution of various requirements in the set, providing data support for subsequent layout design, understanding the user's demand preferences for text and images, and thus being able to arrange the key points of the layout targeted. Type distribution analysis can help the system judge which design elements (text or image) dominate in the requirements, helping designers or the system to prioritize the corresponding design elements.

[0070] More specifically, analyze the emotional characteristics of each requirement description information, give the emotional tendency of each requirement, summarize the emotional tendencies of all requirement descriptions, conduct an overall emotional tendency analysis, and judge the overall emotional trend of the requirement set. For example, calculate the proportions of positive emotions, negative emotions, and neutral emotions in the requirement set, and use emotional tendency scores (such as emotional intensity, emotional labels, etc.) for comprehensive analysis to generate an emotional tendency distribution map. Overall emotional tendency analysis helps to identify the user's emotional expectations for the design finished product (such as tending to be simple and modern, or complex and luxurious), which is of great significance for determining the final layout design style. Emotional analysis can help identify potential negative emotional requirements and adjust the design direction to avoid design outputs that do not meet user expectations.

[0071] More specifically, fuse the analysis results of the two aspects of type distribution (the ratio of text type and image type) and emotional tendency (the distribution of positive, negative, and neutral emotions) to generate a description pattern feature. Finally, merge these features into a high-dimensional feature representation to form the description pattern feature of the requirement description set for subsequent layout design or optimization.

[0072] Preferably, the steps of performing a feature transformation of guiding value on the description pattern feature of the requirement description set to obtain the high-dimensional guiding feature distribution corresponding to the target finished product printing layout of the requirement description set include:

[0073] S221: Perform a meaning parsing and pointing analysis on the requirement description set according to the type distribution of the description set in the description pattern feature to obtain the first meaning parsing vector of each requirement description information in the requirement description set;

[0074] S222: Perform a meaning parsing and pointing analysis on the requirement description set according to the emotional tendency of the description set in the description pattern feature to obtain the second meaning parsing vector of each requirement description information in the requirement description set;

[0075] S223: Based on the first meaning analysis vectors and the second meaning analysis vectors corresponding to each of the requirement description information in the requirement description set, perform a guiding value analysis on each of the requirement description information to obtain the benchmark guiding value features of each of the requirement description information;

[0076] S224: Based on a preset association database, perform an analysis on the relevance expansion of the potential guiding value of the distribution of the description set types and the sentiment tendency of the description set in the description mode features to obtain the expanded guiding value features of the description simulation features;

[0077] S225: Analyze the degree of fit between the benchmark guiding value features and the expanded guiding value features, and select the expanded guiding value features according to the analysis results to obtain the additional guiding value features whose degree of fit meets the predetermined standard in the expanded guiding value features;

[0078] S226: Perform feature combination and weight allocation on the benchmark guiding value features and the additional guiding value features to obtain the high-dimensional guiding feature distribution of the target finished product printing layout.

[0079] Specifically, according to the distribution of the description set types (the ratio of the text type and the image type) extracted in the previous step, classify each requirement description information in the requirement description set, and for each requirement description, perform meaning analysis in combination with its type (text or image) to identify the specific design elements and their functions pointed to by the requirement description.

[0080] More specifically, for the text type, analyze the semantics of the text to identify the key design requirements of the description (such as text size, font, layout method, etc.), and for the image type, analyze the content of the image description to determine its relevant design elements (such as the size, position, style, etc. of the image). Based on the analysis results of the distribution of the description set types, generate the first meaning analysis vector of each requirement description information, and this vector contains the design requirements and objectives related to the description content.

[0081] More specifically, by classifying the requirement description information through the type distribution and performing meaning analysis, the specific design requirements of each requirement can be captured more accurately, the analysis efficiency can be improved, and it can be ensured that the information of each requirement description can effectively point to its corresponding design objectives and elements.

[0082] More specifically, according to the sentiment tendency of the description set in the description mode features, perform meaning parsing and pointing analysis on the requirement description set. According to the sentiment tendency distribution in the description set, perform sentiment analysis on each requirement description in the requirement description set to determine the sentiment type (positive, negative, or neutral) of the requirement description. Based on the analysis results of the sentiment tendency, perform sentiment-driven meaning parsing on each requirement description. For example, positive sentiment may imply the need for simplicity and aesthetics, while negative sentiment may indicate criticism of complex or unexpected design elements. According to the sentiment tendency of each requirement description, generate a second meaning parsing vector related to the sentiment feature, which represents the sentiment goal and sentiment-oriented design requirements in the requirement description.

[0083] More specifically, the meaning parsing combined with sentiment characteristics can help the system identify the user's emotional expectations, enabling the design to better meet the user's emotional needs (such as a preference for a simple or complex design style), providing guidance at the emotional level for subsequent design decisions, and ensuring that the design work can resonate with the user emotionally.

