A product introduction script content intelligent generation method and system

Abstract

The present application relates to the technical field of commodity information management, and in particular to a kind of commodity introduction copy content intelligent generation method and system.The method comprises: obtaining the attribute information of target commodity and its competitive product commodity in each dimension;According to the difference of attribute information of competitive product commodity in the same dimension, the sales situation of competitive product commodity and the overall difference of attribute information of target commodity and competitive product commodity, the basic important factor of each dimension attribute is obtained;According to the difference of attribute information of target commodity and competitive product commodity in the same dimension, the same grade commodity of target commodity in each dimension is screened;According to the number of competitive product commodity belonging to the same grade commodity in each dimension, the proportion of the number of same grade commodity in each dimension and the basic important factor, the emphasis description factor of each dimension is determined, and then the introduction copy of target commodity is generated.The present application improves the personalization, user attraction and conversion effect of the generated commodity introduction copy.

Description

Technical Field

[0001] This invention relates to the field of commodity information management technology, specifically to a method and system for intelligently generating commodity description text. Background Technology

[0002] With the continuous development of e-commerce and intelligent recommendation technology, product descriptions play a crucial role in product display and marketing conversion. High-quality product descriptions not only need to accurately describe the product's various attributes but also need to highlight the product's core selling points to attract user attention and improve click-through rates and conversion rates.

[0003] Existing methods for generating product copy mainly include manual writing and automatic generation based on templates or rules. Among them, manual writing relies on the experience of operations personnel, which is inefficient and difficult to scale; while template-based methods can improve generation efficiency, they usually only fill in fixed attributes and lack flexibility and personalized expression capabilities. Summary of the Invention

[0004] To address the lack of flexibility and personalization in product description text generated by existing methods, the present invention aims to provide an intelligent method and system for generating product description text content. The specific technical solution adopted is as follows: In a first aspect, the present invention provides a method for intelligently generating product description text, the method comprising the following steps: Obtain attribute information of the target product and its competitors' products across various dimensions; Based on the differences in attribute information between competing products in the same dimension, the sales performance of competing products, and the overall differences in attribute information between the target product and competing products, the basic key factors of each dimension's attributes are obtained; based on the differences in attribute information between the target product and competing products in the same dimension, products of the same grade as the target product in each dimension are screened. Based on the number of competing products that belong to the same product category in both dimensions, the proportion of the same product category in each dimension, and the corresponding basic important factors, the key descriptive factors for each dimension are determined. Based on the aforementioned key descriptive factors, generate introductory text for the target product.

[0005] Preferably, the basic key factors for each dimension of attributes are obtained based on the differences in attribute information between competing products in the same dimension, the sales performance of competing products, and the overall differences in attribute information between the target product and competing products. These factors include: Based on the differences in attribute information between each competing product and other types of competing products in the candidate dimensions, the attribute differentiation contribution of each competing product in the candidate dimensions is obtained. By combining the attribute differentiation contribution of all competing products in the candidate dimensions, the sales volume of each competing product, and the differences in attribute information between each competing product and the target product in the candidate dimensions, the basic important factors of the candidate dimension attributes are obtained. The candidate dimension can be any dimension.

[0006] Preferably, the basic key factors for candidate dimension attributes are obtained by comprehensively considering the attribute differentiation contribution of all competing products in the candidate dimension, the sales volume of each competing product, and the differences in attribute information between each competing product and the target product in the candidate dimension. These factors include: Based on the sales share of each competitor product, the proportion of the attribute differentiation contribution of each competitor product in the candidate dimension to the sum of the attribute differentiation contributions of all dimensions of the same competitor product, and the similarity of attribute information between each competitor product and the target product in the candidate dimension, the basic important factors of the candidate dimension attributes are obtained. The similarity of attribute information between each competing product and the target product in the candidate dimension is the negative correlation mapping value of the difference in attribute information between each competing product and the target product in the candidate dimension.

[0007] Preferably, the step of filtering comparable products across all dimensions based on the differences in attribute information between the target product and competing products in the same dimension includes: Based on the differences in attribute information between the target product and each competitor's product in the candidate dimensions, the disparity coefficient between the target product and each competitor's product in the candidate dimensions is obtained. Competitor products whose differential coefficient is less than the preset differential threshold are identified as products of the same grade as the target product in the candidate dimension.

