CN112508612B

CN112508612B - Method for training advertisement creative generation model and generating advertisement creative and related device

Info

Publication number: CN112508612B
Application number: CN202011442552.XA
Authority: CN
Inventors: 陈云峰; 龚良泉
Original assignee: Beijing Sogou Technology Development Co Ltd
Current assignee: Beijing Sogou Technology Development Co Ltd
Priority date: 2020-12-11
Filing date: 2020-12-11
Publication date: 2024-02-27
Anticipated expiration: 2040-12-11
Also published as: CN112508612A

Abstract

The application discloses a training advertisement creative generation model, a method for generating advertisement creative and a related device, wherein the training advertisement creative generation model method comprises the following steps: taking the query text of the user and the corresponding click text as question-answering training text; identifying a first entity word in the query text, a first commodity category corresponding to the first entity word, and clicking a first commodity feature of a commodity in the text; and training a preset text generation network to obtain an advertising creative generation model through the first entity word, the first commodity category, the first commodity feature and the question-answer training text. The method for generating the advertising creative comprises the following steps: obtaining a second commodity category, a second commodity feature and a second entity word in commodity information of the target commodity; and inputting the second entity word, the second commodity category and the second commodity characteristic into the advertisement creative generation model to generate a second question-answer advertisement creative of the target commodity.

Description

Method for training advertisement creative generation model and generating advertisement creative and related device

Technical Field

The present disclosure relates to the field of data processing technologies, and in particular, to a method and related apparatus for training an ad creative generation model and generating an ad creative.

Background

With the continuous development of internet technology, users are continuously receiving various advertisements in the process of using internet services; the large number of advertisements causes the user to become increasingly numbed of advertisements, resulting in a corresponding decrease in the click-through rate of the advertisements. In order to improve the click rate of the advertisement, the commodity characteristics of the commodity can be combined with a one-question and one-answer form which accords with the daily habits of the user, and the question-answer advertisement creative of the commodity is generated.

In the prior art, question-answer advertising creatives for goods are typically generated based on templates. Specifically, a corresponding question-answer advertisement creative template is constructed aiming at the field of commodity; and then, replacing certain contents in the questioning and answering type advertising creative template aiming at different commodities to generate corresponding questioning and answering type advertising creatives.

However, in the prior art, all the question-answer type advertisement creative templates corresponding to different fields need to be constructed manually, and mobility among the question-answer type advertisement creative templates corresponding to different fields is poor; and different commodities in the same field adopt specific question-answer type advertising creative templates, so that the question-answer type advertising creative of different commodities in the same field is single and poor in novelty.

Disclosure of Invention

In view of this, the present application provides a method and related apparatus for training an ad creative generation model, generating an ad creative, and obtaining an ad creative generation model applicable to a wide variety of goods without manually constructing templates or manually constructing training data; the question-answer type advertising creatives of different commodities generated by using the advertising creative generation model are diversified and have novelty.

In a first aspect, embodiments of the present application provide a method of training an advertising creative generation model, the method comprising:

acquiring a query text and a corresponding click text of a user to generate a question-answer training text;

identifying a first entity word in the query text, a first commodity category corresponding to the first entity word, and a first commodity feature of a commodity in the click text;

and training a preset text generation network to obtain an advertising creative generation model based on the first entity word, the first commodity category, the first commodity feature and the question-answer training text.

Optionally, the training a preset text generation network to obtain an advertisement creative generation model based on the first entity word, the first commodity category, the first commodity feature and the question-answer training text includes:

Inputting the first entity word, the first commodity category and the first commodity characteristic into the preset text generation network to generate a first question-answer type advertising creative;

adjusting parameters of the preset text generation network based on the first question-answer type advertising creative, the question-answer type training text and a preset loss function;

and taking the trained preset text generation network as the advertising creative generation model.

Optionally, the method further comprises:

excavating first core commodity characteristics of commodities in the click text;

screening the first commodity features based on the first core commodity features to obtain screened first commodity features;

correspondingly, the training a preset text generation network to obtain an advertisement creative generation model based on the first entity word, the first commodity category, the first commodity feature and the question-answer training text comprises the following steps:

and training the preset text generation network to obtain the advertising creative generation model based on the first entity word, the first commodity category, the screened first commodity characteristics and the question-answering training text.

Optionally, the preset text generation network is specifically a generation type pre-training language GPT model.

In a second aspect, embodiments of the present application provide a method of generating an advertising creative using the advertising creative generation model of any one of the first aspects, the method comprising:

obtaining a second commodity category, a second commodity feature and a second entity word in commodity information of the target commodity;

inputting the second entity word, the second commodity category and the second commodity feature into the advertising creative generation model;

a second question-and-answer advertising creative is generated for the target item.

Optionally, the method further comprises:

determining a weight of the second merchandise feature based on the interests of the target user or the target user population;

screening the second commodity features based on the weight of the second commodity features to obtain screened second commodity features, wherein the number of the screened second commodity features is smaller than that of the second commodity features;

correspondingly, the inputting the second entity word, the second category of merchandise, and the second feature of merchandise into the advertising creative generation model includes:

and inputting the second entity word, the second commodity category and the screened second commodity characteristics into the advertising creative generation model.

