A clothing intelligent design method and system based on social data
By extracting social network data in intelligent design, building an emotional space and style element matrix of clothing aesthetic preferences, and using the Fuzzy-Transformer model to convert data to generate clothing designs that meet users' aesthetic preferences, solving the data sources and recommendation methods in the existing technology The research on the limitations and aesthetic preferences of the lag problem has achieved efficient clothing recommendations.
Patent Information
- Application Number
- CN202310922170.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2023-07-26
- Publication Date
- 2024-11-01
- Estimated Expiration
- 2043-07-26
AI Technical Summary
The existing intelligent design method has limited data sources when recommending clothing, and the recommendation method is single. It is impossible to deeply explore the implicit preferences of users, and it is impossible to transform clothing aesthetic preferences from uncertain and sexy data to deterministic data, resulting in lag in aesthetic preference research. .
By extracting user social network content, building a matrix of clothing aesthetic preference emotional space and style elements, using the Fuzzy-Transformer model to convert high-frequency vocabulary into triangular fuzzy numbers, establishing a dynamic aesthetic preference model of the user, and generating recommended design elements through similarity detection , combine it into a complete clothing style picture.
It breaks through the limitations of data sources and recommendation methods, automatically explores users' implicit preferences, updates clothing recommendations in real time, solves the lag problem of aesthetic preference research, and provides clothing designs that meet users' aesthetic preferences.
Smart Images

Figure CN117315085B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of intelligent design technology, and in particular to a clothing intelligent design method and system based on social data. Background Art
[0002] With the development of science and technology, the development of intelligent design will move towards more advanced technologies, such as design for big data, design based on cloud computing, and even design based on artificial intelligence. In any case, intelligent design will have a significant impact on the design process, changing the traditional design method to better meet user needs.
[0003] Research on intelligent design has presented a variety of solutions. IrfiantiNur et al. analyzed the attributes of fashion products that most affect consumers' purchasing preferences through sampling surveys. Varghese collected information about market trends and product availability of body-shaping clothing through market research, and conducted a verification experiment on consumers to reveal their preferences and needs for the body-shaping clothing category. Hu et al. built an e-commerce visual system platform and used big data to analyze customer groups to solve related problems in visual marketing. Volkov et al. analyzed social network information in the supply chain and provided consumers with goods and services of interest based on their preferences; Wu et al. used keywords extracted from user posts as tags to directly regard them as user preferences and thus as the basis for personalized recommendations. Weng et al. used the LDA topic model to study the content of users' Weibo posts, thereby inferring the user's topic distribution and realizing personalized recommendations based on user preferences.
[0004] Although the existing intelligent design has made considerable progress, there are still the following shortcomings:
[0005] (1) Existing design methods focus on making recommendations based on users’ purchase and browsing history, with limited data sources and a single recommendation method.
[0006] (2) Existing data cannot deeply explore users’ implicit preferences, and the recommendation method is too superficial.
[0007] (3) When obtaining real-time data of users on social networks, existing methods cannot convert the uncertain perceptual data of clothing aesthetic preferences into deterministic data that can be quantitatively processed, thereby establishing a dynamic aesthetic preference model of users, and cannot solve the lag in aesthetic preference research. Summary of the invention
[0008] To this end, an embodiment of the present invention provides a clothing intelligent design method and system based on social data, which is used to solve the problem that the prior art cannot convert the uncertain perceptual data of clothing aesthetic preference into deterministic data that can be quantitatively processed, thereby establishing a dynamic aesthetic preference model of users, and cannot solve the problem of the lag in aesthetic preference research.
[0009] In order to solve the above problems, an embodiment of the present invention provides a clothing intelligent design method based on social data, the method comprising:
[0010] S1: Extract the content posted by users in social networks and integrate the content to form social network text;
[0011] S2: Processing the social network text to obtain high-frequency words;
[0012] S3: constructing a clothing aesthetic preference emotional space, perceptually evaluating high-frequency words from various dimensions of the clothing aesthetic preference emotional space, establishing a connection between high-frequency words and the clothing aesthetic preference emotional space, and obtaining perceptual evaluation results. The set of perceptual evaluation results is the user's clothing aesthetic preference model;
[0013] S4: Construct a clothing style element matrix, conduct a subjective evaluation of each element in the clothing style element matrix from various dimensions of the clothing aesthetic preference emotional space, establish the connection between clothing style design elements and clothing aesthetic preference emotional space, and obtain a clothing style knowledge model;
[0014] S5: performing a similarity test on the evaluation results in the clothing aesthetic preference model and the clothing style knowledge model, and setting the design element corresponding to the one with the highest similarity as the recommended design element;
[0015] S6: Combine the recommended design elements of each garment component to generate a complete garment style diagram.