[0084] More specifically, perform guiding value parsing on the requirement description information according to the first meaning parsing vector and the second meaning parsing vector. According to the first meaning parsing vector (type pointing) and the second meaning parsing vector (sentiment pointing), perform guiding value parsing on each requirement description information. Guiding value analysis focuses on the impact degree and direction of each requirement description on the target finished product, which is usually closely related to factors such as the practicality, emotional appeal, and creativity of the design. Generate a benchmark guiding value feature for each requirement description information, which represents the contribution degree of this requirement to the target design layout. Through guiding value parsing, the system can quantify the influence of each requirement description on the target design, provide a quantitative basis for subsequent optimization, improve the matching degree between the requirements and the design finished product, and ensure that the design can effectively respond to the actual value demands in the requirement description.

[0085] More specifically, based on a preset association database, perform an analysis on the potential guiding value relevance expansion of the type distribution and sentiment tendency distribution in the description set of the description mode features. Use the preset association database to perform an analysis on the potential guiding value relevance expansion. This database contains a large amount of historical data, which describes how requirements of different types and sentiment tendencies are related to specific design elements or layout effects. Through the historical data model in the database, analyze the potential association between the type distribution and sentiment tendency in the description set, and expand possible design solutions and optimization directions. Through relevance analysis, obtain a set of potential expanded guiding value features, which represent possible design optimization directions or additional values under the current description mode.

[0086] More specifically, by using the associated database to expand the description pattern features, potential design optimization space can be discovered, the innovation and practicality of the design can be enhanced, and expanding the guiding value features provides more support for feasibility analysis in the design process, which helps to avoid being limited to the initial requirement scope.

[0087] More specifically, perform a conformity analysis on the benchmark guiding value features and the expanded guiding value features, and select the expanded guiding value features according to the analysis results. Evaluate the conformity of the benchmark guiding value features and the expanded guiding value features through numerical analysis (such as cosine similarity, correlation coefficient, etc.). The higher the conformity, the more the expanded guiding value feature meets the actual requirements. Select the additional guiding value features that meet the predetermined criteria: According to the results of the conformity analysis, select those expanded guiding value features with a conformity higher than the predetermined criteria as the additional guiding value features.

[0088] More specifically, the conformity analysis helps to screen out the guiding value features that best meet the requirements, ensuring that the final effect of the design is maximally matched with the requirements. By reasonably selecting the additional guiding value features, it is possible to avoid the design deviating too much from the actual requirements and ensure that the design optimization meets the user's expectations.

[0089] More specifically, perform feature combination and weight assignment on the benchmark guiding value features and the additional guiding value features to obtain the high-dimensional guiding feature distribution of the target finished product printing layout

[0090] More specifically, merge the benchmark guiding value features and the additional guiding value features to form a unified feature vector, and perform weight assignment according to the importance of each feature. The weight assignment can be based on factors such as the specific content of the requirement description, the priority of market demand, and the actual feasibility of the design. The finally obtained feature combination forms the high-dimensional guiding feature distribution of the target finished product printing layout, which can accurately reflect the design requirements, emotional appeals, and actual layout requirements of the requirement description set.

[0091] More specifically, the feature combination and weight assignment ensure the comprehensiveness and pertinence of the target finished product printing layout design, enabling the design decision to be adjusted according to the actual requirements and emotional preferences of the user. The high-dimensional guiding feature distribution provides precise guidance for the printing layout, ensuring that the final design not only meets the technical requirements but also maximally satisfies the user's emotional needs and design goals.

[0092] Preferably, the step of obtaining the first meaning analysis vector of each requirement description information in the requirement description set by performing meaning analysis and pointing analysis on the requirement description set according to the description set type distribution in the description pattern features includes:

[0093] S2211: Perform proportion analysis and weight conversion of text type and image type on the described set type distribution, so as to obtain the printing layout image analysis weights of the text type requirement description information and the image type requirement description information in the requirement description set corresponding to the described set type distribution;

[0094] S2212: Perform property analysis of the overall meaning on the text type requirement description information and the image type requirement description information in the requirement description set, so as to obtain the proportion of the overall meaning of the text type requirement description information and the image type requirement description information in the requirement description set for the target finished product printing layout;

[0095] S2213: Based on the printing layout image analysis weights of the text type requirement description information and the image type requirement description information in the requirement description set corresponding to the described set type distribution, and the proportion of the overall meaning of the text type requirement description information and the image type requirement description information in the requirement description set for the target finished product printing layout, perform analysis on the layout image conversion form of each requirement description information for the target finished product printing layout, so as to obtain the layout image conversion form of each requirement description information, and use it as the first meaning analysis vector of each requirement description information.

[0096] Specifically, first, it is necessary to analyze the set type distribution in the requirement description set, specifically identify whether each requirement description is of text type or image type, and through the proportion analysis of text type and image type, convert it into the printing layout image analysis weights of different types of descriptions. The weight here mainly represents the importance of each type of description information in the target layout image, and is usually set based on the priority of design requirements and user preferences. For example, if the description requirement of the image is more important, the weight of the image type is higher. Assuming that the proportion of the text type is 70% and the proportion of the image type is 30%, the printing layout image analysis weights can be calculated according to these two proportions, and methods such as weighted average method may be used.

[0097] More specifically, by analyzing the proportion of text and image types, the importance of different types of descriptions in the design process can be quantified, ensuring that reasonable weights are given to different types of requirements during the layout process. The weight conversion provides a more accurate basis for the subsequent steps, enabling the system to automatically adjust the priority of each element in the layout design.