Optionally, the screening the second commodity feature based on the weight of the second commodity feature, to obtain a screened second commodity feature specifically includes:

and selecting the second commodity feature corresponding to the weight larger than or equal to the preset weight as the screened second commodity feature based on the weight of the second commodity feature.

Optionally, the screening the second commodity feature based on the weight of the second commodity feature to obtain a screened second commodity feature includes:

sorting the second commodity features from large to small according to the weights of the second commodity features;

and selecting the first N sorted second commodity features as the screened second commodity features, wherein N is a positive integer.

Optionally, the method further comprises:

judging whether the second question-answer advertising creative meets preset advertising creative indexes or not;

if so, the second question-answer type advertising creative is determined as the target question-answer type advertising creative of the target commodity.

Optionally, the preset advertising creative index includes a text through index, a logical reasonable index and/or a validity index, and the validity index includes a legal regulation index and/or a sensitive information index.

In a third aspect, embodiments of the present application provide an apparatus for training an ad creative generation model, the apparatus comprising:

the acquisition unit is used for acquiring the query text of the user and the corresponding click text to generate a question-answer training text;

the identification unit is used for identifying a first entity word in the query text, a first commodity category corresponding to the first entity word and a first commodity characteristic of the commodity in the click text;

and the training unit is used for training a preset text generation network to obtain an advertising creative generation model based on the first entity word, the first commodity category, the first commodity feature and the question-answering training text.

Optionally, the training unit includes:

the generation subunit is used for inputting the first entity word, the first commodity category and the first commodity characteristic into the preset text generation network to generate a first question-answer advertisement creative;

an adjustment subunit, configured to adjust parameters of the preset text generation network based on the first question-answer advertisement creative, the question-answer training text, and a preset loss function;

and the subunit is used for taking the trained preset text generation network as the advertising creative generation model.

Optionally, the apparatus further includes:

the mining unit is used for mining the first core commodity characteristics of the commodities in the click text;

the first screening unit is used for screening the first commodity characteristics based on the first core commodity characteristics to obtain screened first commodity characteristics;

correspondingly, the training unit is used for:

In a fourth aspect, an embodiment of the present application provides an apparatus for generating an advertising creative using the advertising creative generation model of any one of the first aspects, the apparatus comprising:

the obtaining unit is used for obtaining a second commodity category, a second commodity characteristic and a second entity word in commodity information of the target commodity;

an input unit configured to input the second entity word, the second commodity category, and the second commodity feature into the advertising creative generation model;

and the generation unit is used for generating a second question-answer type advertising creative of the target commodity.

Optionally, the apparatus further includes:

a first determining unit configured to determine a weight of the second commodity feature based on an interest of a target user or a target user group;

a second screening unit, configured to screen the second commodity features based on the weights of the second commodity features, and obtain screened second commodity features, where the number of the screened second commodity features is smaller than the number of the second commodity features;

correspondingly, the input unit is used for:

Optionally, the second screening unit is specifically configured to:

Optionally, the second screening unit includes:

a sorting subunit, configured to sort the second commodity features from big to small according to the weights of the second commodity features;

and the selecting subunit is used for selecting the first N sorted second commodity features as the screened second commodity features, wherein N is a positive integer.

Optionally, the apparatus further includes:

the judging unit is used for judging whether the second question-answer type advertising creative meets preset advertising creative indexes or not;

and the second determining unit is used for determining the second question-answer type advertising creative as the target question-answer type advertising creative of the target commodity if the second question-answer type advertising creative is the target question-answer type advertising creative.

In a fifth aspect, embodiments of the present application provide an apparatus for training an advertising creative generation model, the apparatus comprising a memory, and one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by one or more processors, the one or more programs comprising instructions for:

In a sixth aspect, embodiments of the present application provide an apparatus for generating an advertising creative using the advertising creative generation model of any one of the first aspects, the apparatus comprising a memory, and one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by one or more processors, the one or more programs comprising instructions for:

In a seventh aspect, embodiments of the present application provide a machine readable medium having instructions stored thereon that, when executed by one or more processors, cause an apparatus to perform a method of training an ad creative generation model according to any one of the first aspects above.

In an eighth aspect, embodiments of the present application provide a machine-readable medium having instructions stored thereon that, when executed by one or more processors, cause an apparatus to perform the method of generating an advertising creative of any of the second aspects above.

Compared with the prior art, the application has at least the following advantages:

by adopting the technical scheme of the embodiment of the application, the query text of the user and the corresponding click text are used as question-answering training texts; identifying a first entity word in the query text, a first commodity category corresponding to the first entity word, and clicking a first commodity feature of a commodity in the text; and training a preset text generation network to obtain an advertising creative generation model through the first entity word, the first commodity category, the first commodity feature and the question-answer training text. Therefore, training data is obtained after the query text and the corresponding click text in the search scene are processed, and a preset text generation network is trained to obtain an advertising creative generation model; the method fully mines the association information between the first entity words in the massive query text, the first commodity category corresponding to the first entity words, the first commodity characteristics of the commodities in the click text and the corresponding question-answer training text, so that the advertisement creative generation model is applicable to various commodities.

In addition, a second commodity category, a second commodity feature and a second entity word in commodity information of the target commodity are obtained; and inputting the second entity word, the second commodity category and the second commodity characteristic into the advertisement creative generation model to generate a second question-answer advertisement creative of the target commodity. Based on the characteristics of the advertisement creative generation model, the method is more flexible in generating the question-answer type advertisement creative; therefore, the question-answer type advertising creative of different commodities generated by using the advertising creative generation model is diversified and is rich in novelty.