[0016] Preferably, the method further comprises: recommending the complete clothing style picture to the user, collecting user evaluation information, and calculating the user's satisfaction with the recommended clothing style picture.
[0017] Preferably, the method for processing social network text to obtain high-frequency words is:
[0018] Firstly, the continuous sentences and paragraphs in the user's social network text are split into separate words and symbols through Chinese text segmentation. Then, the words are diverted according to the vocabulary case library to obtain words with research significance. Then, the frequency of occurrence of each word is counted and sorted. Finally, the top x words in frequency ranking are selected as the high-frequency word extraction results of each user.
[0019] Preferably, the vocabulary case library is composed of a stop word library and an adopted word library, the stop word library is used to enter meaningless words and symbols, and the adopted word library is used to enter words with research significance.
[0020] Preferably, the Chinese text word segmentation processing method is:
[0021] Case retrieval is performed based on social network texts and vocabulary case libraries; case matching judgment is performed based on the retrieval results. If the case matching is successful, the vocabulary is retained or eliminated based on the case content; if the case matching fails, manual judgment is performed, and meaningless words, symbols, and words with research significance are entered into the stop word library and the adopted word library respectively, so as to realize the update and learning of the vocabulary case library.
[0022] Preferably, the high-frequency words are perceptually evaluated from various dimensions of the clothing aesthetic preference emotional space, and the connection between the high-frequency words and the clothing aesthetic preference emotional space is established. The method for obtaining the perceptual evaluation result is:
[0023] The Fuzzy-Transformer model is used to perform perceptual evaluation on high-frequency words from various dimensions of clothing aesthetic preference emotional space, establish the connection between clothing style design elements and clothing aesthetic preference emotional space, and convert high-frequency words into triangular fuzzy numbers to obtain perceptual evaluation results.
[0024] Preferably, the conversion process of the Fuzzy-Transformer model is:
[0025] The normalization of user requirements is expressed as {(W i ,λ i ′ )|i=1,2,3}, where W i represents high-frequency words, λ i ′ represents word frequency;
[0026] In fuzzy comprehensive decision making, assuming that N(F c ) is to select F c The number of experts for each evaluation result in , the evaluation result is characterized as:
[0027]
[0028] Use p(F c )replace Represent the output results as quantitative data:
[0029]
[0030] Among them, TFN() represents the triangular fuzzy number, p(F c) represents the probability of high-frequency words at level c in a certain sentiment dimension, It represents the ratio of experts who choose degree c in manual evaluation, that is, the possibility of degree c under this sentiment dimension.
[0031] Preferably, each element in the clothing style element matrix is subjectively evaluated from various dimensions of the clothing aesthetic preference emotional space, and the connection between the clothing style design elements and the clothing aesthetic preference emotional space is established, and the method for obtaining the clothing style knowledge model is:
[0032] Each element in the clothing style element matrix is subjectively evaluated from various dimensions of the emotional space, and the connection between clothing style design elements and the emotional space of clothing aesthetic preference is established. After obtaining the subjective evaluation results, triangular fuzzy numbers are used to quantify them to obtain the clothing style knowledge model.
