Search

CN-121981753-A - Method for constructing personalized design model of brand created by text based on user emotion recognition

CN121981753ACN 121981753 ACN121981753 ACN 121981753ACN-121981753-A

Abstract

The invention discloses a method for constructing a personalized design model of a text-created brand based on user emotion recognition, which relates to the technical field of text-created brand design, and comprises the steps of acquiring and processing emotion data of a user, extracting multidimensional emotion characteristics to form a user emotion characteristic pool, and establishing a multi-level design element pool which comprises visual elements, text elements, style templates and interaction forms for the text-created brand design, wherein each design element is marked with a quantized emotion arousal intensity value and emotion dimension labels. The matching rule provided by the invention can intelligently reconcile the complex emotion demands of users through a layering processing mechanism, and the fuzzy emotion is converted into computable data by means of quantization and vectorization. The method realizes accurate and data-driven matching from emotion to design elements, fundamentally solves the problem of disordered style caused by simple superposition in the prior art, and remarkably improves the emotion accuracy, consistency and overall harmony of the design.

Inventors

  • YU WEIJIE

Assignees

  • 成都职业技术学院

Dates

Publication Date
20260505
Application Date
20260123

Claims (10)

  1. 1. The method for constructing the personalized design model of the brand created by the text based on the emotion recognition of the user is characterized by comprising the following steps: Acquiring and processing emotion data of a user, extracting multidimensional emotion characteristics and forming a user emotion characteristic pool; establishing a multi-level design element pool, wherein the multi-level design element pool comprises visual elements, text elements, style templates and interaction forms for the text-created brand design, and each design element is marked with a quantized emotion arousal intensity value and an emotion dimension label; inputting the user emotion feature pool and the multi-level design element pool into a matching module based on a multi-dimensional emotion matching rule; executing at least one preset matching rule through the matching module so as to map the emotion characteristics of the user into corresponding design elements and output a matching result set; Calling corresponding design elements from the multi-level design element pool according to the matching result set, and combining to generate at least one candidate brand design scheme; sequencing and optimizing the at least one candidate brand design scheme according to a dynamic weight calculation model; And collecting actual interactive feedback data of the applied brand design scheme by the user, and performing iterative optimization on the matching rule and the emotion marking of the design element by utilizing the interactive feedback data.
  2. 2. The method for constructing the personalized design model of the brand created by the text based on the emotion recognition of the user according to claim 1, wherein the method for constructing the feature pool of the emotion of the user comprises the following steps: Collecting at least one of text content, physiological signals, behavior logs and explicit feedback of a user as original emotion data through a data interface; cleaning, denoising and standardizing the original emotion data to form structured data; And processing the structured data by adopting a pre-trained multi-mode fusion analysis model, and extracting multidimensional emotion characteristics comprising emotion titers, emotion awakening degrees and specific emotion type probability distribution.
  3. 3. The method for constructing the personalized design model of the brand created by the text based on the emotion recognition of the user, which is disclosed in claim 2, is characterized in that the emotion valence and the emotion arousal degree form an emotion two-dimensional coordinate for positioning the instantaneous emotion state of the user in a vector space, and the specific emotion type probability distribution is used for carrying out fine granularity classification on the emotion of the user.
  4. 4. The method for constructing the personalized design model of the brand created by the text based on the emotion recognition of the user as set forth in claim 1, wherein the establishing the multi-level design element pool specifically comprises the following steps: Labeling each design element with a corresponding emotion dimension label based on a preset text-created brand emotion arousing knowledge model, wherein the knowledge model stores the mapping relation between the design element characteristics and the emotion dimension labels; Through statistical analysis of historical interaction data or evaluation by combining expert knowledge, each design element is assigned with a numerical value as the emotion arousal intensity value for quantifying the emotion arousal capacity; And establishing an index relation between the marked design elements and the emotion dimension labels according to the types of the marked design elements so as to construct the multi-level design element pool.