Drawings

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

Fig. 1 is a schematic diagram of a system frame related to an application scenario in an embodiment of the present application;

FIG. 2 is a flow chart of a method for training an ad creative generation model according to an embodiment of the present application;

FIG. 3 is a schematic diagram of a framework for training an ad creative generation model provided by embodiments of the present application;

FIG. 4 is a flow diagram of a method of generating an advertising creative according to an embodiment of the present application;

FIG. 5 is a schematic structural diagram of an apparatus for training an ad creative generation model according to an embodiment of the present application;

FIG. 6 is a schematic diagram of an apparatus for generating advertising creatives according to an embodiment of the present application;

FIG. 7 is a schematic diagram of an apparatus for training an ad creative generation model or generating an ad creative according to one embodiment of the present application;

fig. 8 is a schematic structural diagram of a server according to an embodiment of the present application.

Detailed Description

In order to make the present application solution better understood by those skilled in the art, the following description will clearly and completely describe the technical solution in the embodiments of the present application with reference to the accompanying drawings in the embodiments of the present application, and it is apparent that the described embodiments are only some embodiments of the present application, not all embodiments. All other embodiments, which can be made by one of ordinary skill in the art without undue burden from the present disclosure, are within the scope of the present disclosure.

In order to improve the click rate of the advertisement, the commodity characteristics of the commodity are combined with a one-question and one-answer form conforming to the daily habits of the user, and the question-answer advertisement creative of the commodity is generated. At present, a question-answer type advertisement creative of the commodity is generally generated based on a template; specifically, a corresponding question-answer advertisement creative template is constructed aiming at the field of commodity; and then, replacing certain contents in the questioning and answering type advertising creative template aiming at different commodities to generate corresponding questioning and answering type advertising creatives. However, in the prior art, all the question-answer type advertisement creative templates corresponding to different fields need to be constructed manually, and mobility among the question-answer type advertisement creative templates corresponding to different fields is poor; and different commodities in the same field adopt specific question-answer type advertising creative templates, so that the question-answer type advertising creative of different commodities in the same field is single and poor in novelty.

In order to solve the problem, in the embodiment of the application, the query text of the user and the corresponding click text are used as question-answering training text; identifying a first entity word in the query text, a first commodity category corresponding to the first entity word, and clicking a first commodity feature of a commodity in the text; and training a preset text generation network to obtain an advertising creative generation model through the first entity word, the first commodity category, the first commodity feature and the question-answer training text. Namely, processing the query text and the corresponding click text in the search scene to obtain training data, and training a preset text generation network to obtain an advertising creative generation model; the method fully mines the association information between the first entity words in the massive query text, the first commodity category corresponding to the first entity words, the first commodity characteristics of the commodities in the click text and the corresponding question-answer training text, so that the advertisement creative generation model is applicable to various commodities.

For example, one of the scenarios of the embodiments of the present application may be applied to the scenario shown in fig. 1, which includes the user terminal 101 and the processor 102. In the case where each user performs a large number of searches through the user terminal 101, that is, in the case where each user performs a search by inputting text through the user terminal 101 and clicks on a certain text or texts displayed; the processor 102 obtains and stores an ad creative generation model using embodiments of the present application that provide for training ad creative generation models. The user sends the target commodity to the processor 102 through the user terminal 101, and the processor 102 obtains the question-answer ad creative of the target commodity by adopting the specific implementation mode for generating the ad creative provided by the embodiment of the application.

First, in the above application scenario, although the operations of the embodiments of the present application are described as being executed by the processor 102, the present application is not limited in terms of execution subject, as long as the operations disclosed in the embodiments of the present application are executed.

Second, the above scenario is only one example of a scenario provided in the embodiments of the present application, and the embodiments of the present application are not limited to this scenario.

Specific implementations of a method for training an ad creative generation model and generating an ad creative and related devices in embodiments of the present application are described in detail below by way of embodiments with reference to the accompanying drawings.

Exemplary method

Referring to FIG. 2, a flow diagram of a method of training an ad creative generation model in an embodiment of the present application is shown. In this embodiment, the method may include, for example, the steps of:

step 201: and acquiring the query text and the corresponding click text of the user to generate a question-answer training text.

Because in the mode of generating the question-answer type advertisement creative of the commodity based on the template, the corresponding question-answer type advertisement creative template needs to be constructed aiming at each field to which the commodity belongs, and the migration among the question-answer type advertisement creative templates corresponding to different fields is poor; and different commodities in the same field adopt specific question-answer type advertising creative templates, so that the question-answer type advertising creative of different commodities in the same field is single and poor in novelty. Therefore, in the embodiment of the application, in order to avoid manually constructing the question-answer type advertising creative template, an advertising creative generation model applicable to various commodities can be trained by a machine learning mode and used for generating question-answer type advertising creatives of various commodities.