[0033] The embodiment of the present invention also provides a clothing intelligent design system based on social data, the system comprising:
[0034] A content extraction module is used to extract the content posted by the user in the social network and integrate the content to form social network text;
[0035] A social network text processing module, used to process the social network text to obtain high-frequency words;
[0036] A clothing aesthetic preference model acquisition module is used to construct a clothing aesthetic preference emotional space, perform a perceptual evaluation on high-frequency words from various dimensions of the clothing aesthetic preference emotional space, establish a connection between high-frequency words and the clothing aesthetic preference emotional space, and obtain a perceptual evaluation result. The collection of the perceptual evaluation results is the user's clothing aesthetic preference model;
[0037] The clothing style knowledge model acquisition module is used to construct a clothing style element matrix, conduct subjective evaluation on each element in the clothing style element matrix from various dimensions of the clothing aesthetic preference emotional space, establish the connection between clothing style design elements and clothing aesthetic preference emotional space, and obtain the clothing style knowledge model;
[0038] A recommended design element generation module is used to perform similarity detection on the evaluation results in the clothing aesthetic preference model and the clothing style knowledge model, and set the design element corresponding to the one with the highest similarity as the recommended design element;
[0039] The clothing style graph generation module is used to combine the recommended design elements of various clothing parts to generate a complete clothing style graph.
[0040] Preferably, the system further comprises a clothing style picture recommendation module, which is used to recommend complete clothing style pictures to users, collect user evaluation information, and calculate user satisfaction with the recommended clothing style pictures.
[0041] It can be seen from the above technical solutions that the present invention has the following advantages:
[0042] (1) The present invention applies social network data when providing users with clothing that meets their aesthetic preferences, breaking through the limitation of existing methods that only apply data directly related to clothing to product aesthetic preference research.
[0043] (2) The present invention proposes a Chinese text word segmentation processing method based on case-based reasoning technology, which uses the user's high-frequency vocabulary to study the user's personal clothing aesthetic preferences. It breaks through the current path limitation of mainly using data obtained from users' active behavior to study aesthetic preferences, and automatically mines out imperceptible implicit preferences from their social network data without interfering with the user.
[0044] (3) The present invention obtains real-time data of users on social networks and, based on the Fuzzy-Transformer model, converts the uncertain perceptual data of clothing aesthetic preference into deterministic data that can be quantitatively processed, thereby establishing a dynamic aesthetic preference model of users, breaking through the lag in aesthetic preference research. BRIEF DESCRIPTION OF THE DRAWINGS
[0045] In order to more clearly illustrate the implementation cases of the present invention or the technical solutions in the prior art, the following is a brief description of the drawings required for use in the embodiments. By referring to the drawings, the features and advantages of the present invention will be more clearly understood. The drawings are schematic and should not be understood as limiting the present invention in any way. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work. Among them:
[0046] Figure 1 is a flow chart of a clothing intelligent design method based on social data provided in an embodiment;
[0047] Figure 2 This is a workflow diagram of the intelligent design method for a dress in the embodiment;
[0048] Figure 3 Schematic diagram of the Chinese text word segmentation processing result (part) of user 1 in the embodiment;
[0049] Figure 4 This is a schematic diagram of the word frequency statistics result (part) of user 1 in the embodiment;
[0050] Figure 5A schematic diagram of a process flow of a Chinese text word segmentation processing method in an embodiment;
[0051] Figure 6 Schematic diagram of clothing aesthetic preference emotional space in the embodiment;
[0052] Figure 7 A schematic diagram of a conversion method based on a Fuzzy-Transformer model in an embodiment;
[0053] Figure 8 Schematic diagram of the dress style element matrix in the embodiment;
[0054] Fig. 9 It is a fuzzy language evaluation diagram in the embodiment;
[0055] Fig.10 Schematic diagram of triangular fuzzy numbers in the embodiment;
[0056] Fig.11 Schematic diagram of the dress style knowledge model in the embodiment;
[0057] Fig.12 is a dress style picture recommended for each user in the embodiment;
[0058] Fig.13 A block diagram of an intelligent clothing design system based on social data provided in an embodiment. DETAILED DESCRIPTION
[0059] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0060] Embodiment 1
[0061] like Figure 1 As shown, an embodiment of the present invention proposes a clothing intelligent design method based on social data, the method comprising:
[0062] S1: Extract the content posted by users in social networks and integrate the content to form social network text;
[0063] S2: Processing the social network text to obtain high-frequency words;
[0064] S3: constructing a clothing aesthetic preference emotional space, perceptually evaluating high-frequency words from various dimensions of the clothing aesthetic preference emotional space, establishing a connection between high-frequency words and the clothing aesthetic preference emotional space, and obtaining perceptual evaluation results. The set of perceptual evaluation results is the user's clothing aesthetic preference model;
[0065] S4: Construct a clothing style element matrix, conduct a subjective evaluation of each element in the clothing style element matrix from various dimensions of the clothing aesthetic preference emotional space, establish the connection between clothing style design elements and clothing aesthetic preference emotional space, and obtain a clothing style knowledge model;
[0066] S5: performing a similarity test on the evaluation results in the clothing aesthetic preference model and the clothing style knowledge model, and setting the design element corresponding to the one with the highest similarity as the recommended design element;
[0067] S6: Combine the recommended design elements of each garment component to generate a complete garment style diagram.