[0098] More specifically, perform an overall semantic property analysis on the text-type and image-type requirement description information in the requirement description set. Based on the text-type and image-type information in the requirement description set, perform an overall semantic property analysis. This step mainly conducts semantic analysis on each type of requirement description (text and image) to extract its overall semantic impact on the printed layout design of the target finished product.

[0099] More specifically, for text-type requirements, analyze their impacts on aspects such as text layout, font selection, line spacing, etc. For image-type requirements, analyze how aspects such as the size, position, and style of the image affect the layout design. According to the property analysis of each requirement description, calculate the proportion of the impacts of text-type and image-type on the overall design, forming the meaning proportion of text-type and the meaning proportion of image-type. This analysis can help identify the design intentions of each element in the requirement description information, ensure that the design can accurately reflect these intentions. By calculating the meaning proportion, it can provide a basis for subsequent layout image conversion and help the system better understand the comprehensive design requirements of different descriptions.

[0100] More specifically, combine the printing layout image analysis weight and the overall meaning proportion in the above steps to analyze the layout image conversion form for each requirement description. Specifically, the conversion form of text-type requirements may be manifested as impacts on the layout structure (such as paragraphs, text size, font style, etc.), while image-type requirements may involve aspects such as the layout, size, and alignment of the image. Through the weight and meaning proportion, convert the design goals of each requirement description information into corresponding layout image elements, that is, adjust the elements of the layout form (such as text wrapping, the coordination method between image and text, etc.) through weight weighting and meaning proportion. Based on the above analysis, generate a layout image conversion form for each requirement description. This form serves as the first semantic analysis vector of this requirement description information. These vectors contain the degree and direction of the impact on the printed layout design of the target finished product. This step can transform abstract requirement descriptions (such as "the text should appear concise and modern" or "the image needs to highlight the visual focus") into specific layout design elements and formats, providing clear guidance for the final design layout. The generated first semantic analysis vector provides content-based multi-dimensional guidance for layout image conversion, making the layout design more in line with the requirement description.

[0101] It is understandable that by conducting a detailed proportional analysis and weight conversion of the description set type distribution, combined with the meaning analysis of the requirement description, it is possible to accurately grasp the weight and influence of each requirement description in the typesetting design. This process converts the requirement description information into specific typesetting image design requirements, and provides detailed and precise guidance for subsequent typesetting image conversion, so that the typesetting design of the target product is more in line with actual needs. By combining the analysis of text and image types, it can not only meet the needs functionally, but also respond to users' expectations of typesetting style from an emotional level, thereby improving the overall quality of the final design.

[0102] Preferably, the step of performing meaning parsing and pointing analysis on the requirement description set according to the description set sentiment tendency in the description pattern feature to obtain a second meaning parsing vector of each requirement description information in the requirement description set includes:

[0103] S2221: generating key description tags for target finished product printing and typesetting based on the description set sentiment tendency, so as to obtain a plurality of key description tags for the description set sentiment tendency for target finished product printing and typesetting;

[0104] S2222: Perform label pointing analysis on each requirement description information in the requirement description set according to each of the key description labels to obtain a second meaning analysis vector of each of the requirement description information.

[0105] Specifically, first, we need to conduct a comprehensive analysis of the sentiment tendency in the demand description set. Sentiment tendency refers to the emotional attitude expressed by each demand description, such as positive, negative, neutral, etc. This analysis usually relies on natural language processing technology (such as sentiment analysis models) to determine the emotional color of the text.

[0106] More specifically, based on sentiment tendency analysis, several key descriptive tags are generated for the design requirements of the target finished product printing layout. These tags may include a combination of emotions and design requirements, such as "strong sense of modernity", "warm and friendly", "simple and bright", etc. A mapping relationship is established between the sentiment tendency distribution and these key descriptive tags, so that each label corresponds to a certain emotion and design direction, which helps to better express the design intention. Through in-depth analysis of sentiment tendencies, the potential emotional colors in the demand description can be identified, providing emotional guidance for subsequent design. The generated key descriptive tags help the system to clarify the emotional orientation of each design element (such as layout, font, color tone, etc.) in subsequent steps, ensuring that the design can meet the expected emotional expression in terms of visual effects.

[0107] More specifically, for each requirement description information, perform label pointing analysis of the sentiment tendency. Specifically, the system needs to check the sentiment content of each requirement description and match it with the generated key description labels. For example, a certain requirement description may be "The font needs to be simple and elegant". If the sentiment tendency of this description is analyzed as "modern and simple", the matching key label may be "simple and lively". Through label pointing analysis, each requirement description information is matched with the corresponding sentiment tendency label, thereby generating the second meaning analysis vector of each requirement description. This vector will contain the sentiment orientation of the description information and its association with the design goal.

[0108] It can be understood that through label pointing analysis, the sentiment tendency of the requirement description can be further refined, ensuring that the design solution not only meets the functional requirements but also accurately conveys the required emotional effect. The generated second meaning analysis vector can provide richer emotional information for the layout design, enabling designers to accurately control the emotional atmosphere of the design during actual layout, such as gentle, lively, solemn, etc.