  5. 5. The method for constructing a personalized design model of a brand created by text based on user emotion recognition according to claim 1, wherein the step of inputting the user emotion feature pool and the multi-level design element pool to a matching module based on a multidimensional emotion matching rule comprises the following steps: formatting the multidimensional emotion features in the user emotion feature pool, and converting the multidimensional emotion features into a feature vector sequence which can be identified by the matching module; Carrying out vectorization coding on emotion dimension labels of each design element in the multi-level design element pool to generate standardized element label vectors; and packaging the feature vector sequence and the standardized element tag vector according to a preset data structure and transmitting the feature vector sequence and the standardized element tag vector to an input interface of the matching module.
  6. 6. The method for constructing a personalized design model of a brand created by text based on emotion recognition of a user according to claim 1, wherein the executing at least one preset matching rule specifically comprises an emotion space quadrant mapping mode: defining a multi-quadrant emotion space formed by emotion valence dimension and emotion awakening degree dimension division in advance; configuring an associated composite design element module for each emotion quadrant; positioning core emotion features in the user emotion feature pool to corresponding quadrants of the multi-quadrant emotion space; and outputting the composite design element module associated with the quadrant as a matching result.
  7. 7. The method for constructing a personalized design model of a brand created by text based on emotion recognition of a user as set forth in claim 6, wherein the matching rules further comprise hierarchical combination rules, and the execution of the hierarchical combination rules comprises the following steps: the first stage, namely a basic style template matching stage, specifically comprises the following steps: Analyzing the intensity value of each emotion feature in the emotion feature pool of the user and the stability value in a preset time window; Determining at least one emotion feature with an intensity value exceeding a first preset intensity threshold and a stability value exceeding a first preset stability threshold as a core emotion feature; Matching the core emotion characteristics with a preset brand basic style template library, and outputting a basic style framework; The second stage is a personalized element adding stage, which specifically comprises the following steps: taking other emotion characteristics except the core emotion characteristics in the user emotion characteristic pool as secondary emotion characteristics; calculating the similarity between the feature vector of each secondary emotion feature and the label vector of each non-template design element in the multi-level design element pool; selecting at least one design element with similarity exceeding a preset similarity threshold and highest emotion evoked intensity value for each secondary emotion feature, and generating a personalized additive element list; And synthesizing the basic style framework and the personalized additive element list to form a layered combination scheme.
  8. 8. The method for constructing a personalized design model of a brand created by text based on emotion recognition of a user as set forth in claim 7, wherein the combining generates at least one candidate brand design scheme specifically including: analyzing the matching result set to obtain at least one design element identifier matched with each user emotion characteristic; Extracting corresponding design element entities from the multi-level design element pool according to the design element identifiers; And according to preset combination logic and layout constraint, and referring to the structure of the layered combination scheme, arranging and synthesizing the extracted design element entities to generate the at least one candidate brand design scheme.
  9. 9. The method for creating a personalized design model of a brand based on emotion recognition of a user as set forth in claim 1, wherein said ranking and optimizing said at least one candidate brand design scheme according to a dynamic weight calculation model comprises: Acquiring an intensity value of each emotion feature in the emotion feature pool of the user and a stability value in a preset time window; determining a correlation coefficient of each emotion feature and a brand design target according to a preset brand design target; Comprehensively calculating the intensity value, the stability value and the correlation coefficient of each emotion feature through a preset weight calculation function, and calculating corresponding dynamic weights; and according to the dynamic weights, adjusting the comprehensive scores of the candidate brand design schemes, sequencing the comprehensive scores, and outputting the optimized scheme sequences.
  10. 10. The method for constructing the personalized design model of the brand created by the text based on the emotion recognition of the user according to claim 1, wherein the iterative optimization of the matching rule and the emotion marking of the design element by using the interactive feedback data specifically comprises the following steps: converting the interactive feedback data collected in a preset time period into an emotion feedback effect value; calculating a predicted deviation value between the emotion feedback effect value and a predicted emotion arousal value according to which the brand design scheme is generated; carrying out numerical calibration on the emotion arousal intensity value and the emotion dimension label of the related design elements in the multi-level design element pool according to the predicted deviation value; And forming a training sample by the user emotion characteristics in the design task, the output brand design scheme and the corresponding emotion feedback effect value, and optimizing the internal weight parameters of the matching rule.