[0068] The present invention proposes a clothing intelligent design method based on social data. By extracting the content posted in the user's social network and integrating the content to form a social network text, the social network data is used to provide users with clothing that meets their aesthetic preferences, breaking through the limitation of the existing method that only applies data directly related to the product to the study of product aesthetic preferences. When processing the social network text to obtain high-frequency words, a Chinese text word segmentation processing method is proposed based on case reasoning technology. The user's personal clothing aesthetic preferences are studied with the help of the user's high-frequency words, breaking through the current path limitation of mainly using the data obtained from the user's active behavior to study aesthetic preferences, and automatically mining the imperceptible implicit preferences from their social network data without interfering with the user. With the help of the Fuzzy-Transformer model, the clothing aesthetic preference, which is a perceptual data with uncertainty, is converted into deterministic data that can be quantitatively processed, thereby establishing a dynamic aesthetic preference model of the user, breaking through the lag of aesthetic preference research.
[0069] The following takes the design of a dress as an example to specifically explain the design method proposed by the present invention. The workflow diagram of the design method is as follows: Figure 2 shown.
[0070] Furthermore, in step S1, the content posted in the user's social network is extracted and the content is integrated to form social network text. Specifically, the text data of the 100 pages of posts recently posted by each user is extracted by means of a crawler algorithm.
[0071] Furthermore, in step S2, the social network text is processed to obtain high-frequency words. Specifically, in order to obtain high-frequency words, first, the continuous sentences and paragraphs in the user's social network text are split into separate words and symbols through Chinese text segmentation processing. The Chinese text segmentation processing result is as follows: Figure 3 As shown; then, the vocabulary is divided according to the vocabulary case library to obtain vocabulary with research significance; then, the frequency of occurrence of each vocabulary is counted and sorted. The word frequency statistics result is as follows Figure 4 Finally, the top three words in frequency are selected as the high-frequency words of each user and used in subsequent designs. Table 1 below shows the statistical results of the high-frequency words of users.
[0072] Table 1
[0073]
[0074] In addition, since text segmentation is the process of recombining continuous character sequences into word sequences according to certain specifications. Unlike English, where one word is one word, Chinese is written in characters, and there are no direct and obvious marks between words, so Chinese text segmentation is more difficult. Therefore, the present invention proposes a Chinese text segmentation processing method based on case-based reasoning technology, such as Figure 5 shown.
[0075] The vocabulary in the vocabulary case library is obtained from statistics of a large amount of text data, and the number of times these words appear in the text (word frequency) is also counted, which can be used to resolve word segmentation ambiguity. At the same time, the vocabulary case library consists of two sub-libraries. The stop word library contains meaningless words and symbols, and the adopted word library contains words with research significance. Under such a distribution method, the word segmentation system can complete the task of vocabulary diversion while performing Chinese word segmentation, improving the efficiency of processing user texts. Next, the system will count the number of times the retained words appear in the text (word frequency), and after sorting the word frequencies, the top x words can be selected as the user's high-frequency words.
[0076] Specifically, the Chinese text word segmentation processing method is as follows: case retrieval is performed based on social network text and vocabulary case library; case matching judgment is performed based on the retrieval results. If the case matching is successful, the vocabulary is retained or eliminated based on the case content; if the case matching fails, manual judgment is performed, and meaningless words, symbols, and words with research significance are entered into the stop word library and the adopted word library respectively, so as to realize the update and learning of the vocabulary case library.