[0109] More specifically, according to the analysis of the sentiment tendency, adjust the layout elements to reflect the emotional needs of the target finished product's printed layout. For example, if a certain label points to "simple and lively", the layout design may adopt clear and simple fonts and graphic elements; if the label points to "warm and friendly", the design may adopt soft colors, rounded fonts, and a relatively warm image style. Through the second meaning analysis vector, the orientation of layout elements such as text, images, color matching, and layout in emotional expression can be determined, and design decisions can be optimized during the layout process. The close combination of sentiment tendency and layout design ensures that the final design result not only meets the requirements in form but also can resonate with users or the target audience at the emotional level. The second meaning analysis vector enables the design to maintain consistency in terms of emotion and vision, ultimately achieving a layout effect that better meets the emotional needs of the target market or target group.

[0110] It can be understood that through sentiment tendency analysis and the generation of key description labels, the emotional color of each requirement description can be clarified, and these emotional needs can be transformed into specific design elements through label pointing analysis. This not only improves the accuracy of the design but also ensures that the design result can achieve the expected effect at the emotional level. The generated second meaning analysis vector, as a multi-dimensional emotional information carrier, enables the design to accurately convey emotional intentions, enhancing the personalization and accuracy of the layout design. Overall, the combination of sentiment tendency and design requirements improves the emotional compliance of the design result, making the final layout design not only meet the functional requirements but also effectively convey the required emotional atmosphere.

[0111] It should be noted that the analysis of sentiment tendency is not limited to sentiment dimensions such as positive and negative, but also includes tendencies in more sentiment dimensions such as simplicity and complexity. At the same time, it should be noted that the analysis of sentiment dimensions is not only limited to the analysis of the sentiment tendency of words themselves, but also includes the analysis of the sentiment tendency of word attributes, that is, the latent sentiment tendency of the act of using a certain word. Further, the sentiment tendency is not limited to a single word, but also includes the holistic analysis of multiple words.

[0112] Preferably, the pre-training steps of the intelligent printing layout model include:

[0113] S31: Collect printing layout samples to construct a historical data set; wherein, the printing layout samples have corresponding sample construction feature distributions, and the sample construction feature distributions correspond to the high-dimensional guiding feature distributions;

[0114] S32: Divide the historical data set to obtain a training data set, a validation data set, and a test data set;

[0115] S33: Construct a convolutional neural network model composed of a convolutional layer, an activation function layer, a pooling layer, a fully connected layer, and an output layer as the intelligent printing layout model to be trained;

[0116] S34: Substitute the training data set into the convolutional neural network model, so that the convolutional neural network model learns through the backpropagation algorithm according to the training data set, obtains the mapping relationship between the printing layout samples and the sample construction feature distributions, and optimizes the model of the convolutional neural network model according to the mapping relationship;

[0117] S35: During the model training of the convolutional neural network model using the training data set, substitute the validation data set into the convolutional neural network model, and make the convolutional neural network model perform validation processing on the validation data set to evaluate the model performance of the convolutional neural network model during the training process, and optimize the model hyperparameters according to the evaluation results; wherein, the model hyperparameters include the learning rate and the regularization parameter;

[0118] S36: When the model training of the training data set is completed, substitute the test data set into the convolutional neural network model, and make the convolutional neural network model perform test processing on the test data set to finally evaluate the model performance of the convolutional neural network model to determine whether to end the model training.

[0119] Specifically, first, a large number of printed layout samples need to be collected. These samples should cover various layout styles, design elements (such as fonts, layout structures, text and image mixing, etc.), and different design requirements. Each sample should have a corresponding sample construction feature distribution, which includes all key design elements in printed layout (such as text layout, font selection, image size and layout, color usage, etc.). The feature distribution of these layout samples will correspond to the high-dimensional guiding feature distribution, which means that the layout samples will be mapped into the target space for model learning. The output of the model will be compared with these features for optimization. By collecting and organizing layout samples, the model can obtain rich input data, which provides diverse instances and design elements for subsequent training. This correspondence between the sample feature distribution and the high-dimensional guiding features enables the model to understand and capture the diversity and complexity of layout design during the training process.

[0120] More specifically, the collected printed layout samples are divided into three different subsets: the training dataset, the validation dataset, and the test dataset. Usually, the training set accounts for about 70 - 80%, the validation set accounts for 10 - 15%, and the test set accounts for 10 - 15%. The training set is used to train the model, the validation set is used to evaluate the model performance and adjust hyperparameters, and the test set is used for the final model evaluation to ensure that the model has good generalization ability. The division of the dataset ensures that the model can be evaluated and optimized in different ways at different stages, thereby improving the model's generalization ability. A reasonable division of the training set, validation set, and test set helps to avoid overfitting problems and ensure the performance of the model on unseen data.