[0008] Preferably, the step of determining the emphasis descriptive factors for each dimension based on the number of competing products that belong to the same product category in both pairs of dimensions, the proportion of competing products in the same product category in each dimension, and the corresponding basic importance factors includes: For any two dimensions: The proportion of competing products that simultaneously belong to the target product in any two of the stated dimensions is used as the attribute co-occurrence factor for any two dimensions. Based on the co-occurrence factors of the attributes of any two dimensions, the proportion of similar products in all competing products in each of the two dimensions, and the degree of dispersion of the basic importance factors of all dimensions, calculate the joint scarcity factor of any two dimensions. Based on the joint scarcity factor of any two dimensions and the difference between the basic importance factor of any two dimensions and the mean of the basic importance factor of all dimensions, the joint scarcity weight of the target product in any two dimensions is obtained. By combining the fundamental key factors of each dimension's attributes with the combined scarcity weights of the target product in each dimension and other dimensions, we obtain the key descriptive factors for each dimension.

[0009] Preferably, the step of calculating the joint scarcity factor of any two dimensions based on the co-occurrence factors of the attributes of any two dimensions, the proportion of similar products in each of the two dimensions among all competing products, and the degree of dispersion of the basic importance factors of all dimension attributes includes: Obtain the product of the proportion of similar products in all competing products in any two of the given two dimensions; Based on the relative magnitude of the product of the co-occurrence factors of attributes in any two dimensions and the degree of discrete distribution of the fundamental importance factors of all dimensions, the joint scarcity factor of any two dimensions is obtained.

[0010] Preferably, obtaining the joint scarcity weight of the target product in any two dimensions based on the joint scarcity factor of any two dimensions and the difference between the basic importance factor of the attribute in any two dimensions and the mean of the basic importance factor of all attributes in all dimensions includes: Calculate the difference factor between the base importance factors of any two dimensional attributes and the mean of the base importance factors of all dimensional attributes; By combining the difference factor and the joint scarcity factor, the joint scarcity weight of the target product in any two dimensions is obtained.

[0011] Preferably, the degree of discrete distribution of the fundamental important factors of all dimensional attributes is the variance of the fundamental important factors of all dimensional attributes.

[0012] Preferably, the step of generating the introductory text for the target product based on the emphasized descriptive factors includes: Sort all dimensions in descending order according to the emphasized descriptor to obtain the dimension sequence; Select the first preset number of dimensions from the dimension sequence as the core descriptive attributes; The core descriptive attributes and their corresponding emphasis descriptive factors are constructed into generation control information, which is then input into a trained language model to generate the introductory text for the target product.

[0013] Secondly, the present invention provides an intelligent product description copywriting generation system, which implements the method described in the first aspect above. The system includes: The attribute acquisition module is used to acquire attribute information of the target product and its competitors' products in various dimensions. The product filtering module is used to obtain the basic key factors of each dimension of attributes based on the differences in attribute information between competing products in the same dimension, the sales performance of competing products, and the overall differences in attribute information between the target product and competing products; and to filter products of the same grade in each dimension based on the differences in attribute information between the target product and competing products. The attribute analysis module is used to determine the key descriptive factors for each dimension based on the number of competing products that belong to the same product category in both pairs of dimensions, the proportion of the same product category in each dimension, and the corresponding basic important factors. The copywriting generation module is used to generate introductory copy for the target product based on the emphasized descriptive factors.

[0014] The present invention has at least the following beneficial effects: This invention first quantifies the fundamental importance factors of each dimension's attributes based on the attribute differences between competing products in the same dimension, the sales performance of competing products, and the overall attribute differences between the target product and competing products. These fundamental importance factors comprehensively reflect the basic importance of each attribute dimension in terms of market competitor differentiation, mainstream sales preferences, and its suitability for the target product, providing an objective and quantifiable weighting basis for subsequent copywriting generation. Based on the attribute differences between the target product and competing products in the same dimension, it filters out comparable products in each dimension, allowing subsequent processing to focus on competitors at the same competitive level as the target product in that dimension. This eliminates information interference from competitors with excessively large differences or irrelevant information, improving the targeting and accuracy of the analysis. Finally, it calculates the number of competing products that are simultaneously comparable in each pair of dimensions and the percentage of comparable products in each dimension. By combining fundamental key factors, the key descriptive factors for each dimension are determined. This approach not only considers the fundamental importance of a single dimension but also identifies key attribute combinations with scarcity or strong synergistic effects in multi-dimensional attribute combinations by exploring the relationships between different attribute dimensions among similar products. This cross-dimensional joint evaluation method can more accurately capture the unique advantages or differentiated selling points of the target product compared to its competitors. Furthermore, based on the obtained key descriptive factors, the product description copy is generated. The generated description copy can specifically highlight the core attributes of the target product that are most competitive and attractive compared to similar products. This solves the problems of lack of flexibility, low personalization, and difficulty in highlighting the unique selling points of the target product in the existing technology for product copy generation. It improves the personalization, user appeal, and conversion effect of the generated description copy, realizing the intelligent and precise generation of product copy. Attached Figure Description