[0077] Further, in step S3, the emotional space of clothing aesthetic preference is constructed. First, we invited 35 designers and formed an evaluation team. Secondly, each member was asked to list as many words as possible to describe the fashion emotions of clothing. Then, the team members discussed to reduce redundant words, thereby retaining 70 emotional words recognized by all members. Then, the evaluation team matched these words into word pairs, and surveyed 150 female consumers through questionnaires. The respondents were asked to select at least 28 pairs of clothing perception semantic words that they thought were important. Finally, the evaluation team classified the word pairs with a selection rate greater than 60%, and selected 9 pairs of representative clothing perception semantic words, such as Figure 6 These words are selected based on the following principles: wide coverage and high representativeness, which makes the study of clothing aesthetic preferences more multidimensional and comprehensive, and is conducive to improving the accuracy of clothing style recommendations.
[0078] Since the traditional Transformer model only outputs the items with the largest impact factors during the decoding process, the information conveyed by the remaining items is ignored. In addition, this output result only converts the required information into qualitative semantic vocabulary, which is difficult to apply to scientific computing processes and lacks certain scientificity and rigor. The output pre-information of the traditional Transformer model can be represented as a vector (p(F1), p(F2), p(F3), p(F4), p(F5), p(F6), p(F7)), where p(F c ) represents the probability of high-frequency words in a certain emotional dimension. The Transformer model outputs the most likely item based on this information as a representation of the user's aesthetic preference. In order to solve the problem that the Transformer model has low information retention and fuzzy expression during the conversion process, the present invention proposes a quantitative conversion method based on the Fuzzy-Transformer model, such as Figure 7 As shown. First, the user demand is standardized as {(W i ,λ i ′ )|i=1,2,3}, where W i represents high-frequency words, λ i ′ Represents word frequency. In fuzzy comprehensive decision making, assuming that N(F c ) is to select F c The number of experts for each evaluation result in , the evaluation result is characterized as:
[0079]
[0080] Use p(F c )replace Represent the output results as quantitative data:
[0081]
[0082] Among them, TFN() represents the triangular fuzzy number, p(F c ) represents the probability of high-frequency words at level c in a certain sentiment dimension, It represents the ratio of experts who choose degree c in manual evaluation, that is, the possibility of degree c under this sentiment dimension.
[0083] Furthermore, after the construction of the clothing aesthetic preference emotional space is completed, the high-frequency words are perceptually evaluated from various dimensions of the clothing aesthetic preference emotional space with the help of the Fuzzy-Transformer model, the connection between the high-frequency words and the clothing aesthetic preference emotional space is established, and the high-frequency words are converted into triangular fuzzy numbers to obtain perceptual evaluation results. The set of perceptual evaluation results is the user's clothing aesthetic preference model. Table 2 below shows the triangular fuzzy number calculation results of user 1.
[0084] Table 2
[0085]
[0086]
[0087] Furthermore, in step S4, a dress style element matrix is constructed. First, all members of the evaluation team arrange a series of dress style categories (such as silhouette, collar type, sleeve type, etc.) based on their design experience and open resources. Then, the evaluation team selects the most appropriate design style classification by voting. This screening needs to follow the following principles: (1) The basic shape of the dress is decomposed into several independent and variable shape categories, and each category should be complete and non-repetitive; (2) Each clothing shape category can form a new design scheme after being recombined; (3) All members recognize the result.
[0088] Next, the team members listed the design elements of each style category (such as S-shaped, round neck, sleeveless, etc.). Then, the evaluation team voted to select the most appropriate design element classification. Similarly, the following principles are also followed during the selection process: (1) The various styles of the dress are decomposed into several independent and variable design elements, and each design element should be complete and non-repetitive; (2) The various design elements can be recombined to form a new design scheme; (3) All members agree on the result.
[0089] After the above two steps, we divided the dress style design into 11 categories and each category contains a total of 54 design elements, such as Figure 8 shown.
[0090] Furthermore, in order to establish the connection between the dress style design elements and the clothing aesthetic preference emotional space, and then complete the task of building a clothing style knowledge model, we require each member of the evaluation team to conduct a subjective evaluation of each element in the dress style element matrix from each dimension of the emotional space. In order to solve the contradictory state of fuzzy evaluation caused by the inability to accurately measure the performance of the evaluation object, the present invention uses fuzzy tools to quantify the designer's subjective evaluation and use these data to complete the construction of the corresponding model.
