CN117745337A

CN117745337A - Product generation method, device, computer equipment and storage medium

Info

Publication number: CN117745337A
Application number: CN202410045316.6A
Authority: CN
Inventors: 蔡龙兴; 赵驰; 于跃; 许宁; 于金楠
Original assignee: Industrial and Commercial Bank of China Ltd ICBC
Current assignee: Industrial and Commercial Bank of China Ltd ICBC
Priority date: 2024-01-12
Filing date: 2024-01-12
Publication date: 2024-03-22

Abstract

The application relates to a method, a device, a computer device and a storage medium for generating products, relates to the technical field of computers, and can be used in the technical field of finance or other related fields. The method comprises the following steps: acquiring target hot event data related to the attribute category of the reference product in a preset area; generating product description data of a target product under a reference product attribute category according to the target hot event data; acquiring product interaction data of associated products of a target product at different historical moments; based on the time sequence model, predicting interaction prediction data of a target product at different historical moments corresponding to future moments according to the product interaction data; and generating product content data of the target product according to the product description data and the interaction prediction data corresponding to the target product. By adopting the method, the generated product description data can be closer to the social hot spot, and the novelty of the generated target product is improved.

Description

Product generation method, device, computer equipment and storage medium

Technical Field

The present invention relates to the field of computer technology, and may be used in the field of financial technology or other related fields, and in particular, to a method, an apparatus, a computer device, and a storage medium for generating a product.

Background

With the improvement of life quality, the demands of people on financial products are increasing, and the current sales mode is greatly influenced by the innovation of the financial products, so that how to improve the innovation of the financial products is of great importance.

In the conventional technology, new financial products are often generated by analyzing sales conditions of historical financial products and selecting financial products with good sales conditions from the sales conditions for updating and iterating.

Although this method can produce new financial products, there is a problem in that the product novelty is low.

Disclosure of Invention

In view of the foregoing, it is desirable to provide a product generation method, apparatus, computer device, and storage medium that can improve product novelty.

In a first aspect, the present application provides a method of generating a product, comprising:

acquiring target hot event data related to the attribute category of the reference product in a preset area;

generating product description data of a target product under a reference product attribute category according to the target hot event data;

acquiring product interaction data of associated products of a target product at different historical moments;

based on the time sequence model, predicting interaction prediction data of a target product at different historical moments corresponding to future moments according to the product interaction data;

And generating product content data of the target product according to the product description data and the interaction prediction data corresponding to the target product.

In one embodiment, obtaining product interaction data corresponding to an associated product of a target product at different historical moments includes: acquiring product attribute data corresponding to the associated product of the target product at different historical moments; and selecting product interaction data related to the preset interaction task attribute from the product attribute data.

In one embodiment, selecting product interaction data related to a preset interaction task attribute from the product attribute data includes: respectively determining first correlation coefficients between product attribute data in different dimensions and preset interaction task attributes; and selecting the product attribute data with the first correlation coefficient larger than a preset first correlation threshold value as product interaction data.

In one embodiment, the method further comprises: and referring to other products in the product attribute category as associated products of the target product.

In one embodiment, the method further comprises: acquiring product description data of the associated product; determining a second correlation coefficient of product description data of the associated product and the target hot event data; and eliminating the associated products with the second correlation number smaller than a preset second correlation threshold value to update the associated products.

In one embodiment, generating product description data of a target product under a reference product attribute category according to target hot event data includes: based on the generated pre-training model, generating product subject data of the target product under the reference product attribute category according to the target hot event data; and generating product description data of the target product according to the product subject data of the target product based on the multi-mode content generation model.

In one embodiment, generating product description data of a target product according to product theme data of the target product includes: determining a product display category of the target product according to the reference product attribute category; and generating product description data of the target product according to the product subject data of the target product and the corresponding product display category based on the multi-mode content generation model.

In one embodiment, the product description data includes a product entity picture and/or a product text description; correspondingly, based on the multi-mode content generation model, generating product description data of the target product according to the product subject data of the target product and the corresponding product display category, including: if the product display category of the target product comprises an entity category, generating a model based on the multi-mode content, and generating a product entity picture of the target product according to the product subject data of the target product; if the product display category of the target product comprises the virtual category, generating a model based on the multi-mode content, and generating a product text description of the target product according to the product subject data of the target product.

In one embodiment, obtaining target hotspot event data related to a reference product attribute category in a preset area includes: acquiring event description data of a hot event in a preset area; feature data related to the reference product attribute category is extracted from the event description data as target hot event data.

In a second aspect, the present application also provides a product generating apparatus, comprising:

the hot spot acquisition module is used for acquiring target hot spot event data related to the attribute category of the reference product in the preset area;

the description acquisition module is used for generating product description data of the target product under the reference product attribute category according to the target hot event data;

the interaction acquisition module is used for acquiring product interaction data of the associated product of the target product at different historical moments;

the interaction prediction module is used for predicting interaction prediction data of a target product at different historical moments corresponding to future moments according to the product interaction data based on the time sequence model;

and the content generation module is used for generating product content data of the target product according to the product description data and the interaction prediction data corresponding to the target product.