In order to train and obtain the advertising creative generation model, training data are needed to be obtained firstly; considering that the purpose of the advertising creative generation model is to generate question-answer advertising creatives for a wide variety of products, question-answer training texts need to be obtained. In order to avoid manual construction of a question-answer training text, based on the searching behavior of each user in daily life, the text input by the user and used for searching can be obtained as a query text, and one or more texts displayed by clicking by the user are taken as clicking texts corresponding to the query text, so that a text in a form of one question and one answer is generated as the question-answer training text.

As one example, why is the text entered by the user for searching "what is the lipstick of color number good? "the text displayed by clicking by the user is" brand new lankou lipstick_152 is hit "; the query text is "what color number lipstick looks good? The click text corresponding to the query text is 'brand new lankou lipstick_152 new color attack'; the generated question-answering training text is a question: what color number lipstick looks good? The method comprises the steps of carrying out a first treatment on the surface of the Answering: all new lankoun lipstick_152 new color comes.

Step 202: and identifying the first entity word in the query text, the first commodity category corresponding to the first entity word, and the first commodity characteristic of the commodity in the click text.

After obtaining the question-answer training text in step 201, other training data except the question-answer training text in the training data needs to be obtained; specifically, a first entity word in a query text, a first commodity category corresponding to the first entity word, and a first commodity feature of a commodity in a click text are inquired; wherein the merchandise characteristics are used to characterize the point of sale of the merchandise, i.e., the nature, characteristics, etc. of the merchandise. Thus, after step 201, query text and click text need to be identified to obtain the other training data; the specific recognition means may be, for example, semantic understanding means, knowledge graph means, etc., and will not be described in detail here.

As one example, on the basis of the above example, the query text "what color number lipstick looks good? The first entity word in the method is an entity word=lipstick, and the first commodity category corresponding to the first entity word lipstick is a commodity category=cosmetic; clicking on the first merchandise feature of the merchandise in the text "all new lankoog lipstick_152 new color attack" is "brand=lankoog", "model=152".

Step 203: and training a preset text generation network to obtain an advertising creative generation model based on the first entity word, the first commodity category, the first commodity feature and the question-answer training text.

In step 201, a question-answer training text is obtained, and in step 202, a first entity word, a first commodity category corresponding to the first entity word is obtained, and after clicking on a first commodity feature of a commodity in the text, it is indicated that training data is obtained. On the basis, a preset text generation network, namely, a preset text generation network can be trained through the training data; and fully mining the first entity word in the query text, the first commodity category corresponding to the first entity word, clicking the first commodity feature of the commodity in the text, and carrying out repeated iterative training on the associated information between the corresponding question-answer training text through massive training data to obtain an advertisement creative generation model, wherein the advertisement creative generation model is applicable to various commodities.

When step 203 is implemented, first, a first entity word, a first commodity category and a first commodity feature need to be input into a preset text generation network, and the preset text generation network can generate and output a text in a question-answer form as a first question-answer advertising creative; then, calculating the loss of the first question-answer type advertisement creative and the question-answer type training text by using a loss function so as to adjust parameters of a preset text generation network; and performing iterative training for a plurality of times until the preset iterative training times are reached or the convergence of the preset text generation network is finished, and taking the trained preset text generation network as an advertising creative generation model. Thus, in an alternative implementation of the embodiments of the present application, step 203 may include, for example, the steps of:

Step A: inputting the first entity word, the first commodity category and the first commodity characteristic into the preset text generation network to generate a first question-answer type advertising creative;

and (B) step (B): adjusting parameters of the preset text generation network based on the first question-answer type advertising creative, the question-answer type training text and a preset loss function;

step C: and taking the trained preset text generation network as the advertising creative generation model.

As an example, a frame diagram of a training advertising creative generation model is shown in fig. 3, on the basis of the above example, a first entity word is "entity word=lipstick", a first commodity category is "commodity category=cosmetic", and a first commodity feature is "brand=lankomon", "model=152" is input into a preset text generation network to generate a first question-answer advertising creative; based on the first question-answer ad creative, question-answer training text "question: what color number lipstick looks good? The method comprises the steps of carrying out a first treatment on the surface of the Answering: the method comprises the steps of (1) adjusting parameters of a preset text generation network according to a new lankoog_152 new color attack and a preset loss function; and taking the trained preset text generation network as an advertising creative generation model.

It should be noted that the preset text generation network may be, for example, a generated Pre-Training language (abbreviated as GPT) model, a BERT (abbreviated as Bidirectional Encoder Representation from Transformers) model, an ELMO (abbreviated as Deep contextualized word representations) model, a Long-short-term memory network (abbreviated as Long-Short Term Memory) or a Pointer network (abbreviated as PN). The GPT model is better in text generation, and the generated text exceeds the expectations of people on the language model of the current stage in terms of context consistency and emotion expression; the method is set to be a preset text generation network, and compared with other text generation models such as BERT models, ELMO models, LSTM models and PN models, the method has better effect. Thus, in an alternative implementation manner of the embodiment of the present application, the preset text generating network is specifically a generating pre-training language GPT model.

In addition, considering that the first merchandise feature of the merchandise in the click text identified in step 202 may have a non-core merchandise feature of the merchandise, the non-core merchandise feature is used to train the preset text generation network to obtain the advertisement creative generation model, which may make the advertisement creative generation model have poor effect. Therefore, in the embodiment of the application, the first core commodity feature of the commodity in the click text can be mined by other sources and the like, and the first core commodity feature is used for screening the first commodity feature so as to filter the non-core commodity feature; based on the method, when the preset text generation network is trained to obtain the advertising creative generation model, the screened first commodity characteristics can be adopted, so that the advertising creative generation model obtained through training is better in effect.