[0015] To more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0016] Figure 1 A flowchart illustrating a method for intelligently generating product description text content provided in an embodiment of the present invention; Figure 2 This is a structural block diagram of a product description copywriting intelligent generation system provided in an embodiment of the present invention. Detailed Implementation

[0017] To further illustrate the technical means and effects adopted by the present invention to achieve the intended purpose, the following detailed description of a method and system for intelligent generation of product description text content based on the present invention, in conjunction with the accompanying drawings and preferred embodiments, is provided below.

[0018] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0019] The following description, in conjunction with the accompanying drawings, details the specific solution of the intelligent generation method and system for product description text provided by this invention.

[0020] An example of a method for intelligently generating product description text: This embodiment proposes a method for intelligently generating product description text, such as... Figure 1 As shown, the intelligent generation method for product description text in this embodiment includes the following steps: Step S1: Obtain attribute information of the target product and its competitors' products in various dimensions.

[0021] When generating product descriptions, firstly, the product for which descriptions are to be generated is designated as the target product. All products that are already sold, have existing descriptions, and compete with the target product are designated as competitor products. Then, attribute information for different dimensions of both the target product and each competitor product is obtained. Attribute information may include, but is not limited to, weight, volume, thickness, battery life, price, and hardness. Weight is one dimension of attribute information. For example, volume is one dimension, thickness is another, battery life is yet another, and so on. In practical applications, implementers can determine the attribute information based on the specific type of product and its selling points.

[0022] For each dimension's attribute information, normalization is performed, and the normalized attribute information is recorded as the attribute value. Normalization can be performed using the maximum-minimum normalization method or other existing data normalization methods. It should be noted that the attribute information and attribute values ​​mentioned below refer to the normalized attribute values. Specifically, if the maximum value of the dataset for the dimension to be normalized equals the minimum value, then all normalization results for that dataset are directly set to 0.

[0023] Step S2: Based on the differences in attribute information between competing products in the same dimension, the sales performance of competing products, and the overall differences in attribute information between the target product and competing products, obtain the basic key factors of each dimension attribute; based on the differences in attribute information between the target product and competing products in the same dimension, filter products of the same grade as the target product in each dimension.

[0024] When generating introductory copy for a target product, it's necessary to consider the similarity of attribute information between the target product and competing products. The higher the similarity and the greater the sales volume of the corresponding competing product, the greater its reference value and the more attention it should receive. Therefore, it's essential to evaluate the differences in attribute information across the same dimensions between competing and target products.

[0025] The following explanation uses one-dimensional attribute information as an example. Other dimensions can be processed using the method provided in this embodiment.

[0026] Specifically, any dimension is designated as a candidate dimension. Based on the differences in attribute information between each competing product and other types of competing products on the candidate dimension, the attribute differentiation contribution of each competing product on the candidate dimension is obtained. The attribute value of a competing product on a dimension is the average of the attribute values ​​of all products of that competing product on that dimension.

[0027] As one specific implementation method, the first The first of the competing products The contribution of each dimension of attributes can be represented as:

[0028] in, Indicates the first The first of the competing products Each dimension of attribute distinguishes contribution. This indicates the variety and quantity of competing products. Indicates the first The first of the competing products The attribute values ​​of each dimension, Indicates the first The first of the competing products The attribute values ​​of each dimension, Indicates the absolute value sign. This represents the normalization function.

[0029] Indicates the first Competing products and the first The competing products in the first The difference in attribute values ​​across all dimensions; the larger the value, the greater the difference between the two. Characterizing the first The overall difference in attribute information between this type of competing product and other types of competing products; the larger the value, the better. Competing products and the first The greater the overall difference in attribute information among competing products.

[0030] No. The first of the competing products The attribute contribution of the first dimension is used to reflect the distinguishing contribution of the second dimension. The competing products in the first The degree of difference between a product and other competing products in each dimension of attributes. The greater the difference between a product and other competing products in a given dimension, the higher its distinguishing contribution, indicating that the attribute in that dimension has a stronger distinguishing ability among the competing products.