In a third aspect, the present application also provides a computer device comprising a memory and a processor, the memory storing a computer program, the processor implementing the following steps when executing the computer program:

In a fourth aspect, the present application also provides a computer readable storage medium having stored thereon a computer program which when executed by a processor performs the steps of:

In a fifth aspect, the present application also provides a computer program product comprising a computer program which, when executed by a processor, performs the steps of:

The product generation method, the device, the computer equipment and the storage medium acquire target hot event data related to the reference product attribute category in a preset area; generating product description data of a target product under a reference product attribute category according to the target hot event data; acquiring product interaction data of associated products of a target product at different historical moments; based on the time sequence model, predicting interaction prediction data of a target product at different historical moments corresponding to future moments according to the product interaction data; and generating product content data of the target product according to the product description data and the interaction prediction data corresponding to the target product. Compared with the traditional technology, only the mode of updating and iterating the historical product is adopted, the target hot spot event data is considered, so that the generated product description data is closer to the social hot spot, meanwhile, the product interaction data at the future moment is predicted according to the product interaction data through a time sequence model, the interaction condition of the target product at the future moment can be predicted, and finally the product description data and the product interaction data are combined to generate the target product, so that the generated target product meets the current social requirement and the interaction requirement, the novelty of the product is improved, and the interaction relevance of the target product is ensured.

Drawings

In order to more clearly illustrate the embodiments of the present application or the technical solutions in the related art, the drawings that are required to be used in the embodiments or the related technical descriptions will be briefly described below, and it is obvious that the drawings in the following description are only some embodiments of the present application, and other drawings may be obtained according to the drawings without inventive effort for a person having ordinary skill in the art.

Fig. 1 is an application environment diagram of a product generating method provided in this embodiment;

fig. 2 is a schematic flow chart of a first product generating method according to the embodiment;

fig. 3 is a schematic flow chart of selecting product interaction data according to the embodiment;

FIG. 4 is a schematic flow chart of updating associated products according to the present embodiment;

FIG. 5 is a flow chart of generating product description data according to the present embodiment;

FIG. 6 is a flow chart of a second method for generating a product according to the present embodiment;

fig. 7 is a block diagram of a product generating apparatus according to the present embodiment;

fig. 8 is a block diagram of a product generating apparatus according to the present embodiment;

fig. 9 is an internal structure diagram of a computer device according to the present embodiment.

Detailed Description

In order to make the objects, technical solutions and advantages of the present application more apparent, the present application will be further described in detail with reference to the accompanying drawings and examples. It should be understood that the specific embodiments described herein are for purposes of illustration only and are not intended to limit the present application.

The product generation method provided by the embodiment of the application can be applied to an application environment shown in fig. 1. Wherein the terminal 102 communicates with the server 104 via a network. The data storage system may store data that the server 104 needs to process. The data storage system may be integrated on the server 104 or may be located on a cloud or other network server. Target hot event data related to the reference product attribute category in a preset area can be acquired through the terminal 102/the server 104; generating product description data of a target product under a reference product attribute category according to the target hot event data through the terminal 102/the server 104; acquiring product interaction data of the associated product of the target product at different historical moments through the terminal 102/the server 104; based on the time sequence model, the terminal 102/the server 104 predicts interactive prediction data of a target product at different historical moments corresponding to future moments according to the product interactive data; product content data of the target product is generated by the terminal 102/server 104 according to the product description data and the interaction prediction data corresponding to the target product. The terminal 102 may be, but not limited to, various personal computers, notebook computers, smart phones, tablet computers, internet of things devices, and portable wearable devices, where the internet of things devices may be smart speakers, smart televisions, smart air conditioners, smart vehicle devices, and the like. The portable wearable device may be a smart watch, smart bracelet, headset, or the like. The server 104 may be implemented as a stand-alone server or as a server cluster of multiple servers.

In an exemplary embodiment, as shown in fig. 2, a product generation method is provided, and an example of application of the method to the server 104 in fig. 1 is described, including the following steps S201 to S205.

Wherein:

s201, acquiring target hot event data related to the attribute category of the reference product in a preset area.

The preset area may be a network platform including a hotspot event, a platform release area, or an occurrence area of the hotspot event, and may be set or adjusted by a technician according to actual needs or experience. The reference product attribute category may be a category that classifies a product based on attribute characteristics of the reference product. In one particular implementation, if the product is a financial product, the corresponding reference product attribute category may include at least one of personal financial, loan, investment, and the like. The target hotspot event data may include event content-related data for the hotspot event.

In one optional implementation manner, a hot event corresponding to the reference product attribute category may be selected from a preset area, and the event content of the hot event is used as target hot event data.

In another optional implementation manner, event description data of a hot event in a preset area may be obtained; feature data related to the reference product attribute category is extracted from the event description data as target hot event data.

Specifically, each hot event can be selected from a preset area, word segmentation processing is carried out on the hot event, a word segmentation result of the hot event is obtained, and the word segmentation result is used as event description data; performing feature coding processing on different event description data to obtain event description features of corresponding event description data; determining a correlation coefficient between each event description feature and a reference product attribute category; and selecting event description characteristics with the correlation coefficient larger than a preset correlation coefficient threshold value as target hot event data. The correlation coefficient in the present embodiment may include at least one of a pearson correlation coefficient, a spearman scale correlation coefficient, and the like. The feature encoding process may be implemented by at least one feature encoding method in the related art, which is not limited in this application.

The method for determining the target hot event data has the advantages that the event description data can be analyzed from the hot event, and the correlation between each event description data and the reference product attribute category is determined, so that the target hot event data related to the reference product attribute category is extracted from the hot event more accurately, the product description data with stronger correlation with the hot event can be generated based on the target hot event data, and the innovation and hot correlation of the target product are further improved.

Illustratively, there are hot events: "the first gift of the mother who the son's mother sends is carnation and necklace", the event description data for determining the hot event includes: "mother's day", "mother", "carnation", "necklace", "first gift", "son". Assume that the reference product attribute category is "personal finance category"; therefore, a correlation coefficient between the attribute category of the reference product and each event description data (such as a correlation coefficient between "personal finance category" and "mother's day", a correlation coefficient between "personal finance category" and "necklace", etc.) is determined, wherein the event description data with the correlation coefficient higher than the preset correlation coefficient threshold value is determined to be "necklace", and then the coding feature corresponding to the "necklace" is taken as the target hot event data.