That is, in an alternative implementation manner of the embodiment of the present application, after step 202 and before step 203, for example, the following steps may be further included:

step D: and mining the first core commodity characteristics of the commodities in the click text.

Step E: and screening the first commodity features based on the first core commodity features to obtain screened first commodity features.

Based on the above description, the step 203 may include, for example: and training the preset text generation network to obtain the advertising creative generation model based on the first entity word, the first commodity category, the screened first commodity characteristics and the question-answering training text. The specific implementation refers to the steps a to C, and only the first commodity feature is replaced by the first commodity feature after screening, which is not described herein.

Through the various implementation manners provided by the embodiment, the query text and the corresponding click text of the user are used as question-answering training texts; identifying a first entity word in the query text, a first commodity category corresponding to the first entity word, and clicking a first commodity feature of a commodity in the text; and training a preset text generation network to obtain an advertising creative generation model through the first entity word, the first commodity category, the first commodity feature and the question-answer training text. Therefore, training data is obtained after the query text and the corresponding click text in the search scene are processed, and a preset text generation network is trained to obtain an advertising creative generation model; the method fully mines the association information between the first entity words in the massive query text, the first commodity category corresponding to the first entity words, the first commodity characteristics of the commodities in the click text and the corresponding question-answer training text, so that the advertisement creative generation model is applicable to various commodities.

On the basis of the embodiment, the advertisement creative generation model obtained by training is used for generating question-answer advertisement creatives of various commodities; therefore, by taking any commodity as a target commodity, a second commodity category, a second commodity characteristic and a second entity word in commodity information of the target commodity can be obtained first; and inputting the data into the advertising creative generation model to generate and output a text in a question-answer form as a second question-answer advertising creative of the target commodity. The second question-answer ad creative for the target good is more innovative than the question-answer ad creative for the target good generated based on templates in the prior art.

Referring to FIG. 4, a flow diagram of another method of generating an advertising creative in an embodiment of the present application is shown. In this embodiment, using the ad creative generation model described in the foregoing embodiment, the method may include, for example, the following steps:

step 401: and obtaining a second commodity category, a second commodity characteristic and a second entity word in commodity information of the target commodity.

Step 402: and inputting the second entity word, the second commodity category and the second commodity characteristic into the advertising creative generation model.

Step 403: a second question-and-answer advertising creative is generated for the target item.

It should be noted that, considering that the generated second question-answer type advertising creative of the target commodity needs to be recommended to different users or different user groups, interests of different users are different, and emphasis of the commodity features of interest is also different; similarly, different user groups have different interests and the emphasis on the characteristics of the merchandise they are focusing on. Therefore, in the embodiment of the present application, any one of the users is taken as the target user, or any one of the user groups is taken as the target user group, and different weights may be set for the second commodity feature according to the interests of the target user or the target user group, where the weights of the second commodity feature represent the interest degree of the target user or the target user group for the second commodity feature; and screening part of the second commodity features from the second commodity features through the weights of the second commodity features, wherein the screened second commodity features are more in line with the interests of the target users or target user groups. Based on the method, when the second question-answer type advertisement creative of the target commodity is generated by utilizing the advertisement creative generation model, the screened second commodity characteristics can be adopted, so that the generated second question-answer type advertisement creative of the target commodity is more in line with the interests of target users or target user groups, and the click rate of advertisements is further improved.

That is, in an alternative implementation manner of the embodiment of the present application, after the step 401 and before the step 402, for example, the following steps may be further included:

step F: the weight of the second merchandise feature is determined based on the interests of the target user or target user population.

Step G: screening the second commodity features based on the weight of the second commodity features to obtain screened second commodity features, wherein the number of the screened second commodity features is smaller than that of the second commodity features;

in this embodiment, step G may be performed by at least two embodiments:

in a first alternative embodiment of step G, a greater weight of the second merchandise feature indicates that the target user or group of target users is interested in the second merchandise feature, and a lesser weight of the second merchandise feature indicates that the target user or group of target users is less interested in the second merchandise feature; in order to distinguish whether the target user or the target user group is interested in the second commodity feature through the weight, a weight can be preset as a preset weight, the lower limit of the interest of the target user or the target user group is represented, and then the second commodity feature with the weight smaller than the preset weight is required to be filtered out from the second commodity feature, and the second commodity feature with the weight larger than or equal to the preset weight is filtered and used as the filtered second commodity feature. Thus, in an alternative implementation manner of the embodiment of the present application, the step G may be, for example, specifically: and selecting the second commodity feature corresponding to the weight larger than or equal to the preset weight as the screened second commodity feature based on the weight of the second commodity feature.

In a second alternative embodiment of step G, a greater weight of the second merchandise feature indicates that the target user or target user population is more interested in the second merchandise feature, and a lesser weight of the second merchandise feature indicates that the target user or target user population is less interested in the second merchandise feature; in order to display the ordering of the interest of the target user or the target user group to the second commodity features, the second commodity features are ordered from large to small according to the weight of the second commodity features, and the first N ordered second commodity features are selected after the ordering to serve as the screened second commodity features. Thus, in an alternative implementation of the embodiment of the present application, the step G may include, for example, the following steps:

step G1: sorting the second commodity features from large to small according to the weights of the second commodity features;

step G2: and selecting the first N sorted second commodity features as the screened second commodity features, wherein N is a positive integer.