[0121] More specifically, convolutional layer: Extract local features of the input layout image through the convolutional layer. These features can reflect basic information such as texture and edges in the image. Activation function layer: Usually use the ReLU (Rectified Linear Unit) activation function to introduce non-linearity and increase the model's expressive ability. Pooling layer: The pooling layer (such as the max pooling layer) is used to reduce the spatial dimension of the feature map while retaining important feature information and reducing the computational amount. Fully connected layer: Map the pooled features to the final output space through the fully connected layer, enabling the model to perform complex feature combinations. Output layer: The output layer is usually a softmax layer for classification tasks, outputting a probability distribution corresponding to the sample category (or design element category). Through hierarchical feature learning, the convolutional neural network can gradually extract key information from printed layout samples from low-level to high-level. This hierarchical structure enables the model to process image features at different scales and ultimately learn how to perform effective layout design.

[0122] More specifically, the training dataset is input into the constructed convolutional neural network model. The convolutional neural network learns through the backpropagation algorithm. According to the error between the output and the target label, the parameters of each layer in the network are adjusted. The model learns the mapping relationship between the printed layout samples and the feature distribution of the samples through training, that is, it learns how to extract features from the input layout samples and generate corresponding layout designs. According to the results of backpropagation, the weight parameters of the network are adjusted to gradually optimize the model. Through backpropagation and gradient descent, the model can efficiently learn the features of the layout design and continuously adjust during the training process, making the output of the model closer and closer to the target design requirements. By continuously optimizing the model, the accuracy and generalization ability of the model are improved, and it can better adapt to various design requirements.

[0123] More specifically, during the model training process, the validation dataset is substituted into the convolutional neural network model. For each iterative training, the validation dataset is used to evaluate the performance of the model and adjust the hyperparameters (such as the learning rate, regularization parameter, etc.) according to the performance of the model on the validation set. For example, by adjusting the learning rate, the learning rate of the model is controlled to avoid too fast or too slow convergence; the regularization parameter is used to control the model complexity to prevent overfitting. The use of the validation set ensures that the model will not overfit the training data during the training process, and at the same time, the hyperparameters can be continuously optimized according to the performance on the validation set to improve the model performance. Appropriate hyperparameter tuning can significantly improve the learning efficiency and final performance of the model.

[0124] More specifically, when the model training is completed, the test dataset is substituted into the convolutional neural network model for testing. By evaluating the test set, the performance of the model on unseen data is checked to ensure that it can be effectively generalized to practical applications. If the test results show that the model performance meets the expected standard, the training can be ended. If not, it may be necessary to return and adjust the training set, hyperparameters, or model structure. Through the final evaluation on the test dataset, it can be judged whether the model has sufficient generalization ability and practical application value. The evaluation of the test set is a key step in the model training process, and it can effectively verify every decision in the training process.

[0125] It can be understood that through the training process of the convolutional neural network (CNN), the model can automatically extract the features in the layout samples and perform effective optimization. The final convolutional neural network model has strong adaptability and can provide high-quality design suggestions in different printed layout tasks. Reasonable dataset division, backpropagation optimization, and hyperparameter tuning enable the model to effectively avoid overfitting and ensure good generalization ability. Through the cyclic evaluation and optimization of the training, validation, and test datasets, the performance stability of the model is ensured, and its ability to adapt to different design scenarios is verified.

[0126] Preferably, the step of correcting the high-dimensional guiding feature distribution according to the optimization suggestion includes:

[0127] S51: Analyze the direct correction features of the optimization suggestion to obtain the direct correction features based on the optimization suggestion;

[0128] S52: Analyze the sentiment tendency of the optimization suggestion to obtain the potential correction features reflected by the optimization suggestion, and perform a correlation analysis on the potential correction features according to the direct correction features to obtain the correlation analysis result of the potential correction features and the direct correction features;

[0129] S53: Select the potential correction features according to the correlation analysis result to obtain additional correction features;

[0130] S54: Assign weights to the direct correction features and the additional correction features, and correct the high-dimensional guiding feature distribution based on the direct correction features and the additional correction features after weight assignment.

[0131] Specifically, the optimization suggestion is analyzed step by step to identify the direct correction features therein. These features are usually elements directly affecting the typesetting design effect extracted from actual feedback. For example, a certain suggestion may involve directly adjustable typesetting attributes such as the font size, line spacing, and paragraph alignment of the text. The specific adjustment items are extracted from the optimization suggestion. These items are usually specific and clear and can directly affect the visual effect of the typesetting sample. The direct correction features provide a clear guiding direction for subsequent feature correction, ensuring that the model optimization process targets actual needs and has practical operability. This parsing process enables the model to accurately extract the key features that need to be corrected from the optimization suggestion without relying too much on guesswork or irrelevant features.

[0132] More specifically, perform a sentiment tendency analysis on the optimization suggestion to identify the potential sentiment information therein, such as whether the optimization suggestion favors a certain typesetting style (such as modernity, simplicity, etc.) or expresses a certain design concept (such as high-end feeling, relaxed feeling, etc.). Through sentiment tendency analysis, the potential correction features in the optimization suggestion can be obtained. These features may not be intuitive modification items, but they reflect the design intention or sensory needs behind the optimization suggestion. Example: If an optimization suggestion mentions "hoping to increase the modernity of the typesetting", then the sentiment tendency analysis may identify the design goal of "modernity" and associate some potential correction features with it, such as font selection, typesetting structure, color matching, etc.