S202, generating product description data of a target product under a reference product attribute category according to the target hot event data.

The target product may be a new product that is expected to be promoted for sale at a future time, such as a financial product in a financial institution.

In one alternative implementation manner, the target hot event data and the reference product attribute category may be input into a pre-trained product description generation model, where the product description generation model may parse the target hot event data and the reference product attribute category, thereby generating description data related to the reference product attribute category, and take the product description data as product description data of the target product.

In another alternative implementation, the product description data of the target product under the reference product attribute category may be generated according to the target hotspot event data based on the neural network model with the content generation function.

S203, acquiring product interaction data of the associated product of the target product at different historical moments.

The related product may be a product having a relationship with the target product in any dimension, for example, may be the same product as at least one of a generating subject (e.g., generating participant), a generating area, a publishing subject (e.g., publishing participant), and a publishing area of the target product.

In an alternative implementation, the associated product may be other products of the same reference product attribute category as the target product. Optionally, determining a reference product attribute category of the target product; taking other products under the attribute category of the reference product as related products of the target product; and acquiring product interaction data of the associated product at different historical moments.

The product interaction data may be data associated with a product interaction, and optionally, the product interaction data may include at least one of interaction data amount (such as marketing amount), customer consumption data (i.e. consumption data of an existing product by a customer), product scale data, product influence data (such as awareness data and customer loyalty data), interaction channel data (i.e. sales channel of an existing product), and product value data (such as product price), etc.

S204, based on the time sequence model, according to the product interaction data, the interaction prediction data of the target product at different historical moments corresponding to future moments are predicted.

The time series model (TimesNet model) may be a model for predicting interaction condition of the target product at a future time, and it should be noted that the model may be a two-dimensional convolution kernel model. The interaction forecast data may be interaction data for the target product at a future time, such as a visual sales curve for the financial product at the future time.

Optionally, product interaction data of the associated product at different historical moments is input into a time sequence model, and the time sequence model analyzes the product interaction data, so that interaction prediction data of the target product at future moments is obtained through prediction.

It should be noted that, the time sequence model can remodel sequences in two-dimensional space, namely simulate the changes in period and period, fold one-dimensional time sequences based on multiple periods, and simultaneously rapidly identify two-dimensional time sequences (such as daily period and annual period) in product interaction data through fourier transformation, so as to extract two-dimensional time sequence change characterization (2D presentation), meanwhile, since each column and row of the two-dimensional time sequence respectively reflect the time sequence changes in period and period, two-dimensional time sequence change (samples 2D-variations) can be determined, and finally, in order to ensure the accurate fusion of multiple periods, the two-dimensional time sequence change characterization is developed into one-dimensional time sequence. Therefore, a blank technology for analyzing interaction trend of some products in practical application is supplemented, for example, the tendency of the products of personal finance targets in quarters is different due to the influence of factors such as holidays and the like.

S205, generating product content data of the target product according to the product description data and the interaction prediction data corresponding to the target product.

Wherein the product content data may be data for describing the product content of the target product.

Optionally, determining product description data and interaction prediction data corresponding to the target product, and performing fusion processing on the product description data and the interaction prediction data, so as to obtain content data, and taking the content data as product content data of the target product.

It should be noted that, in step S202 of this embodiment, the interactive propulsion scheme of the target product may be generated by generating the pre-training model according to the target hotspot event data and the reference product attribute category. Therefore, the embodiment can also generate the product content data of the target product according to the product description data, the interaction prediction data and the interaction propulsion scheme corresponding to the target product.

According to the product generation method, target hot event data related to the reference product attribute category in the preset area is obtained; generating product description data of a target product under a reference product attribute category according to the target hot event data; acquiring product interaction data of associated products of a target product at different historical moments; based on the time sequence model, predicting interaction prediction data of a target product at different historical moments corresponding to future moments according to the product interaction data; and generating product content data of the target product according to the product description data and the interaction prediction data corresponding to the target product. Compared with the traditional technology, only the mode of updating and iterating the historical product is adopted, the target hot spot event data is considered, so that the generated product description data is closer to the social hot spot, meanwhile, the product interaction data at the future moment is predicted according to the product interaction data through a time sequence model, the interaction condition of the target product at the future moment can be predicted, and finally the product description data and the product interaction data are combined to generate the target product, so that the generated target product meets the current social requirement and the interaction requirement, the novelty of the product is improved, and the interaction relevance of the target product is ensured.

In an alternative embodiment, the training phase of the time series model can also be realized by optimizingA modeling algorithm (e.g., a random gradient descent algorithm) adjusts model parameters of the time series model to minimize prediction errors. Wherein the random gradient descent algorithm (Stochastic Gradient Descent, SGD) is a commonly used optimization algorithm, i.e. given a loss function L (θ) (where θ is a model parameter of a time series model), a single sample or small batch of samples is randomly selected to determine the gradient of the loss function (for a single sample, using the sign)Representing the gradient; for small batches of samples the symbol +.>Representing the gradient), and then using the learning rate η to adjust the step size of the gradient (e.g., subtracting the learning rate from the gradient direction and multiplying the gradient to reduce the loss function) until a specified iteration termination condition is reached (e.g., a maximum number of iterations is reached or the loss function converges), thereby completing the updating of the model parameter θ. It is worth noting that the updating of the model parameters can be completed by adopting small batches of samples, and as the small batches of samples can ensure the stability of the time sequence model in the model parameter adjustment process, the calculated amount of model parameter adjustment can be reduced by utilizing a batch of sample matrix, and the robustness and the training efficiency of the trained time sequence model can be obviously improved.