As can be seen from the above description, the step 402 may include, for example: and inputting the second entity word, the second commodity category and the screened second commodity characteristics into the advertising creative generation model.

It should be further noted that, after the second question-answer type ad creative of the target product is generated in step 402, one or more indexes related to the ad creative may be preset as preset ad creative indexes, considering whether the generated second question-answer type ad creative can be actually applied to the ad; it is necessary to determine whether the second question-answer ad creative meets the preset ad creative index, and only if so, the second question-answer ad creative can be used as the target question-answer ad creative of the target commodity. Thus, in an alternative implementation of the embodiment of the present application, after said step 402, for example, the following steps may be further included:

step H: judging whether the second question-answer advertising creative meets preset advertising creative indexes or not; if yes, executing the step I.

The second question-answer type advertisement creative needs to have characteristics of text smoothness, reasonable logic, legitimacy and the like in consideration of the fact that the generated second question-answer type advertisement creative is actually applied to advertisements, and the legitimacy shows that the second question-answer type advertisement creative meets requirements of laws and regulations, sensitive information and the like. Thus, in an alternative implementation of the embodiments of the present application, the preset advertising creative indicators include a text-passing indicator, a logical reasonable indicator, and/or a validity indicator, and the validity indicator includes a legal regulation indicator and/or a sensitive information indicator.

Step I: the second question-answer advertising creative is determined as a target question-answer advertising creative for the target item.

Through the various implementation manners provided by the embodiment, a second commodity category, a second commodity feature and a second entity word in commodity information of the target commodity are obtained; and inputting the second entity word, the second commodity category and the second commodity characteristic into the advertisement creative generation model to generate a second question-answer advertisement creative of the target commodity. Based on the characteristics of the advertisement creative generation model, the method is more flexible in generating the question-answer type advertisement creative; therefore, the question-answer type advertising creative of different commodities generated by using the advertising creative generation model is diversified and is rich in novelty.

Exemplary apparatus

Referring to FIG. 5, a schematic diagram of an apparatus for training an ad creative generation model in an embodiment of the present application is shown. In this embodiment, the apparatus may specifically include, for example:

an obtaining unit 501, configured to obtain a query text of a user and a corresponding click text to generate a question-answer training text;

the identifying unit 502 is configured to identify a first entity word in the query text, a first commodity category corresponding to the first entity word, and a first commodity feature of a commodity in the click text;

And a training unit 503, configured to train a preset text generation network to obtain an advertisement creative generation model based on the first entity word, the first commodity category, the first commodity feature and the question-answer training text.

In an alternative implementation manner of the embodiment of the present application, the training unit 503 includes:

In an alternative implementation manner of the embodiment of the present application, the apparatus further includes:

Correspondingly, the training unit 503 is configured to:

In an optional implementation manner of the embodiment of the present application, the preset text generation network is specifically a generated pre-training language GPT model.

Referring to FIG. 6, a schematic diagram of an apparatus for generating advertising creatives in an embodiment of the present application is shown. In this embodiment, with the advertising creative generation model described in the foregoing embodiment, the apparatus may specifically include:

an obtaining unit 601, configured to obtain a second entity word in a second commodity category, a second commodity feature, and commodity information of the target commodity;

an input unit 602 configured to input the second entity word, the second item category, and the second item feature into the advertising creative generation model;

a generating unit 603, configured to generate a second question-answer type advertising creative of the target commodity.

correspondingly, the input unit 602 is configured to:

In an optional implementation manner of the embodiment of the present application, the second screening unit is specifically configured to:

In an optional implementation manner of the embodiment of the present application, the second screening unit includes:

In an optional implementation manner of the embodiment of the present application, the preset advertising creative index includes a text through index, a logical reasonable index and/or a validity index, and the validity index includes a legal regulation index and/or a sensitive information index.

Figure 7 is a block diagram illustrating an apparatus 700 for training an ad creative generation model or generating an ad creative in accordance with an exemplary embodiment. For example, apparatus 700 may be a mobile phone, computer, digital broadcast terminal, messaging device, game console, tablet device, medical device, exercise device, personal digital assistant, or the like.

Referring to fig. 7, an apparatus 700 may include one or more of the following components: a processing component 702, a memory 704, a power component 706, a multimedia component 708, an audio component 710, an input/output (I/O) interface 712, a sensor component 714, and a communication component 716.

The processing component 702 generally controls overall operation of the apparatus 700, such as operations associated with display, telephone calls, data communications, camera operations, and recording operations. The processing component 702 may include one or more processors 720 to execute instructions to perform all or part of the steps of the methods described above. Further, the processing component 702 can include one or more modules that facilitate interaction between the processing component 702 and other components. For example, the processing component 702 may include a multimedia module to facilitate interaction between the multimedia component 708 and the processing component 702.

Memory 704 is configured to store various types of data to support operations at device 700. Examples of such data include instructions for any application or method operating on the apparatus 700, contact data, phonebook data, messages, pictures, videos, and the like. The memory 704 may be implemented by any type or combination of volatile or nonvolatile memory devices such as Static Random Access Memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic or optical disk.