[0091] For example, when evaluating a designer’s fashion sense for a dress style design element in the “feminine-neutral” dimension, we use fuzzy language scoring criteria to divide the emotional dimension, such as Fig. 9 If the letters A and B are used to represent the two poles of perceptual semantic words, then each group of perceptual semantic words can be expressed in the form of "AB", that is, it can be divided into 7 levels: very A, very A, relatively A, average, relatively B, very B, and very B.
[0092] With the help of this rating chart, we can use different clothing perception semantic words to describe the designer's fashion perception. Based on fuzzy set theory, these perception words will be quantified into fuzzy numbers. In the present invention, triangular fuzzy numbers are used to quantify clothing perception semantic words, as shown in Table 3 below.
[0093] Table 3
[0094]
[0095] Triangular fuzzy numbers are a mathematical tool developed from "fuzzy sets" to deal with fuzzy and uncertain problems. They can convert fuzzy and uncertain language variables into definite numerical values. They are defined as follows: If there is a fuzzy set M on a given domain X, for any x∈X there is a number μ M (x)∈[0,1] corresponds to μ M (x) is called the membership of x to M. The triangular fuzzy number contains three parameters, which represent linguistic variables related to the membership of 0 or 1. That is, the above triangular fuzzy number can be recorded as M = (a, b, c), where a, b, and c represent the minimum value, expected value, and maximum value of the fuzzy event, respectively. Fig.10 shown.
[0096] With the help of the expert evaluation results and the triangular fuzzy numbers of each level of perceptual semantic vocabulary, this study can calculate the triangular fuzzy number of each dress style design element in each clothing perceptual semantic vocabulary dimension through the triangular fuzzy number formula. The triangular fuzzy number calculation formula is as follows:
[0097]
[0098] Taking the situation in Table 4 as an example, the expert scoring results of the dress style design element "off-shoulder neckline" in the "feminine-neutral" dimension are: 12 people think it is "Extremely feminine", 14 people think it is "Very feminine", and 9 people think it is "Alittle feminine". Then it can be concluded that the triangular fuzzy number of "off-shoulder neckline" in the "feminine-neutral" dimension is ((12*0+14*0+9*1) / 35, (12*0+14*1+9*3) / 35), (12*1+14*3+9*5) / 35), that is, (0.257, 1.171, 2.829).
[0099] Table 4
[0100]
[0101] After collecting the evaluation results of all members, we can quantify them using triangular fuzzy numbers to complete the construction of the dress style knowledge model, such as Fig.11 shown.
[0102] Furthermore, in step S5, a similarity test is performed on the evaluation results in the clothing aesthetic preference model and the dress style knowledge model, and the design element corresponding to the one with the highest similarity is set as the recommended design element.
[0103] Furthermore, in step S6, the recommended design elements of the various parts of the dress are combined. Table 5 below is a recommended design element combination of the dress style of user 1, and finally a complete dress style diagram is generated.
[0104] Table 5
[0105]
[0106] Furthermore, according to Table 5, we draw the corresponding dress style diagram and make recommendations. Among the five users, four users each get a dress style. Since user 4 has similar similarity detection results of multiple elements, the present invention provides two recommended style element combinations for him. Fig.12 Dress style pictures recommended for each user.
[0107] At the same time, the present invention also collects the evaluation information of users on the recommended dress style pictures, and the user's satisfaction evaluation on the recommended dress style pictures is shown in Table 6 below.
[0108] Table 6
[0109]
[0110] The variance of the satisfaction ratings of the five users for a total of six items was calculated, and the result was 1.139. In addition, since the rating of user 5 was somewhat different from that of others, this study excluded user 5 and recalculated the variance of the satisfaction ratings of the first four users, and the result was 0.24.