Illustratively, the random gradient descent for a single sample, equation (1-1), is:

wherein, theta { t+1} is a model parameter of the time series model in the t+1th iteration process, theta { t } is a model parameter of the time series model in the t iteration process, eta is a learning rate,for the gradient of the time series model in the t-th iteration, L (θ { t }) is the loss function of the time series model in the t-th iteration.

Illustratively, the random gradient descent equation (1-2) for small sample updates is:

wherein, theta { t+1} is a model parameter of the time series model in the t+1th iteration process, theta { t } is a model parameter of the time series model in the t iteration process, eta is a learning rate,is the gradient of the time series model in the t-th iteration process.

FIG. 3 is a flow chart of selecting product interaction data in one embodiment. The embodiment provides an optional way for selecting product interaction data, which comprises the following steps:

s301, obtaining product attribute data corresponding to the associated product of the target product at different historical moments.

The product attribute data may be product attribute data of the associated product, such as applicable age, sales area, sales mode, etc. of the associated product.

Optionally, determining a reference product attribute category of the target product, and taking other products in the reference product attribute category as associated products to obtain product attribute data of the associated products at different historical moments.

S302, selecting product interaction data related to preset interaction task attributes from the product attribute data.

Optionally, determining first correlation coefficients between product attribute data in different dimensions and preset interaction task attributes respectively; and selecting the product attribute data with the first correlation coefficient larger than a preset first correlation threshold value as product interaction data. The first correlation threshold may be set by a skilled person according to needs or experience, or may be repeatedly determined through a plurality of experiments, which is not limited in any way in the present application.

For example, product attribute data of the associated product under different dimensions can be determined, a first correlation coefficient between each product attribute data and a preset interaction task attribute is determined through at least one of a pearson correlation coefficient and a spearman level correlation coefficient, and product attribute data with the first correlation coefficient larger than a preset first correlation threshold value is selected to be used as product interaction data of the target product.

It is noted that when the first correlation coefficient is a correlation coefficient of a different class, the determination of the final first correlation coefficient may be performed by means of weighted summation. The weights corresponding to the correlation coefficients of different classes can be set by a technician according to needs or experience, or can be repeatedly determined through extensive experiments, and the application is not limited in any way.

According to the technical scheme, the product attribute data with higher correlation with the interaction task attribute is more accurately determined in a first correlation coefficient quantification mode, so that the product interaction data of the associated product can be more accurately determined according to the product attribute data.

According to the technical scheme, the product interaction data meeting the interaction task attribute can be more accurately determined by acquiring the product attribute data corresponding to the associated product of the target product at different historical moments and selecting the product interaction data related to the preset interaction task attribute from the product attribute data, so that the interaction prediction data of the target product at the future moment can be more accurately determined for the subsequent interaction prediction data based on the product interaction data.

FIG. 4 is a flow diagram of updating an associated product in one embodiment. In order to more accurately determine product interaction data of a target product, in this embodiment, when selecting an associated product of the target product, correlation between the associated product and the target product and correlation with a hot event should be considered at the same time, and specifically, this embodiment provides an alternative way for updating the associated product, which includes the following steps:

S401, taking other products under the reference product attribute category as related products of the target product.

Optionally, determining a reference product attribute category of the target product, and selecting other products under the reference product attribute category as associated products.

S402, acquiring product description data of the associated product.

Optionally, analyzing the associated product to obtain product description data of the associated product.

S403, determining a second correlation coefficient of the product description data of the associated product and the target hot event data.

Optionally, product description data of the associated product and each target hot event data are acquired, and a correlation coefficient between each product description data and the target hot event data is determined as a second correlation coefficient. Wherein the correlation coefficient may include at least one of a pearson correlation coefficient, a spearman scale correlation coefficient, and the like.

In an alternative implementation, event description data in the target hotspot event data may be extracted; a second correlation coefficient is determined for the product description data and the event description data for the associated product.

Wherein the event description data is used for characterizing event content of the hot event. Illustratively, there are hot events: "the first gift of the mother who the son's mother sends is carnation and necklace", the event description data for determining the hot event includes: at least one of "mother", "carnation", "necklace", "first gift", and "son", etc.

In another alternative implementation, hotspot environment data in the target hotspot event data may be extracted; a second relationship number is determined for the product description data of the associated product and the hotspot environment data.

Wherein the hot spot environment data is used for characterizing an atmosphere environment of a hot spot event, such as holidays and the like. For example, if there is a hot event: the first gift of the mother who sends the child's mother's day is carnation and necklace, and the hot spot environmental data of the hot spot event is determined as ' mother's day '.

It is noted that when the second correlation coefficient is a correlation coefficient of a different class, the determination of the final second correlation coefficient may be performed by means of weighted summation. The weights corresponding to the correlation coefficients of different classes can be set by a technician according to needs or experience, or can be repeatedly determined through extensive experiments, and the application is not limited in any way.

S404, eliminating the associated products with the second correlation number smaller than a preset second correlation threshold value to update the associated products.

Optionally, based on the determined second correlation numbers of the product description data and the target hot event data, selecting an associated product with the second correlation number smaller than a preset second correlation threshold value from the second correlation numbers, and removing the associated product, so that the associated product is updated, and the updated associated product does not contain a product with lower correlation with the hot event.

According to the method for updating the associated product, other products of the attribute category of the reference product with the target product are preferentially selected as the associated product, so that the attribute category consistency of the associated product and the target product is ensured; and then removing the associated product with lower correlation according to the correlation between the product description data of the associated product and the target hot event data, thereby ensuring the correlation between the associated product and the hot event. By adopting the method, the reference product attribute category of the reserved associated product can be the same as that of the target product, the correlation with the hot event is higher, the attribute correlation and the hot correlation of the selected associated product and the target product are improved, further, the product attribute data corresponding to the associated product at different historical moments is used as the product attribute data of the target product at different historical moments, the interactive prediction data of the target product at the subsequent future moment is determined, and the interactive prediction data is more accurate and has the referential property.