[0133] More specifically, through sentiment analysis, the model can understand the implicit intentions and design feelings behind the optimization suggestions, so as to not only correct the surface features, but also optimize the overall style and emotional experience of the layout design. This kind of analysis ensures that the model can better meet the emotional needs of users when optimizing the layout, enhancing the affinity of the design and the user experience.

[0134] More specifically, a correlation analysis is carried out on the direct correction features extracted from the optimization suggestions and the potential correction features obtained through sentiment analysis. By calculating the correlation between these two types of features, the strength and relationship of their interaction during the correction process are determined. Statistical methods (such as Pearson correlation coefficient, cosine similarity, etc.) or machine learning methods can be used to measure the relationship between the features. For example, the correlation between the direct correction feature "font size" and the potential correction feature "modern sense" may indicate that the adjustment of the font size has a greater impact on conveying the modern sense. Through the correlation analysis, it is found which potential correction features have a strong correlation with the adjustment of the direct correction features. This helps to more accurately select and optimize features during the correction process. The correlation analysis can reveal the mutual relationship between different features, thereby guiding the prioritization and importance ranking of the features, ensuring that the correction process is both accurate and efficient. This kind of analysis ensures the rationality of the correction process, avoids redundant adjustments, and optimizes the overall strategy of feature correction.

[0135] More specifically, according to the results of the correlation analysis, those features with a high correlation with the direct correction features are selected from the potential correction features. These features will be used as additional correction features to further improve the correction plan. The additional correction features may include detailed adjustments related to emotions, design styles, layout structures, etc. These adjustments can not only correct the technical problems of the layout itself, but also further enhance the overall design effect. If the direct correction feature includes "paragraph spacing", and the potential correction feature "simplicity" has a high correlation with "paragraph spacing", then the additional correction feature of "simplicity" may be reflected by further adjusting the paragraph spacing.

[0136] More specifically, by selecting the additional correction features, the correction process becomes more comprehensive, not only limited to the directly visible correction features, but also able to penetrate into aspects such as design styles and emotional expressions, making the layout design more delicate and meeting the user's needs. This selection process improves the intelligence level of the correction plan, enabling the model to better understand the user feedback and conduct multi-dimensional optimizations.

[0137] More specifically, according to the contribution of each correction feature to the correction of the high-dimensional guiding feature distribution, weights are assigned to the direct correction features and the additional correction features. These weights may be determined based on factors such as the results of correlation analysis, the priority of design intent, and the urgency of optimization suggestions. During the actual optimization process, these weights may need to be dynamically adjusted to ensure the best correction effect. For example, if the direct correction feature has a greater impact on the typesetting effect, a higher weight can be given, while the additional correction features are appropriately weighted according to their impact on the overall effect.

[0138] More specifically, weight assignment ensures the reasonable role of different features in the correction process, avoids excessive or insufficient correction, and achieves the best correction effect. Dynamic weight adjustment makes the model more flexible and can optimize the accuracy and effect of feature correction according to different typesetting tasks or user requirements.

[0139] More specifically, based on the previously determined direct correction features, additional correction features, and their weights, the high-dimensional guiding feature distribution is actually corrected. This step usually involves modifying the input features of the model, adjusting the distribution of feature vectors to make it more in line with the optimized design goal, and applying the corrected features to the input of the model for prediction and output. The corrected feature distribution will guide the model to generate a typesetting result that is more in line with the optimization suggestions. This step ensures that the correction of the high-dimensional guiding feature distribution can directly affect the output of the model, ultimately improving the quality and effect of the typesetting design. Through the precise correction of the feature distribution, the model can generate a design output that is more in line with user requirements, improving the accuracy and personalization level of the design.

[0140] Refer to Figure 2 As shown, in a second aspect, the present invention provides an intelligent printing typesetting device based on machine learning for implementing the intelligent printing typesetting method according to any one of the first aspects, including:

[0141] A requirement acquisition module for acquiring a set of requirement descriptions of the user for the target finished product printing typesetting; wherein, the set of requirement descriptions includes several requirement description information, and the requirement description information is used to describe the typesetting requirements of the user for the target finished product printing typesetting;

[0142] A guiding analysis module for performing feature extraction of the description pattern and feature transformation of the guiding value on the set of requirement descriptions to obtain the high-dimensional guiding feature distribution corresponding to the target finished product printing typesetting;

[0143] A preliminary simulation module for performing multiple data simulations on the high-dimensional guiding feature distribution according to a pre-trained intelligent printing typesetting model to obtain several original version printing typesettings corresponding to the high-dimensional guiding feature distribution;

[0144] An optimization feedback module, configured to build a user feedback channel based on the printing layout of each of the original versions, and receive optimization suggestions from users for the printing layout of each of the original versions through the user feedback channel;

[0145] A typesetting correction module, configured to correct the high-dimensional guiding feature distribution according to the optimization suggestions, and perform data simulation on the corrected high-dimensional guiding feature distribution according to a pre-trained intelligent printing layout model to obtain a target finished printing layout.