The power supply component 706 provides power to the various components of the device 700. The power components 706 may include a power management system, one or more power sources, and other components associated with generating, managing, and distributing power for the device 700.

The multimedia component 708 includes a screen between the device 700 and the user that provides an output interface. In some embodiments, the screen may include a Liquid Crystal Display (LCD) and a Touch Panel (TP). If the screen includes a touch panel, the screen may be implemented as a touch screen to receive input signals from a user. The touch panel includes one or more touch sensors to sense touches, swipes, and gestures on the touch panel. The touch sensor may sense not only the boundary of a touch or sliding action, but also the duration and pressure associated with the touch or sliding operation. In some embodiments, the multimedia component 708 includes a front-facing camera and/or a rear-facing camera. The front-facing camera and/or the rear-facing camera may receive external multimedia data when the device 700 is in an operational mode, such as a shooting mode or a video mode. Each front camera and rear camera may be a fixed optical lens system or have focal length and optical zoom capabilities.

The audio component 710 is configured to output and/or input audio signals. For example, the audio component 710 includes a Microphone (MIC) configured to receive external audio signals when the device 700 is in an operational mode, such as a call mode, a recording mode, and a voice recognition mode. The received audio signals may be further stored in the memory 704 or transmitted via the communication component 716. In some embodiments, the audio component 710 further includes a speaker for outputting audio signals.

The I/O interface 712 provides an interface between the processing component 702 and peripheral interface modules, which may be a keyboard, click wheel, buttons, etc. These buttons may include, but are not limited to: homepage button, volume button, start button, and lock button.

The sensor assembly 714 includes one or more sensors for providing status assessment of aspects of the apparatus 700. For example, the sensor assembly 714 may detect an on/off state of the device 700, a relative positioning of the components, such as a display and keypad of the apparatus 700, a change in position of the apparatus 700 or one component of the apparatus 700, the presence or absence of user contact with the apparatus 700, an orientation or acceleration/deceleration of the apparatus 700, and a change in temperature of the apparatus 700. The sensor assembly 714 may include a proximity sensor configured to detect the presence of nearby objects without any physical contact. The sensor assembly 714 may also include a light sensor, such as a CMOS or CCD image sensor, for use in imaging applications. In some embodiments, the sensor assembly 714 may also include an acceleration sensor, a gyroscopic sensor, a magnetic sensor, a pressure sensor, or a temperature sensor.

The communication component 716 is configured to facilitate communication between the apparatus 700 and other devices in a wired or wireless manner. The apparatus 700 may access a wireless network based on a communication standard, such as WiFi,2G or 3G, or a combination thereof. In one exemplary embodiment, the communication part 716 receives a broadcast signal or broadcast related information from an external broadcast management system via a broadcast channel. In an exemplary embodiment, the communication component 716 further includes a Near Field Communication (NFC) module to facilitate short range communications. For example, the NFC module may be implemented based on Radio Frequency Identification (RFID) technology, infrared data association (IrDA) technology, ultra Wideband (UWB) technology, bluetooth (BT) technology, and other technologies.

In an exemplary embodiment, the apparatus 700 may be implemented by one or more Application Specific Integrated Circuits (ASICs), digital Signal Processors (DSPs), digital Signal Processing Devices (DSPDs), programmable Logic Devices (PLDs), field Programmable Gate Arrays (FPGAs), controllers, microcontrollers, microprocessors, or other electronic components for executing the methods described above.

In an exemplary embodiment, a non-transitory computer readable storage medium is also provided, such as memory 704, including instructions executable by processor 720 of apparatus 700 to perform the above-described method. For example, the non-transitory computer readable storage medium may be ROM, random Access Memory (RAM), CD-ROM, magnetic tape, floppy disk, optical data storage device, etc.

A non-transitory computer readable storage medium, which when executed by a processor of a mobile terminal, causes the mobile terminal to perform a method of training an ad creative generation model, the method comprising:

A non-transitory computer readable storage medium, which when executed by a processor of a mobile terminal, causes the mobile terminal to perform a method of generating an advertising creative, the method comprising:

Fig. 8 is a schematic structural diagram of a server in an embodiment of the present application. The server 800 may vary considerably in configuration or performance and may include one or more central processing units (central processing units, CPUs) 822 (e.g., one or more processors) and memory 832, one or more storage media 830 (e.g., one or more mass storage devices) storing applications 842 or data 844. Wherein the memory 832 and the storage medium 830 may be transitory or persistent. The program stored in the storage medium 830 may include one or more modules (not shown), each of which may include a series of instruction operations on a server. Still further, the central processor 822 may be configured to communicate with the storage medium 830 to execute a series of instruction operations in the storage medium 830 on the server 800.

The server 800 may also include one or more power supplies 826, one or more wired or wireless network interfaces 850, one or more input/output interfaces 858, one or more keyboards 856, and/or one or more operating systems 841, such as Windows ServerTM, mac OS XTM, unixTM, linuxTM, freeBSDTM, and the like.

In this specification, the embodiments are described in a progressive manner, and each embodiment focuses on the difference from other embodiments, and the same similar parts between the embodiments are referred to each other. For the device disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant points refer to the description of the method section.