[0111] Embodiment 2
[0112] like Fig.13 As shown, the present invention provides a clothing intelligent design system based on social data, the system comprising:
[0113] The content extraction module 10 is used to extract the content published in the user's social network and integrate the content to form a social network text;
[0114] A social network text processing module 20, used to process the social network text to obtain high-frequency words;
[0115] A clothing aesthetic preference model acquisition module 30 is used to construct a clothing aesthetic preference emotional space, perform a perceptual evaluation on high-frequency words from various dimensions of the clothing aesthetic preference emotional space, establish a connection between the high-frequency words and the clothing aesthetic preference emotional space, and obtain a perceptual evaluation result. The set of perceptual evaluation results is the user's clothing aesthetic preference model;
[0116] A clothing style knowledge model acquisition module 40 is used to construct a clothing style element matrix, perform subjective evaluation on each element in the clothing style element matrix from various dimensions of the clothing aesthetic preference emotional space, establish a connection between clothing style design elements and clothing aesthetic preference emotional space, and obtain a clothing style knowledge model;
[0117] A recommended design element generation module 50 is used to perform similarity detection on the evaluation results in the clothing aesthetic preference model and the clothing style knowledge model, and set the design element corresponding to the one with the highest similarity as the recommended design element;
[0118] A clothing style diagram generating module 60 is used to combine the recommended design elements of various clothing components to generate a complete clothing style diagram;
[0119] The clothing style picture recommendation module 70 is used to recommend complete clothing style pictures to users, collect user evaluation information, and calculate user satisfaction with the recommended clothing style pictures.
[0120] The system is used to implement the above-mentioned intelligent clothing design method based on social data. In order to avoid redundancy, it will not be described here.
[0121] An embodiment of the present invention also provides an electronic device, which includes a processor, a memory and a bus system, wherein the processor and the memory are connected through the bus system, the memory is used to store instructions, and the processor is used to execute the instructions stored in the memory to implement the above-mentioned intelligent clothing design method based on social data.
[0122] An embodiment of the present invention further provides a computer storage medium, which stores a computer software product. The computer software product includes several instructions for enabling a computer device to execute the above-mentioned intelligent clothing design method based on social data.
[0123] Note that the above are only preferred embodiments of the present invention and the technical principles used. Those skilled in the art will understand that the present invention is not limited to the specific embodiments described herein, and that various obvious changes, readjustments and substitutions can be made by those skilled in the art without departing from the scope of protection of the present invention. Therefore, although the present invention has been described in more detail through the above embodiments, the present invention is not limited to the above embodiments, and may include more other equivalent embodiments without departing from the concept of the present invention, and the scope of the present invention is determined by the scope of the appended claims.
[0124] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application may adopt the form of a computer program product implemented in one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that include computer-usable program code.
[0125] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.
[0126] These computer program instructions may also be stored in a computer readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture including an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 These computer program instructions can also be loaded onto a computer or other programmable data processing device, so that a series of operation steps are executed on the computer or other programmable device to produce a computer-implemented process, so that the instructions executed on the computer or other programmable device provide for implementing the process in the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.
[0127] Obviously, the above embodiments are merely examples for clear explanation and are not intended to limit the implementation methods. For those skilled in the art, other different forms of changes or modifications can be made based on the above description. It is not necessary and impossible to list all the implementation methods here. The obvious changes or modifications derived from these are still within the protection scope of the invention.
Claims
1. A clothing intelligent design method based on social data, characterized in that: include: S1: Extract the content posted by users in social networks and integrate the content to form social network text; S2: Processing the social network text to obtain high-frequency words; S3: Constructing a clothing aesthetic preference emotional space, perceptually evaluating high-frequency words from various dimensions of the clothing aesthetic preference emotional space, establishing a connection between high-frequency words and the clothing aesthetic preference emotional space, and obtaining perceptual evaluation results. The set of perceptual evaluation results is the user's clothing aesthetic preference model, which specifically includes: Using the Fuzzy-Transformer model, the high-frequency words are perceptually evaluated from various dimensions of the clothing aesthetic preference emotional space, the connection between the high-frequency words and the clothing aesthetic preference emotional space is established, and the high-frequency words are converted into triangular fuzzy numbers to obtain the perceptual evaluation results. The conversion process of the Fuzzy-Transformer model is: The normalization of user requirements is expressed as {(W i ,λ i ′ )|i=1,2,3}, where W i represents high-frequency words, λ i ′ represents word frequency; In fuzzy comprehensive decision making, assuming that N(F c ) is to select F c The number of experts for each evaluation result in , the evaluation result is characterized as: Use p(F c )replace Represent the output results as quantitative data: Among them, TFN() represents the triangular fuzzy number, p(F c ) represents the probability of high-frequency words at level c in a certain sentiment dimension, It represents the ratio of experts who choose degree c in manual evaluation, that is, the possibility of degree c under this sentiment dimension; S4: Construct a clothing style element matrix, conduct a subjective evaluation of each element in the clothing style element matrix from various dimensions of the clothing aesthetic preference emotional space, establish the connection between clothing style design elements and clothing aesthetic preference emotional space, and obtain a clothing style knowledge model; S5: performing a similarity test on the evaluation results in the clothing aesthetic preference model and the clothing style knowledge model, and setting the design element corresponding to the one with the highest similarity as the recommended design element; S6: Combine the recommended design elements of each garment component to generate a complete garment style diagram.