[0146] In this embodiment, for the specific implementation of each module in the above device embodiment, please refer to the description in the above method embodiment, and details are not described herein again.

[0147] The above are only the preferred embodiments of the present invention, and are not intended to limit the present invention. Any modifications, equivalent replacements, and improvements made within the spirit and principles of the present invention shall be included in the protection scope of the present invention.

Claims

1. An intelligent printing typesetting method based on machine learning, characterized in that: include: Obtaining a user's requirement description set for printing and typesetting of a target finished product; wherein the requirement description set includes a plurality of requirement description information, and the requirement description information is used to describe the user's typesetting requirements for printing and typesetting of the target finished product; Extracting the feature of the description mode and converting the feature of the guiding value of the demand description set to obtain the high-dimensional guiding feature distribution of the demand description set corresponding to the target finished product printing typesetting, including: The demand description set includes description set type distribution and description set sentiment tendency, which together constitute the description pattern characteristics of the demand description set. The requirement description set is subjected to meaning parsing pointing analysis according to the description set type distribution in the description mode feature to obtain a first meaning parsing vector of each requirement description information in the requirement description set, including: Based on the printing layout image analysis weights of the text-type demand description information and the image-type demand description information in the demand description set corresponding to the type distribution of the description set, and the overall meaning proportions of the text-type demand description information and the image-type demand description information in the demand description set for the target finished product printing layout, the typesetting image conversion forms of the target finished product printing layout are analyzed for each of the demand description information, so as to obtain the typesetting image conversion forms of each of the demand description information, as the first meaning analysis vectors of each of the demand description information; Performing a meaning parsing and pointing analysis on the requirement description set according to the description set sentiment tendency in the description pattern feature to obtain a second meaning parsing vector of each requirement description information in the requirement description set; According to the first meaning parsing vector and the second meaning parsing vector corresponding to each of the requirement description information in the requirement description set, each of the requirement description information is respectively subjected to orientation value parsing to obtain a benchmark orientation value feature of each of the requirement description information; Based on a preset association database, a potential value-oriented correlation expansion analysis is performed on the description set type distribution in the description pattern feature and the description set sentiment tendency, so as to obtain an expansion-oriented value feature of the description pattern feature; Analyzing the degree of fit between the benchmark guiding value feature and the extended guiding value feature, and selecting the extended guiding value feature according to the result of the analysis, so as to obtain the additional guiding value feature whose degree of fit meets the predetermined standard among the extended guiding value features; Combining and weighting the baseline guiding value features and the additional guiding value features to obtain a high-dimensional guiding feature distribution of the target finished product printing layout; Performing multiple data simulations on the high-dimensional guiding feature distribution according to a pre-trained intelligent printing typesetting model to obtain a plurality of original version printing typesets corresponding to the high-dimensional guiding feature distribution; Building a user feedback channel based on each of the original version printing layouts, and accepting user optimization suggestions for each of the original version printing layouts through the user feedback channel; The high-dimensional guiding feature distribution is modified according to the optimization suggestion, and data simulation is performed on the modified high-dimensional guiding feature distribution according to the pre-trained intelligent printing typesetting model to obtain the target finished product printing typesetting.

2. The intelligent printing typesetting method based on machine learning according to claim 1, characterized in that: The steps of obtaining a set of user's requirement descriptions for the target finished product printing and typesetting include: Receive the required description information of the target finished product printing and typesetting provided by the user through the information interaction port; Evaluate the completeness and rationality of the requirement description information provided by the user, and mark the vulnerabilities of the requirement description information according to the evaluation results to obtain a correction mark based on the requirement description information; Receiving information correction processing corresponding to the correction mark from the user through the information interaction port, so as to correct the vulnerability of the requirement description information and obtain the final version of the requirement description information; All final versions of the requirement description information provided by the user through the information interaction port are combined to obtain the requirement description set.

3. The intelligent printing typesetting method based on machine learning according to claim 1, characterized in that: The steps of extracting the feature of the description mode and converting the feature of the guidance value of the demand description set to obtain the high-dimensional guidance feature distribution of the demand description set corresponding to the target finished product printing typesetting include: The feature extraction of the description pattern of the requirement description set is performed to obtain the description pattern features of the requirement description set; the feature conversion of the description pattern features of the requirement description set is performed to guide the value to obtain the high-dimensional guiding feature distribution of the requirement description set corresponding to the target finished product printing layout.

4. The intelligent printing typesetting method based on machine learning as claimed in claim 3, characterized in that: The step of extracting the features of the description pattern of the requirement description set to obtain the description pattern features of the requirement description set includes: Performing multi-dimensional information characteristic analysis on each demand description information in the demand description set to obtain type characteristics and emotional characteristics of each demand description information in the demand description set; wherein the type characteristics include text type and image type; Performing an overview analysis of type distribution of the requirement description set according to type characteristics of each requirement description information in the requirement description set to obtain a description set type distribution of the requirement description set; The overall sentiment tendency analysis of the demand description set is performed according to the sentiment characteristics of each demand description information in the demand description set to obtain the description set sentiment tendency of the demand description set.