[0031] Using the above method, the attribute differentiation contribution of each dimension of each competitor product can be obtained. By calculating the attribute differentiation contribution, the interference of highly concentrated distribution of attributes in each dimension on subsequent weight calculation can be avoided, thereby improving the accuracy and stability of attribute importance assessment.

[0032] Furthermore, by comprehensively considering the attribute differentiation contribution of all competing products in the candidate dimensions, the sales volume of each competing product, and the differences in attribute information between each competing product and the target product in the candidate dimensions, the basic important factors of the candidate dimension attributes are obtained. Specifically, based on the sales volume ratio of each competing product, the proportion of the attribute differentiation contribution of each competing product in the candidate dimensions to the sum of the attribute differentiation contributions of all dimensions of the same competing product, and the similarity of attribute information between each competing product and the target product in the candidate dimensions, the basic important factors of the candidate dimension attributes are obtained; among them, the similarity of attribute information between each competing product and the target product in the candidate dimensions is the negative correlation mapping value of the differences in attribute information between each competing product and the target product in the candidate dimensions.

[0033] As a specific implementation method, a detailed calculation formula for the basic importance factor is given. The basic importance factor of the u-th dimension attribute can be expressed as:

[0034] in, This represents the fundamental importance factor of the u-th dimension attribute. This indicates the variety and quantity of competing products. This represents the sales volume of the i-th competing product. This represents the sales volume of the j-th competing product. This represents the total number of dimensions. This represents the distinguishing contribution of the u-th dimension attribute of the i-th competing product. This represents the distinguishing contribution of the v-th dimension attribute of the i-th competing product. This represents the attribute value of the u-th dimension of the target product. This represents the attribute value of the u-th dimension of the i-th competing product. Represents the normalization function. Indicates the absolute value sign.

[0035] In this embodiment, the data normalization process uses the maximum-minimum value normalization method, which is existing technology and will not be described in detail here. As other implementation methods, other existing data normalization methods can also be used for processing.

[0036] This represents the sales share of the i-th competing product. The proportion of the attribute differentiation contribution of the i-th competitor product in the u-th dimension to the sum of the attribute differentiation contributions of the i-th competitor product across all dimensions is given. Specifically, if... If the value of is 0, then directly let The value is 0.

[0037] The basic importance factor represents the similarity of attribute information between the i-th competing product and the target product in the u-th dimension. The larger the value, the more similar the i-th competing product and the target product are in the u-th dimension. The basic importance factor comprehensively considers the sales volume information of competing products, the contribution of attribute differentiation, and the similarity between the target product and competing products. Specifically, it reflects the true market preference through sales volume share, suppresses the influence of homogeneous attributes through attribute differentiation contribution, and characterizes the degree of matching between the target product and high-selling competing products through attribute information similarity. This allows the basic importance factor to simultaneously reflect mainstream market trends and the matching degree of target product characteristics, providing a reliable weighting basis for subsequent copywriting generation.

[0038] Further, taking candidate dimensions as an example, specifically, the absolute value of the difference between the attribute information of the target product and each competitor's candidate dimension is calculated. This absolute value is taken as the difference between the attribute information of the target product and each competitor's candidate dimension. This difference is then normalized, and the normalized result is taken as the disparity coefficient between the target product and each competitor's candidate dimension. In this embodiment, the maximum-minimum normalization method is used to normalize the difference. Competitors with disparity coefficients less than a preset disparity threshold are identified as products of the same level as the target product in the candidate dimension. The preset disparity threshold can be dynamically set. In this embodiment, 10% of the standard deviation of the attribute values ​​(normalized values) of all competitor's candidate dimensions is set as the preset disparity threshold.

[0039] The above methods enable the screening of similar products across various dimensions. This screening process effectively eliminates interference from competing products with significant differences, allowing subsequent calculations of co-occurrence factors to focus more on the competitive landscape of the target product, thereby improving the relevance and effectiveness of attribute combination analysis.

[0040] Step S3: Determine the key descriptive factors for each dimension based on the number of competing products that belong to the same product category in both dimensions, the proportion of the same product category in each dimension, and the corresponding basic important factors.