FIG. 5 is a flow diagram of generating product description data in one embodiment. The present embodiment provides an alternative way of generating product description data, comprising the steps of:

S501, based on the generated pre-training model, generating product theme data of the target product under the reference product attribute category according to the target hot event data.

The product theme data may be understood as a theme summary of the target product, and briefly describes the product content of the target product.

For example, an access interface of a generative pre-training model may be invoked, while entering the access rights and keys of the model; the computing equipment carrying the generated pre-training model authenticates the access right and the secret key and feeds back an authentication result; generating a hypertext transfer protocol (Hypertext Transfer Protocol, HTTP) request according to the target hot event data and the reference product attribute category under the condition that authentication is passed, and calling an access interface of the generated pre-training model based on the HTTP request; the method comprises the steps of carrying a computing device of a generating type pre-training model, analyzing an HTTP request to obtain target hot event data and a reference product attribute type, inputting an analysis result into the generating type pre-training model, generating theme data related to the reference product attribute type, and taking the theme data as product theme data of a target product to be generated. Wherein the target product belongs to the category of the attribute of the reference product.

In an alternative embodiment, the Generative Pre-training model may be a natural language processing model, such as a GPT-4 (Generative Pre-trained Transformer) natural language processing model. The GPT-4 natural language processing model has the advantages that the model can understand and generate natural language and is suitable for various natural language processing tasks, so that the product subject data of a target product under a reference product attribute category can be accurately generated according to the target hot spot event data.

It should be noted that, in this embodiment, according to the target hotspot event data and the reference product attribute category, an interactive propulsion scheme for the target product, for example, a sales description scheme for the financial product, may be generated through a pre-training model.

S502, generating a model based on the multi-mode content, and generating product description data of the target product according to the product subject data of the target product.

The product description data may be understood as the product content of the target product, and describes the product content of the target product in detail.

Wherein the multimodal content generating model may be a contrast language-picture pre-training (CLIP) model. The use of a multimodal model has the advantage that cross-modal reasoning and information processing can be performed, for example, understanding the relevant text description in the case of a given image, or the relevant image content in the case of a given text, so that product description data can be accurately generated from the product topic data of a target product.

In one alternative manner, the embodiment may continuously call the access interface of the multi-mode content generation model, and simultaneously input the access right and the key of the model; and the computing equipment bearing the multi-mode content generation model authenticates the access right and the secret key and feeds back an authentication result. Under the condition that authentication is passed, the product theme data are input into a multi-mode content generation model, the multi-mode content generation model is used for analyzing the product theme data, so that description data are generated, and the description data are used as product description data of a target product.

In another optional implementation manner, the embodiment may further determine a product display category of the target product according to the reference product attribute category, generate a model based on the multi-mode content, and generate product description data of the target product according to the product subject data of the target product and the corresponding product display category.

The product display category may be a category for displaying a product, and optionally, the product display category includes an entity class, a virtual class, and the like. For example, the virtual class may be a language class.

Specifically, determining a product display category of a reference product attribute category, taking the product display category as a product display category of a target product, calling an access interface of a multi-mode content generation model, and simultaneously inputting an access right and a key of the model; and the computing equipment bearing the multi-mode content generation model authenticates the access right and the secret key and feeds back an authentication result. Under the condition that authentication is passed, the product theme data of which the product display category is determined is input into a multi-mode content generation model, the product theme data and the product display category are analyzed through the multi-mode content generation model, so that description data are generated, and the description data are used as product description data of a target product.

The multi-mode content generation model is adopted to generate the product description data, and has the advantages that the product display types of target products are considered, the target products of different display types are ensured, different product description data can be generated according to the needs, the richness and the diversity of the product description data are improved, and meanwhile, the data transmission quantity between the multi-mode content generation model and the computing equipment for executing the product generation method caused by the unnecessary generation of the product description data is reduced.

It should be noted that, the determination manner of the product display category with reference to the product attribute category may be set by a technician according to needs or experience, or may be repeatedly determined through a lot of experiments, which is not limited in this application.

Further, if the product display category of the target product comprises an entity category, generating a model based on the multi-mode content, and generating a product entity picture of the target product according to the product subject data of the target product; if the product display category of the target product comprises the virtual category, generating a model based on the multi-mode content, and generating a product text description of the target product according to the product subject data of the target product.

It should be noted that the single product corresponding product display category may be an entity category or a virtual category, or a combination of an entity category and a virtual category, which is not limited in this application.

In an alternative embodiment, the calling frequency of the access interface, the corresponding frequency and the application degree integrated by the access interface corresponding to the multi-mode content generation model in the embodiment and the interaction frequency of the model can be set as required.

In an alternative embodiment, in the training stage of the multi-modal content generation model, the multi-modal content generation model can be trained through a large amount of image-text pair data (such as images and corresponding text descriptions), and the multi-modal content generation model can learn the correlation between the images and the texts without locally labeling the images or specially labeling the texts, so that the training efficiency of the model is improved. The embodiment can also perform model training on the multi-modal content generation model through contrast learning, so that the model can better understand and represent semantic association between the image and the text. Meanwhile, because the multi-modal content generation model does not depend on labeling data in a specific field, the embodiment can also perform model training on the multi-modal content generation model through zero sample learning, so that the multi-modal content generation model can infer and understand new image and text pairs even under the condition that specific tasks are not specially trained.