5. The intelligent printing typesetting method based on machine learning as claimed in claim 4, characterized in that: The step of performing meaning parsing and pointing analysis on the requirement description set according to the description set type distribution in the description mode feature to obtain a first meaning parsing vector of each requirement description information in the requirement description set includes: Performing a ratio analysis and weight conversion of text type and image type on the description set type distribution to obtain a printing typesetting image parsing weight of the text type requirement description information and the image type requirement description information in the requirement description set corresponding to the description set type distribution; Performing overall meaning property analysis on the text-type requirement description information and the image-type requirement description information in the requirement description set to obtain the overall meaning proportion of the text-type requirement description information and the image-type requirement description information in the requirement description set to the target finished product printing typesetting; Based on the printing and typesetting image analysis weights of the text-type requirement description information and the image-type requirement description information in the requirement description set corresponding to the description set type distribution, and the overall meaning proportion of the text-type requirement description information and the image-type requirement description information in the requirement description set to the target finished product printing and typesetting, the typesetting image conversion form of each requirement description information for the target finished product printing and typesetting is analyzed to obtain the typesetting image conversion form of each requirement description information as the first meaning analysis vector of each requirement description information.

6. The intelligent printing typesetting method based on machine learning as claimed in claim 5, characterized in that: The step of performing a meaning parsing and pointing analysis on the requirement description set according to the description set sentiment tendency in the description pattern feature to obtain a second meaning parsing vector of each requirement description information in the requirement description set includes: Generate key description tags for target finished product printing and typesetting based on the description set sentiment tendency, so as to obtain several key description tags for the description set sentiment tendency for target finished product printing and typesetting; Label pointing parsing is performed on each requirement description information in the requirement description set according to each of the key description labels to obtain a second meaning parsing vector of each of the requirement description information.

7. The intelligent printing typesetting method based on machine learning according to claim 1, characterized in that: The pre-training steps of the intelligent printing typesetting model include: Collect printing and typesetting samples to construct a historical data set; wherein the printing and typesetting samples have corresponding sample construction feature distributions, and the sample construction feature distributions correspond to the high-dimensional guided feature distributions; Dividing the historical data set into a training data set, a verification data set and a test data set; Construct a convolutional neural network model consisting of a convolutional layer, an activation function layer, a pooling layer, a fully connected layer, and an output layer as the intelligent printing typesetting model to be trained; Substituting the training data set into the convolutional neural network model, so that the convolutional neural network model performs back propagation algorithm learning according to the training data set, obtains the mapping relationship between the printing typesetting sample and the sample construction feature distribution, and optimizes the convolutional neural network model according to the mapping relationship; In the process of training the convolutional neural network model using the training data set, the verification data set is substituted into the convolutional neural network model, and the convolutional neural network model is made to perform verification processing on the verification data set, so as to evaluate the model performance of the convolutional neural network model during the training process, and tune the model hyperparameters according to the evaluation results; wherein the model hyperparameters include a learning rate and a regularization parameter; After completing the model training of the training data set, the test data set is substituted into the convolutional neural network model, and the convolutional neural network model is made to perform test processing on the test data set to make a final evaluation of the model performance of the convolutional neural network model to determine whether to end the model training.

8. The intelligent printing typesetting method based on machine learning as claimed in claim 1, characterized in that: The step of modifying the high-dimensional guided feature distribution according to the optimization suggestion comprises: parsing the optimization suggestion for direct correction features to obtain direct correction features based on the optimization suggestion; Performing sentiment analysis on the optimization suggestion to obtain potential correction features fed back by the optimization suggestion, and performing correlation analysis on the potential correction features according to the direct correction features to obtain correlation analysis results between the potential correction features and the direct correction features; Selecting the potential correction features according to the association analysis result to obtain additional correction features; The direct correction feature and the additional correction feature are weighted, and the high-dimensional guided feature distribution is corrected based on the weighted direct correction feature and the additional correction feature.

9. An intelligent printing and typesetting device based on machine learning, characterized in that: A method for intelligent printing typesetting based on machine learning for implementing any one of claims 1 to 8, comprising: A demand acquisition module is used to acquire a user's demand description set for printing and typesetting of a target finished product; wherein the demand description set includes a plurality of demand description information, and the demand description information is used to describe the user's typesetting demand for printing and typesetting of the target finished product; A guidance analysis module, used for extracting the features of the description mode and converting the features of the guidance value of the demand description set to obtain a high-dimensional guidance feature distribution of the demand description set corresponding to the target finished product printing typesetting; A preliminary simulation module, used for performing multiple data simulations on the high-dimensional guiding feature distribution according to a pre-trained intelligent printing typesetting model, so as to obtain a plurality of original versions of printing typesetting corresponding to the high-dimensional guiding feature distribution; An optimization feedback module, used to construct a user feedback channel based on each of the original version printing layouts, and accept optimization suggestions from users on each of the original version printing layouts through the user feedback channel; The typesetting correction module is used to correct the high-dimensional guiding feature distribution according to the optimization suggestion, and perform data simulation on the corrected high-dimensional guiding feature distribution according to the pre-trained intelligent printing typesetting model to obtain the target finished product printing typesetting.

Images