[0041] Considering that the generated product descriptions should align with consumers' concerns, in order to achieve this goal, it is necessary to select core descriptive attributes from all dimensions of attribute information based on information of similar products, and use these as the key dimensions to focus on during the subsequent description generation process.

[0042] Specifically, for any two dimensions: the proportion of competing products of the same grade as the target product in both dimensions is used as the attribute co-occurrence factor for these two dimensions. That is, the ratio of the number of competing products of the same grade as the target product in both dimensions to the total number of all competing products is used as the attribute co-occurrence factor for these two dimensions. The product of the proportions of the same grade of products in these two dimensions among all competing products is obtained. Based on the relative magnitude of the attribute co-occurrence factors of these two dimensions and this product, as well as the degree of dispersion of the basic importance factors of all dimensions, the joint scarcity factor of these two dimensions is obtained.

[0043] As a specific implementation method, a specific calculation formula for the joint scarcity factor is given, the first... The dimension and the first The joint scarcity factor across all dimensions can be expressed as:

[0044] in, Indicates the first The dimension and the first A joint scarcity factor across multiple dimensions. Indicates the target product in the first... The dimension and the first Co-occurrence factors of attributes in each dimension, Indicates the first The percentage of similar products in each dimension among all competing products. Indicates the first The percentage of similar products in each dimension among all competing products. This represents an exponential function with the natural constant as its base. This represents the variance of the fundamental important factors for all dimensional attributes.

[0045] Specifically, if the first The proportion of similar products in each dimension among all competing products is 0 or the highest. If the proportion of similar products in a given dimension among all competing products is 0, then directly set the percentage of similar products in the first dimension to 0. The dimension and the first The joint scarcity factor for each dimension is 0, so the above formula for calculating the joint scarcity factor is no longer used.

[0046] The variance of the fundamental importance factors of all dimensional attributes is used to characterize the degree of dispersion of the fundamental importance factors of all dimensional attributes. The larger the value, the more dispersed the distribution of the fundamental importance factors. The variance of the fundamental importance factors of all dimensional attributes is used as an adjustment coefficient. When the fundamental importance of each dimensional attribute differs greatly, the weight of the rare attribute combination is increased to uncover differentiated selling points.

[0047] The joint scarcity factor achieves a quantitative characterization of the scarcity of attribute combinations by negatively mapping the co-occurrence factor of attributes to the quantity ratio of similar products in a single dimension. This is achieved by introducing the first... The dimension and the first The proportion of each product in the same category among all competing products in each dimension is used to achieve the goal of [the following]. The dimension and the first Normalization of attribute co-occurrence factors across all dimensions can eliminate the influence of the frequency of a single attribute on the assessment of combined scarcity, thus more accurately reflecting the synergistic or repulsive relationships between attributes; The dimension and the first The larger the value of the joint scarcity factor in the dimension, the stronger the scarcity factor. The dimension and the first Each dimension has high differentiation potential in the competitive environment of the target product.

[0048] Furthermore, the mean of the basic importance factors of all dimensional attributes is calculated, and the difference factor is obtained based on the difference between the basic importance factors of these two dimensional attributes and the mean. Combining the difference factor and the joint scarcity factor, the joint scarcity weight of the target product in these two dimensions is obtained. Specifically, the joint scarcity weight of the target product in these two dimensions is obtained based on the joint scarcity factor of these two dimensions and the difference between the basic importance factors of these two dimensions and the basic importance factors of all dimensional attributes.

[0049] As one specific implementation method, a detailed calculation formula for the joint scarcity weight is given, where the target commodity is in the [missing information]. The dimension and the first The joint scarcity weights of the dimensions can be expressed as:

[0050] in, Indicates the target product in the first... The dimension and the first Joint scarcity weights across multiple dimensions Indicates the first The dimension and the first A joint scarcity factor across multiple dimensions. Indicates the first The fundamental and important factors of each dimension attribute Indicates the first The fundamental and important factors of each dimension attribute This represents the average of the fundamental importance factors for all dimensional attributes. This represents the normalization function.

[0051] This represents the difference between the sum of the fundamental importance factors of these two dimensions and the mean of the fundamental importance factors of all dimensions. The difference factor is used to screen for scarcity. Only when the relevant dimension itself has high basic importance will its corresponding scarce combination be enhanced, thereby avoiding misjudging the accidental combination of low-value attributes as key selling points and improving the robustness and practical application value of the overall weight calculation.