According to the technical scheme, the generation of the product description data is comprehensively carried out by introducing the generation pre-training model and the multi-mode content generation model, the product subject data related to the reference product attribute category and the hot event can be preferentially generated, and then the product description data is further generated according to the product subject data, so that the description data generation of the details after summarization is realized, the generation direction of the product description data is conveniently and integrally controlled, and meanwhile, the correlation of the target product corresponding to the product description data, the hot event and the reference product attribute category is improved. In addition, the product description data is generated based on the existing generation type pre-training model and the multi-mode content generation model, a large amount of sample data is not required to be acquired, complicated model training is not required to be carried out, convenience and universality of the generation process of the product description data are improved, and labor cost and time cost are reduced.

In one embodiment, this embodiment gives an alternative way of generating a product, and the method is applied to a server for illustration. As shown in fig. 6, the method includes the steps of:

s601, acquiring event description data of a hot event in a preset area.

S602, extracting feature data related to the attribute category of the reference product from the event description data as target hot event data.

S603, generating product description data of the target product under the reference product attribute category according to the target hot event data.

S604, taking other products under the reference product attribute category as associated products of the target product.

S605, acquiring product description data of the associated product.

S606, determining a second correlation coefficient of the product description data of the associated product and the target hot event data.

S607, eliminating the associated products with the second correlation number smaller than the preset second correlation threshold value to update the associated products.

S608, obtaining product attribute data corresponding to the associated product of the target product at different historical moments.

S609, determining first correlation coefficients between product attribute data in different dimensions and preset interaction task attributes respectively.

S610, selecting product attribute data with the first correlation coefficient larger than a preset first correlation threshold value as product interaction data.

S611, based on the time sequence model and the generation type pre-training model, generating the product subject data of the target product under the reference product attribute category according to the target hot event data.

S612, determining the product display category of the target product according to the reference product attribute category.

S613, generating product description data of the target product according to the product subject data of the target product and the corresponding product display category based on the multi-mode content generation model.

Wherein the product description data comprises a product entity picture and/or a product text description.

Optionally, if the product display category of the target product includes an entity category, generating a model based on the multi-mode content, and generating a product entity picture of the target product according to the product subject data of the target product; if the product display category of the target product comprises the virtual category, generating a model based on the multi-mode content, and generating a product text description of the target product according to the product subject data of the target product.

S614, product content data of the target product is generated according to the product description data and the interaction prediction data corresponding to the target product.

It should be understood that, although the steps in the flowcharts related to the embodiments described above are sequentially shown as indicated by arrows, these steps are not necessarily sequentially performed in the order indicated by the arrows. The steps are not strictly limited to the order of execution unless explicitly recited herein, and the steps may be executed in other orders. Moreover, at least some of the steps in the flowcharts described in the above embodiments may include a plurality of steps or a plurality of stages, which are not necessarily performed at the same time, but may be performed at different times, and the order of the steps or stages is not necessarily performed sequentially, but may be performed alternately or alternately with at least some of the other steps or stages.

Based on the same inventive concept, the embodiment of the application also provides a product generating device for realizing the above related product generating method. The implementation of the solution provided by the device is similar to the implementation described in the above method, so the specific limitation in one or more embodiments of the product generating device provided below may refer to the limitation of the product generating method hereinabove, and will not be repeated here.

In an exemplary embodiment, as shown in fig. 7, there is provided a product generating apparatus 1, comprising: a hotspot acquisition module 10, a description acquisition module 11, an interaction acquisition module 12, an interaction prediction module 13, and a content generation module 14, wherein:

the hotspot obtaining module 10 is configured to obtain target hotspot event data related to a reference product attribute category in a preset area;

the description acquisition module 11 is configured to generate product description data of a target product under a reference product attribute category according to the target hotspot event data;

the interaction acquisition module 12 is used for acquiring product interaction data of the associated product of the target product at different historical moments;

the interaction prediction module 13 is configured to predict interaction prediction data of a target product at different historical moments corresponding to future moments according to product interaction data based on a time sequence model;

The content generating module 14 is configured to generate product content data of the target product according to the product description data and the interaction prediction data corresponding to the target product.

In one embodiment, the interaction acquisition module 12 of FIG. 7 includes: the attribute acquisition unit is used for acquiring product attribute data corresponding to the associated product of the target product at different historical moments; and the interaction selection unit is used for selecting product interaction data related to the preset interaction task attribute from the product attribute data.

In one embodiment, the interaction selection unit comprises: the coefficient determination subunit is used for respectively determining first correlation coefficients between product attribute data under different dimensions and preset interaction task attributes; and the interaction selection subunit is used for selecting the product attribute data with the first correlation coefficient larger than a preset first correlation threshold value as the product interaction data.

In one embodiment, as shown in fig. 8, the product generating apparatus 1 in fig. 7 includes: the product determining module 15 is configured to take other products in the reference product attribute category as associated products of the target product. A product update module 16 for obtaining product description data of the associated product; determining a second correlation coefficient of product description data of the associated product and the target hot event data; and eliminating the associated products with the second correlation number smaller than a preset second correlation threshold value to update the associated products.

In one embodiment, the description acquisition module 11 in fig. 7 includes: the theme determining unit is used for generating product theme data of the target product under the reference product attribute category according to the target hot event data based on the generated pre-training model; and the description generating unit is used for generating a model based on the multi-mode content and generating product description data of the target product according to the product subject data of the target product.

In one embodiment, the description generation unit includes: a category determination subunit, configured to determine a product display category of the target product according to the reference product attribute category; the description generation subunit is used for generating a model based on the multi-mode content and generating product description data of the target product according to the product subject data of the target product and the corresponding product display category.

In one embodiment, the product description data includes a product entity picture and/or a product text description; correspondingly, the description generation subunit is further configured to generate a product entity picture of the target product based on the multi-mode content generation model if the product display category of the target product includes an entity category, and according to the product subject data of the target product; if the product display category of the target product comprises the virtual category, generating a model based on the multi-mode content, and generating a product text description of the target product according to the product subject data of the target product.