Those of skill would further appreciate that the various illustrative elements and algorithm steps described in connection with the embodiments disclosed herein may be implemented as electronic hardware, computer software, or combinations of both, and that the various illustrative elements and steps are described above generally in terms of functionality in order to clearly illustrate the interchangeability of hardware and software. Whether such functionality is implemented as hardware or software depends upon the particular application and design constraints imposed on the solution. Skilled artisans may implement the described functionality in varying ways for each particular application, but such implementation decisions should not be interpreted as causing a departure from the scope of the present application.

It is noted that relational terms such as first and second, and the like are used solely to distinguish one entity or action from another entity or action without necessarily requiring or implying any actual such relationship or order between such entities or actions. The terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one … …" does not exclude the presence of other like elements in a process, method, article, or apparatus that comprises the element.

The foregoing description is only of the preferred embodiments of the present application and is not intended to limit the present application in any way. While the present application has been described with reference to the preferred embodiments, it is not intended to limit the present application. Any person skilled in the art may make many possible variations and modifications to the technical solution of the present application, or modify equivalent embodiments, using the methods and technical contents disclosed above, without departing from the scope of the technical solution of the present application. Therefore, any simple modification, equivalent variation and modification of the above embodiments according to the technical substance of the present application, which do not depart from the content of the technical solution of the present application, still fall within the scope of protection of the technical solution of the present application.

Claims

1. A method of training an advertising creative generation model, comprising:

training a preset text generation network to obtain an advertising creative generation model based on the first entity word, the first commodity category, the first commodity feature and the question-answering training text;

The training a preset text generation network to obtain an advertising creative generation model based on the first entity word, the first commodity category, the first commodity feature and the question-answer training text comprises the following steps:

2. The method as recited in claim 1, further comprising:

3. The method according to any of claims 1 to 2, wherein the pre-set text generation network is in particular a generated pre-training language, GPT, model.

4. A method of generating an advertising creative using the advertising creative generation model of any one of claims 1 to 3, the method comprising:

5. The method as recited in claim 4, further comprising:

6. The method according to claim 5, wherein the screening the second commodity feature based on the weight of the second commodity feature, and obtaining the screened second commodity feature specifically includes:

7. The method of claim 5, wherein the screening the second merchandise feature based on the weight of the second merchandise feature to obtain a screened second merchandise feature comprises:

8. The method as recited in claim 4, further comprising:

9. The method of claim 8, wherein the pre-set ad creative indicators comprise text-passing indicators, logical reasonable indicators, and/or legal indicators, the legal indicators comprising legal regulation indicators and/or sensitive information indicators.

10. An apparatus for training an advertising creative generation model, comprising:

the training unit is used for training a preset text generation network to obtain an advertising creative generation model based on the first entity word, the first commodity category, the first commodity feature and the question-answering training text;

the training unit includes:

11. The apparatus of claim 10, wherein the apparatus further comprises:

correspondingly, the training unit is used for:

12. The apparatus according to any one of claims 10 to 11, wherein the pre-set text generation network is embodied as a generated pre-training language, GPT, model.

13. An apparatus for generating an advertising creative using the advertising creative generation model of any one of claims 1 to 3, the apparatus comprising:

14. The apparatus of claim 13, wherein the apparatus further comprises:

correspondingly, the input unit is used for:

15. The apparatus according to claim 14, wherein the second screening unit is specifically configured to:

16. The apparatus of claim 14, wherein the second screening unit comprises:

17. The apparatus of claim 13, wherein the apparatus further comprises:

18. The apparatus of claim 17, wherein the pre-set ad creative indicators comprise text-passing indicators, logical reasonable indicators, and/or legitimacy indicators, the legitimacy indicators comprising legal regulation indicators and/or sensitive information indicators.

19. An apparatus for training an ad creative generation model, comprising a memory, and one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by one or more processors, the one or more programs comprising instructions for:

20. The device of claim 19, wherein the device is further configured to be executed by one or more processors the one or more programs include instructions for:

21. The apparatus according to any one of claims 19 to 20, wherein the pre-set text generation network is embodied as a generated pre-training language, GPT, model.

22. An apparatus for generating an advertising creative using the advertising creative generation model of any one of claims 1-3, the apparatus comprising a memory and one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the one or more processors, the one or more programs comprising instructions for:

23. The device of claim 22, wherein the device is further configured to be executed by one or more processors the one or more programs include instructions for:

24. The apparatus of claim 23, wherein the screening the second merchandise feature based on the weight of the second merchandise feature obtains a screened second merchandise feature, specifically:

25. The apparatus of claim 23, wherein the screening the second merchandise feature based on the weight of the second merchandise feature to obtain a screened second merchandise feature comprises:

26. The device of claim 22, wherein the device is further configured to be executed by one or more processors the one or more programs include instructions for:

27. The apparatus of claim 26, wherein the pre-set ad creative indicators comprise text-passing indicators, logical reasonable indicators, and/or legitimacy indicators, the legitimacy indicators comprising legal regulations indicators and/or sensitive information indicators.

28. A machine readable medium having instructions stored thereon that, when executed by one or more processors, cause an apparatus to perform the method of training an advertising creative generation model of any of claims 1-3.

29. A machine-readable medium having instructions stored thereon that, when executed by one or more processors, cause an apparatus to perform the method of generating an advertising creative of any of claims 4-9.