Description

Method for constructing personalized design model of brand created by text based on user emotion recognition Technical Field The invention relates to the technical field of text-created brand design, in particular to a method for constructing a text-created brand personalized design model based on user emotion recognition. Background With the rise of the digital creative industry, the literature brand design is gradually getting rid of the traditional mode depending on manpower and experience, and evolving towards the direction of intelligence, individualization and data driving. And new possibilities are opened for design automation by applying the technologies of emotion analysis, intelligent generation and the like. However, when faced with the emotional need for users to diversify, interleave, and even conflict with each other, which is the key to achieving a deep personalized design, the prior art still shows significant shortcomings. Currently common personalized recommendation or automatic generation systems rely on single or linear matching logic. For example, a style (e.g., national tide) may be simply associated with a set of fixed elements, or content may be mechanically superimposed according to several discrete labels. Such a method is often difficult to properly cope with when facing the complex design intent that users want to convey the sense of softness and thickness of traditional culture and also want to inject the light future feeling of modern technology. The system is either hard to mix elements of different gases, resulting in a confusing finished product style, a blurred theme, or can only discard part of the requirements, causing the output result to flow on the surface or lack depth. Disclosure of Invention The invention aims to provide a method for constructing a personalized design model of a brand created by text based on emotion recognition of a user, so as to solve the defects in the background technology. In order to achieve the above purpose, the invention provides the following technical scheme that the method for constructing the personalized design model of the brand created by the text based on the emotion recognition of the user comprises the following steps: Acquiring and processing emotion data of a user, extracting multidimensional emotion characteristics and forming a user emotion characteristic pool; establishing a multi-level design element pool, wherein the multi-level design element pool comprises visual elements, text elements, style templates and interaction forms for the text-created brand design, and each design element is marked with a quantized emotion arousal intensity value and an emotion dimension label; inputting the user emotion feature pool and the multi-level design element pool into a matching module based on a multi-dimensional emotion matching rule; executing at least one preset matching rule through the matching module so as to map the emotion characteristics of the user into corresponding design elements and output a matching result set; Calling corresponding design elements from the multi-level design element pool according to the matching result set, and combining to generate at least one candidate brand design scheme; sequencing and optimizing the at least one candidate brand design scheme according to a dynamic weight calculation model; And collecting actual interactive feedback data of the applied brand design scheme by the user, and performing iterative optimization on the matching rule and the emotion marking of the design element by utilizing the interactive feedback data. In a preferred embodiment, the forming the user emotion feature pool specifically includes: Collecting at least one of text content, physiological signals, behavior logs and explicit feedback of a user as original emotion data through a data interface; cleaning, denoising and standardizing the original emotion data to form structured data; And processing the structured data by adopting a pre-trained multi-mode fusion analysis model, and extracting multidimensional emotion characteristics comprising emotion titers, emotion awakening degrees and specific emotion type probability distribution. In a preferred embodiment, the emotion valence and the emotion arousal degree form an emotion two-dimensional coordinate for locating the instantaneous emotion state of the user in a vector space, and the specific emotion type probability distribution is used for fine-grained classification of the emotion of the user. In a preferred embodiment, the creating a multi-level design element pool specifically includes: Labeling each design element with a corresponding emotion dimension label based on a preset text-created brand emotion arousing knowledge model, wherein the knowledge model stores the mapping relation between the design element characteristics and the emotion dimension labels; Through statistical analysis of historical interaction data or evaluation by combining expert kno