[0052] Furthermore, by combining the fundamental important factors of each dimension's attributes, the joint scarcity weight of the target product in each dimension and other dimensions, and the degree of discrete distribution of the fundamental important factors of all dimension attributes, the key descriptive factors for each dimension are obtained.

[0053] As a specific implementation method, the degree of discrete distribution of the fundamental important factors of all dimensional attributes can be the variance of the fundamental important factors of all dimensional attributes.

[0054] As one specific implementation method, a detailed calculation formula for the emphasis descriptor is given, the first... The key descriptive factors for each dimension can be expressed as:

[0055] in, Indicates the first The key descriptive factors of each dimension Indicates the first The fundamental and important factors of each dimension attribute This represents the variance of the fundamental important factors for all dimensional attributes. Indicates the target product in the first... The dimension and the first Joint scarcity weights across multiple dimensions This represents the normalization function.

[0056] The descriptive factors comprehensively consider the fundamental importance of a single-dimensional attribute and its joint scarcity relationship with other dimensional attributes. The variance of the fundamental importance factors of all dimensional attributes is used to characterize the degree of dispersion of the fundamental importance factors of all dimensional attributes. The larger the value, the more dispersed the distribution of the fundamental importance factors of all dimensional attributes. When the variance is larger, the difference in importance between dimensions is greater, and it is necessary to amplify the joint scarcity weight of attributes to discover long-tail selling points. The smaller the variance, the more necessary it is to maintain the basic importance ranking. By introducing the variance of the fundamental importance factors of all dimensional attributes to weight and adjust the joint scarcity weight, the ability to express differentiation is enhanced in scenarios with large differences in attribute distribution, while the weight remains stable in scenarios with relatively uniform attribute distribution. Finally, the weights are normalized through a weight normalization function to make the weights of each dimensional attribute comparable and can be directly used to guide the description priority and expression intensity of each attribute in product description copy.

[0057] Using the methods described above, we can obtain the key descriptive factors for each dimension.

[0058] Step S4: Generate the introductory text for the target product based on the emphasized descriptive factors.

[0059] In step S3, the emphasis descriptive factors for each dimension are obtained. Next, the core descriptive attributes will be determined based on the values ​​of the emphasis descriptive factors for different dimensions, thereby increasing the focus on the core descriptive attributes and making the generated description copy for the target product more in line with the interests of buyers.

[0060] Specifically, all dimensions are sorted in descending order according to the emphasis descriptor to obtain a dimension sequence; the first preset number of dimensions in the dimension sequence are selected as core descriptive attributes; the preset number is dynamically set according to the total number of dimensions, and the preset number is generally set to 40%~65% of the total number of dimensions. In this embodiment, the preset number is set to 50% of the total number of dimensions. For example, if the total number of attribute dimensions is 10, then the preset number is 5. In specific applications, the implementer can set it according to the specific situation.

[0061] The core descriptive attributes and their corresponding emphasis factors are constructed as generation control information and input into the trained language model. Specifically, the emphasis factors are mapped to degree words, such as "extremely high" and "significant," according to numerical ranges, and concatenated with the corresponding attribute names to form control instructions, such as: "Emphasizing the thickness attribute of this product, its advantage level is extremely high," which is then input into the trained language model. The numerical range is set by the user according to specific circumstances, and will not be elaborated further here. By introducing emphasis factors in the generation prompts or weighting the model's attention allocation, the model prioritizes high-weight attributes during text generation to generate introductory text for the target product. The language model uses a neural network model based on the Transformer architecture. The training of the neural network model is a current technology, and will not be elaborated further here.

[0062] Thus, the above method was used to generate introductory text for the target product.

[0063] This embodiment first quantifies the fundamental importance factors of each dimension's attributes based on the attribute differences between competing products in the same dimension, the sales performance of competing products, and the overall attribute differences between the target product and competing products. These fundamental importance factors comprehensively reflect the basic importance of each attribute dimension in terms of market competitor differentiation, mainstream sales preferences, and the degree of matching with the target product itself, providing an objective and quantifiable weighting basis for subsequent copywriting generation. Based on the attribute differences between the target product and competing products in the same dimension, it filters out products in the same category as the target product in each dimension. This allows subsequent processing to focus on competitors at the same competitive level as the target product in that dimension, effectively eliminating information interference from competitors with excessively large differences or irrelevant information, thus improving the targeting and accuracy of the analysis. Finally, it calculates the number of competing products that are in the same category in each pair of dimensions and the percentage of products in the same category in each dimension. By combining fundamental key factors, the system identifies key descriptive factors for each dimension. This approach not only considers the basic importance of a single dimension but also explores the relationships between different attribute dimensions within the same product category. It identifies key attribute combinations with scarcity or strong synergistic effects in multi-dimensional attribute combinations. This cross-dimensional joint evaluation method can more accurately capture the unique advantages or differentiated selling points of the target product compared to competitors. Furthermore, based on the obtained key descriptive factors, the system generates introductory text for the target product. The generated text effectively highlights the core attributes that make the target product most competitive and attractive compared to competitors, solving the problems of inflexibility, low personalization, and difficulty in highlighting the unique selling points of the target product in existing product copywriting generation technologies. This enhances the marketing persuasiveness, user appeal, and conversion rate of the generated text, achieving intelligent and precise generation of product copywriting.