2. The clothing intelligent design method based on social data according to claim 1 is characterized in that: Also includes: Recommend complete clothing style pictures to users, collect user evaluation information, and calculate user satisfaction with the recommended clothing style pictures.
3. The clothing intelligent design method based on social data according to claim 1 is characterized in that: The method for processing social network text to obtain high-frequency words is: Firstly, the continuous sentences and paragraphs in the user's social network text are split into separate words and symbols through Chinese text segmentation. Then, the words are diverted according to the vocabulary case library to obtain words with research significance. Then, the frequency of occurrence of each word is counted and sorted. Finally, the top x words in frequency ranking are selected as the high-frequency word extraction results of each user.
4. The clothing intelligent design method based on social data according to claim 3 is characterized in that: The vocabulary case library is composed of a stop word library and an adopted word library. The stop word library is used to enter meaningless words and symbols, and the adopted word library is used to enter words with research significance.
5. The clothing intelligent design method based on social data according to claim 3 is characterized in that: The Chinese text word segmentation processing method is: Case retrieval is performed based on social network texts and vocabulary case libraries; case matching judgment is performed based on the retrieval results. If the case matching is successful, the vocabulary is retained or eliminated based on the case content; if the case matching fails, manual judgment is performed, and meaningless words, symbols, and words with research significance are entered into the stop word library and the adopted word library respectively, so as to realize the update and learning of the vocabulary case library.
6. The clothing intelligent design method based on social data according to claim 1 is characterized in that: Each element in the clothing style element matrix is subjectively evaluated from various dimensions of the clothing aesthetic preference emotional space, and the connection between clothing style design elements and clothing aesthetic preference emotional space is established. The method to obtain the clothing style knowledge model is as follows: Each element in the clothing style element matrix is subjectively evaluated from various dimensions of the emotional space, and the connection between clothing style design elements and the emotional space of clothing aesthetic preference is established. After obtaining the subjective evaluation results, triangular fuzzy numbers are used to quantify them to obtain the clothing style knowledge model.
7. A clothing intelligent design system based on social data, characterized in that: The system is used to implement the social data-based intelligent clothing design method according to any one of claims 1 to 6, comprising: A content extraction module is used to extract the content posted by the user in the social network and integrate the content to form social network text; A social network text processing module, used to process the social network text to obtain high-frequency words; A clothing aesthetic preference model acquisition module is used to construct a clothing aesthetic preference emotional space, perform a perceptual evaluation on high-frequency words from various dimensions of the clothing aesthetic preference emotional space, establish a connection between high-frequency words and the clothing aesthetic preference emotional space, and obtain a perceptual evaluation result. The collection of the perceptual evaluation results is the user's clothing aesthetic preference model; The clothing style knowledge model acquisition module is used to construct a clothing style element matrix, conduct subjective evaluation on each element in the clothing style element matrix from various dimensions of the clothing aesthetic preference emotional space, establish the connection between clothing style design elements and clothing aesthetic preference emotional space, and obtain the clothing style knowledge model; A recommended design element generation module is used to perform similarity detection on the evaluation results in the clothing aesthetic preference model and the clothing style knowledge model, and set the design element corresponding to the one with the highest similarity as the recommended design element; The clothing style graph generation module is used to combine the recommended design elements of various clothing parts to generate a complete clothing style graph.
8. The clothing intelligent design system based on social data according to claim 7 is characterized in that: It also includes a clothing style picture recommendation module, which is used to recommend complete clothing style pictures to users, collect user evaluation information, and calculate user satisfaction with the recommended clothing style pictures.