In one embodiment, the hotspot acquisition module 10 in fig. 7 comprises: the description acquisition unit is used for acquiring event description data of the hot events in the preset area; and the feature extraction unit is used for extracting feature data related to the attribute category of the reference product from the event description data as target hot event data.

The respective modules in the above-described product generating apparatus may be implemented in whole or in part by software, hardware, and combinations thereof. The above modules may be embedded in hardware or may be independent of a processor in the computer device, or may be stored in software in a memory in the computer device, so that the processor may call and execute operations corresponding to the above modules.

In an exemplary embodiment, a computer device, which may be a terminal, is provided, and an internal structure thereof may be as shown in fig. 9. The computer device includes a processor, a memory, an input/output interface, a communication interface, a display unit, and an input means. The processor, the memory and the input/output interface are connected through a system bus, and the communication interface, the display unit and the input device are connected to the system bus through the input/output interface. Wherein the processor of the computer device is configured to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage media. The input/output interface of the computer device is used to exchange information between the processor and the external device. The communication interface of the computer device is used for carrying out wired or wireless communication with an external terminal, and the wireless mode can be realized through WIFI, a mobile cellular network, NFC (near field communication) or other technologies. The computer program is executed by a processor to implement a product generation method. The display unit of the computer device is used for forming a visual picture, and can be a display screen, a projection device or a virtual reality imaging device. The display screen can be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer equipment can be a touch layer covered on the display screen, can also be a key, a track ball or a touch pad arranged on the shell of the computer equipment, and can also be an external keyboard, a touch pad or a mouse and the like.

It will be appreciated by those skilled in the art that the structure shown in fig. 9 is merely a block diagram of a portion of the structure associated with the present application and is not limiting of the computer device to which the present application applies, and that a particular computer device may include more or fewer components than shown, or may combine some of the components, or have a different arrangement of components.

In one exemplary embodiment, a computer device is provided comprising a memory and a processor, the memory having stored therein a computer program, the processor when executing the computer program performing the steps of:

In one embodiment, the processor when executing the computer program further performs the steps of: acquiring product attribute data corresponding to the associated product of the target product at different historical moments; and selecting product interaction data related to the preset interaction task attribute from the product attribute data.

In one embodiment, the processor when executing the computer program further performs the steps of: respectively determining first correlation coefficients between product attribute data in different dimensions and preset interaction task attributes; and selecting the product attribute data with the first correlation coefficient larger than a preset first correlation threshold value as product interaction data.

In one embodiment, the processor when executing the computer program further performs the steps of: and referring to other products in the product attribute category as associated products of the target product.

In one embodiment, the processor when executing the computer program further performs the steps of: acquiring product description data of the associated product; determining a second correlation coefficient of product description data of the associated product and the target hot event data; and eliminating the associated products with the second correlation number smaller than a preset second correlation threshold value to update the associated products.

In one embodiment, the processor when executing the computer program further performs the steps of: based on the generated pre-training model, generating product subject data of the target product under the reference product attribute category according to the target hot event data; and generating product description data of the target product according to the product subject data of the target product based on the multi-mode content generation model.

In one embodiment, the processor when executing the computer program further performs the steps of: determining a product display category of the target product according to the reference product attribute category; and generating product description data of the target product according to the product subject data of the target product and the corresponding product display category based on the multi-mode content generation model.

In one embodiment, the processor when executing the computer program further performs the steps of: if the product display category of the target product comprises an entity category, generating a model based on the multi-mode content, and generating a product entity picture of the target product according to the product subject data of the target product; if the product display category of the target product comprises the virtual category, generating a model based on the multi-mode content, and generating a product text description of the target product according to the product subject data of the target product.

In one embodiment, the processor when executing the computer program further performs the steps of: acquiring event description data of a hot event in a preset area; feature data related to the reference product attribute category is extracted from the event description data as target hot event data.

In one embodiment, a computer readable storage medium is provided having a computer program stored thereon, which when executed by a processor, performs the steps of:

In one embodiment, the computer program when executed by the processor further performs the steps of: acquiring product attribute data corresponding to the associated product of the target product at different historical moments; and selecting product interaction data related to the preset interaction task attribute from the product attribute data.

In one embodiment, the computer program when executed by the processor further performs the steps of: respectively determining first correlation coefficients between product attribute data in different dimensions and preset interaction task attributes; and selecting the product attribute data with the first correlation coefficient larger than a preset first correlation threshold value as product interaction data.

In one embodiment, the computer program when executed by the processor further performs the steps of: and referring to other products in the product attribute category as associated products of the target product.

In one embodiment, the computer program when executed by the processor further performs the steps of: acquiring product description data of the associated product; determining a second correlation coefficient of product description data of the associated product and the target hot event data; and eliminating the associated products with the second correlation number smaller than a preset second correlation threshold value to update the associated products.

In one embodiment, the computer program when executed by the processor further performs the steps of: based on the generated pre-training model, generating product subject data of the target product under the reference product attribute category according to the target hot event data; and generating product description data of the target product according to the product subject data of the target product based on the multi-mode content generation model.

In one embodiment, the computer program when executed by the processor further performs the steps of: determining a product display category of the target product according to the reference product attribute category; and generating product description data of the target product according to the product subject data of the target product and the corresponding product display category based on the multi-mode content generation model.

In one embodiment, the computer program when executed by the processor further performs the steps of: if the product display category of the target product comprises an entity category, generating a model based on the multi-mode content, and generating a product entity picture of the target product according to the product subject data of the target product; if the product display category of the target product comprises the virtual category, generating a model based on the multi-mode content, and generating a product text description of the target product according to the product subject data of the target product.

In one embodiment, the computer program when executed by the processor further performs the steps of: acquiring event description data of a hot event in a preset area; feature data related to the reference product attribute category is extracted from the event description data as target hot event data.