[0064] Example of an intelligent product description copy generation system: See Figure 2 The diagram illustrates a structural block diagram of an intelligent product description copywriting generation system according to an embodiment of the present invention. The system may include an attribute acquisition module, a product filtering module, an attribute analysis module, and a copywriting generation module.

[0065] Among them, the attribute acquisition module is used to acquire attribute information of the target product and its competitors' products in various dimensions; The product filtering module is used to obtain the basic key factors of each dimension of attributes based on the differences in attribute information between competing products in the same dimension, the sales performance of competing products, and the overall differences in attribute information between the target product and competing products; and to filter products of the same grade in each dimension based on the differences in attribute information between the target product and competing products. The attribute analysis module is used to determine the key descriptive factors for each dimension based on the number of competing products that belong to the same product category in both pairs of dimensions, the proportion of the same product category in each dimension, and the corresponding basic important factors. The copywriting generation module is used to generate introductory copy for the target product based on the emphasized descriptive factors.

[0066] It should be understood that Figure 2 The structural block diagram and modules of the intelligent product description copy generation system shown can be implemented in various ways. For example, in some embodiments, the system and its modules can be implemented by hardware, software, or a combination of software and hardware. The hardware portion can be implemented using dedicated logic; the software portion can be stored in memory and executed by appropriate instructions, such as a microprocessor or dedicated hardware. Those skilled in the art will understand that the above-described methods and systems can be implemented using computer-executable instructions and / or included in processor control code, for example, such code provided on a carrier medium such as a disk, CD, or DVD-ROM, a programmable memory such as read-only memory (firmware), or a data carrier such as an optical or electronic signal carrier. The system and its modules in this specification can be implemented not only by hardware circuits such as very large-scale integrated circuits or gate arrays, semiconductors such as logic chips, transistors, or programmable hardware devices such as field-programmable gate arrays, programmable logic devices, etc., but also by software, for example, executed by various types of processors, or by a combination of the above-described hardware circuits and software (e.g., firmware).

[0067] For more details about the above modules, please refer to other parts of this manual; they will not be repeated here.

[0068] It should be noted that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for intelligently generating product description text, characterized in that, The method includes the following steps: Obtain attribute information of the target product and its competitors' products across various dimensions; Based on the differences in attribute information between competing products in the same dimension, the sales performance of competing products, and the overall differences in attribute information between the target product and competing products, the basic key factors of each dimension's attributes are obtained; based on the differences in attribute information between the target product and competing products in the same dimension, products of the same grade as the target product in each dimension are screened. Based on the number of competing products that belong to the same product category in both pairs of dimensions, the proportion of competing products in the same product category in each dimension, and the corresponding basic important factors, the key descriptive factors for each dimension are determined, including: For any two dimensions: The proportion of competing products that simultaneously belong to the target product in any two of the stated dimensions is used as the attribute co-occurrence factor for any two dimensions. Based on the co-occurrence factors of the attributes of any two dimensions, the proportion of similar products in all competing products in each of the two dimensions, and the degree of dispersion of the basic importance factors of all dimensions, calculate the joint scarcity factor of any two dimensions. Based on the joint scarcity factor of any two dimensions and the difference between the basic importance factor of any two dimensions and the mean of the basic importance factor of all dimensions, the joint scarcity weight of the target product in any two dimensions is obtained. By combining the fundamental key factors of each dimension's attributes with the joint scarcity weights of the target product in each dimension and other dimensions, the key descriptive factors for each dimension are obtained. Based on the aforementioned key descriptive factors, generate introductory text for the target product.

2. The intelligent generation method for product description text according to claim 1, characterized in that, Based on the differences in attribute information between competing products in the same dimension, the sales performance of competing products, and the overall differences in attribute information between the target product and competing products, the basic key factors for each dimension of attributes are obtained, including: Based on the differences in attribute information between each competing product and other types of competing products in the candidate dimensions, the attribute differentiation contribution of each competing product in the candidate dimensions is obtained. By combining the attribute differentiation contribution of all competing products in the candidate dimensions, the sales volume of each competing product, and the differences in attribute information between each competing product and the target product in the candidate dimensions, the basic important factors of the candidate dimension attributes are obtained. The candidate dimension can be any dimension.

3. The intelligent generation method for product description text according to claim 2, characterized in that, The basic key factors of the candidate dimension attributes are obtained by comprehensively considering the attribute differentiation contribution of all competing products in the candidate dimension, the sales volume of each competing product, and the differences in attribute information between each competing product and the target product in the candidate dimension. These factors include: Based on the sales share of each competitor product, the proportion of the attribute differentiation contribution of each competitor product in the candidate dimension to the sum of the attribute differentiation contributions of all dimensions of the same competitor product, and the similarity of attribute information between each competitor product and the target product in the candidate dimension, the basic important factors of the candidate dimension attributes are obtained. The similarity of attribute information between each competing product and the target product in the candidate dimension is the negative correlation mapping value of the difference in attribute information between each competing product and the target product in the candidate dimension.

4. The intelligent generation method for product description text according to claim 2, characterized in that, The process of filtering comparable products across all dimensions based on differences in attribute information between the target product and competing products in the same dimension includes: Based on the differences in attribute information between the target product and each competitor's product in the candidate dimensions, the disparity coefficient between the target product and each competitor's product in the candidate dimensions is obtained. Competitor products whose differential coefficient is less than the preset differential threshold are identified as products of the same grade as the target product in the candidate dimension.

5. The intelligent generation method for product description text according to claim 4, characterized in that, The step of calculating the joint scarcity factor of any two dimensions based on the co-occurrence factors of the attributes of any two dimensions, the proportion of similar products in each of the two dimensions among all competing products, and the degree of dispersion of the basic importance factors of all dimension attributes, includes: Obtain the product of the proportion of similar products in all competing products in any two of the given two dimensions; Based on the relative magnitude of the product of the co-occurrence factors of attributes in any two dimensions and the degree of discrete distribution of the fundamental importance factors of all dimensions, the joint scarcity factor of any two dimensions is obtained.

6. The intelligent generation method for product description text according to claim 4, characterized in that, The step of obtaining the joint scarcity weight of the target product in any two dimensions based on the joint scarcity factor of any two dimensions and the difference between the basic importance factor of the attribute in any two dimensions and the mean of the basic importance factor of all attributes in all dimensions includes: Calculate the difference factor between the base importance factors of any two dimensional attributes and the mean of the base importance factors of all dimensional attributes; By combining the difference factor and the joint scarcity factor, the joint scarcity weight of the target product in any two dimensions is obtained.

7. The intelligent generation method for product description text according to claim 4, characterized in that, The degree of discrete distribution of the fundamental important factors of all dimensional attributes is the variance of the fundamental important factors of all dimensional attributes.

8. The intelligent generation method for product description text according to claim 1, characterized in that, The process of generating introductory text for the target product based on the emphasized descriptive factors includes: Sort all dimensions in descending order according to the emphasized descriptor to obtain the dimension sequence; Select the first preset number of dimensions from the dimension sequence as the core descriptive attributes; The core descriptive attributes and their corresponding emphasis descriptive factors are constructed into generation control information, which is then input into a trained language model to generate the introductory text for the target product.

9. A product description copywriting intelligent generation system, the system being used to implement the method of claim 1, characterized in that, The system includes: The attribute acquisition module is used to acquire attribute information of the target product and its competitors' products in various dimensions. The product filtering module is used to obtain the basic key factors of each dimension of attributes based on the differences in attribute information between competing products in the same dimension, the sales performance of competing products, and the overall differences in attribute information between the target product and competing products; and to filter products of the same grade in each dimension based on the differences in attribute information between the target product and competing products. The attribute analysis module is used to determine the key descriptive factors for each dimension based on the number of competing products that belong to the same product category in both pairs of dimensions, the proportion of the same product category in each dimension, and the corresponding basic important factors. The copywriting generation module is used to generate introductory copy for the target product based on the emphasized descriptive factors.

Images