In one embodiment, a computer program product is provided comprising a computer program which, when executed by a processor, performs the steps of:

It should be noted that, the data (including, but not limited to, data for analysis, stored data, presented data, etc.) referred to in the present application are all information and data authorized by the user or sufficiently authorized by each party, and the collection, use, and processing of the relevant data are required to meet the relevant regulations.

Those skilled in the art will appreciate that implementing all or part of the above described methods may be accomplished by way of a computer program stored on a non-transitory computer readable storage medium, which when executed, may comprise the steps of the embodiments of the methods described above. Any reference to memory, database, or other medium used in the various embodiments provided herein may include at least one of non-volatile and volatile memory. The nonvolatile Memory may include Read-Only Memory (ROM), magnetic tape, floppy disk, flash Memory, optical Memory, high density embedded nonvolatile Memory, resistive random access Memory (ReRAM), magnetic random access Memory (Magnetoresistive Random Access Memory, MRAM), ferroelectric Memory (Ferroelectric Random Access Memory, FRAM), phase change Memory (Phase ChangeMemory, PCM), graphene Memory, and the like. Volatile memory can include random access memory (Random Access Memory, RAM) or external cache memory, and the like. By way of illustration, and not limitation, RAM can be in the form of a variety of forms, such as Static Random access memory (Static Random access memory AccessMemory, SRAM) or dynamic Random access memory (Dynamic Random Access Memory, DRAM), and the like. The databases referred to in the various embodiments provided herein may include at least one of relational databases and non-relational databases. The non-relational database may include, but is not limited to, a blockchain-based distributed database, and the like. The processors referred to in the embodiments provided herein may be general purpose processors, central processing units, graphics processors, digital signal processors, programmable logic units, quantum computing-based data processing logic units, etc., without being limited thereto.

The technical features of the above embodiments may be arbitrarily combined, and all possible combinations of the technical features in the above embodiments are not described for brevity of description, however, as long as there is no contradiction between the combinations of the technical features, they should be considered as the scope of the description.

The above examples only represent a few embodiments of the present application, which are described in more detail and are not to be construed as limiting the scope of the present application. It should be noted that it would be apparent to those skilled in the art that various modifications and improvements could be made without departing from the spirit of the present application, which would be within the scope of the present application. Accordingly, the scope of protection of the present application shall be subject to the appended claims.

Claims

1. A method of generating a product, the method comprising:

generating product description data of the target product under the reference product attribute category according to the target hotspot event data;

acquiring product interaction data of the associated product of the target product at different historical moments;

Based on a time sequence model, predicting interaction prediction data of the target product at future moments corresponding to different historical moments according to the product interaction data;

2. The method of claim 1, wherein the obtaining product interaction data corresponding to the associated product of the target product at different historical moments comprises:

acquiring product attribute data corresponding to the associated product of the target product at different historical moments;

and selecting product interaction data related to the preset interaction task attribute from the product attribute data.

3. The method according to claim 2, wherein selecting product interaction data related to a preset interaction task attribute from the product attribute data comprises:

respectively determining first correlation coefficients between product attribute data in different dimensions and preset interaction task attributes;

and selecting the product attribute data with the first correlation coefficient larger than a preset first correlation threshold value as the product interaction data.

4. A method according to any one of claims 1-3, wherein the method further comprises:

and taking other products in the reference product attribute category as associated products of the target product.

5. The method according to claim 4, wherein the method further comprises:

acquiring product description data of the associated product;

determining a second correlation coefficient of the product description data of the associated product and the target hot event data;

and eliminating the associated products with the second correlation number smaller than a preset second correlation threshold value to update the associated products.

6. A method according to any one of claims 1-3, wherein said generating product description data of a target product under said reference product attribute category from said target hotspot event data comprises:

based on a generated pre-training model, generating product theme data of a target product under the reference product attribute category according to the target hot event data;

and generating product description data of the target product according to the product subject data of the target product based on the multi-mode content generation model.

7. The method of claim 6, wherein generating product description data of the target product from product topic data of the target product based on the multi-modal content generation model comprises:

Determining a product display category of the target product according to the reference product attribute category;

and generating product description data of the target product according to the product subject data of the target product and the corresponding product display category based on the multi-mode content generation model.

8. The method of claim 7, wherein the product description data comprises a product entity picture and/or a product text description; correspondingly, the generating model based on the multi-mode content generates product description data of the target product according to the product subject data of the target product and the corresponding product display category, and the generating model comprises the following steps:

if the product display category of the target product comprises an entity category, generating a model based on multi-mode content, and generating a product entity picture of the target product according to the product subject data of the target product;

and if the product display category of the target product comprises a virtual category, generating a model based on the multi-mode content, and generating a product text description of the target product according to the product subject data of the target product.

9. A method according to any one of claims 1-3, wherein the obtaining target hotspot event data associated with the reference product attribute category in the preset area comprises:

Acquiring event description data of a hot event in a preset area;

and extracting characteristic data related to the reference product attribute category from the event description data as the target hot event data.

10. A product generating apparatus, the apparatus comprising:

the interaction prediction module is used for predicting interaction prediction data of the target product at the future time corresponding to the different historical moments according to the product interaction data based on a time sequence model;

11. A computer device comprising a memory and a processor, the memory storing a computer program, characterized in that the processor implements the steps of the method of any one of claims 1 to 9 when the computer program is executed.

12. A computer readable storage medium, on which a computer program is stored, characterized in that the computer program, when being executed by a processor, implements the steps of the method of any of claims 1 to 9.

13. A computer program product comprising a computer program, characterized in that the computer program, when being executed by a processor, implements the steps of the method of any one of claims 1 to 9.