Method and apparatus for generating copy in advertisement style, and electronic device and medium

Abstract

Provided in the embodiments of the present disclosure are a method and apparatus for generating copy in an advertisement style, and an electronic device and a medium. The method comprises: acquiring source data associated with a target advertisement; using a trained generative model to generate, from the source data, candidate copy of a target element in an advertisement style; performing post-processing on the candidate copy, so as to acquire available copy for the target element; and delivering the target advertisement to a user equipment, wherein the target advertisement is displayed on the user equipment on the basis of the advertisement style, and at least one piece of copy among the available copy is displayed on the target element in the advertisement style.

Description

Method, device, electronic device and medium for generating copy of advertisement style TECHNICAL FIELD

[0001] Embodiments of the present disclosure relate to the field of computer technology, and more particularly, to a method, device, electronic device, medium and computer program product for generating copy of advertisement style. BACKGROUND

[0002] It has been widely recognized that the effective information contained in the advertisement is more capable of effectively attracting the attention of users and helping users make decisions than the layout of the advertisement. The existing advertisement style has insufficient information, single copy content and lack of personalization, which leads to low click-through rate and conversion rate of the advertisement.

[0003] Artificial intelligence-generated content (AIGC) refers to any content created or generated by an artificial intelligence system under the guidance of humans or in the case of participation. With the advancement of natural language processing (NLP), machine learning and deep learning algorithms, artificial intelligence can now generate various types of content, including text, images, videos, music, etc. How to use AIGC technology to generate high-quality advertisement copy poses a challenge to practitioners.

[0004] SUMMARY

[0005] In view of this, embodiments of the present disclosure provide a technical solution for generating copy of advertisement style using a generative model. By inputting information to be put into an advertisement as source data into a generative model, more abundant and high-quality copy is generated and applied to various advertisement styles, thereby improving the information quantity and effectiveness of the advertisement style and reducing the workload of the advertiser.

[0006] According to a first aspect of the present disclosure, a method for generating copy of advertisement style is provided. The method comprises: obtaining source data associated with a target advertisement; generating candidate copy of a target element in an advertisement style from the source data by using a trained generative model; post-processing the candidate copy to obtain available copy for the target element; and putting the target advertisement to a user device, wherein the target advertisement is displayed on the user device according to the advertisement style, and at least one of the available copy is displayed on the target element of the advertisement style.

[0007] According to a second aspect of the present disclosure, there is provided an apparatus for generating copy in an advertisement style. The apparatus comprises: a source data obtaining unit configured to obtain source data associated with a target advertisement; a copy generating unit configured to generate, by using a trained generative model, candidate copy for a target element in an advertisement style from the source data; a post-processing unit configured to post-process the candidate copy to obtain available copy for the target element; and an advertisement serving unit configured to serve the target advertisement to a user device, wherein the target advertisement is displayed on the user device according to the advertisement style, and at least one of the available copy is displayed on the target element of the advertisement style.

[0008] According to a third aspect of the present disclosure, there is provided an electronic device, comprising: one or more processors; and memory storing one or more programs for execution by the one or more processors, which, when executed by the one or more processors, cause the one or more processors to implement the method according to the first aspect of the present disclosure.

[0009] According to a fourth aspect of the present disclosure, there is provided a computer-readable storage medium having stored thereon a computer program, which, when executed by a processor, implements the method according to the first aspect of the present disclosure.

[0010] According to a fifth aspect of the present disclosure, there is provided a computer program product comprising a computer program which, when executed by a processor, implements the method according to the first aspect of the present disclosure.

[0011] The summary is provided to introduce a selection of concepts, in a simplified form, that are further described below in the DETAILED DESCRIPTION. This summary is not intended to identify key or essential features of the claimed subject matter, nor is it intended to limit the scope of the claimed subject matter. BRIEF DESCRIPTION OF DRAWINGS

[0012] The above and other objects, features and advantages of the present disclosure will become more apparent from the following detailed description when taken in conjunction with the accompanying drawings in which like reference characters refer to like parts throughout the figures, and in which:

[0013] FIG. 1 shows a schematic diagram of an exemplary environment in which embodiments of the present disclosure can be implemented;

[0014] FIG. 2 shows a schematic diagram of an exemplary advertisement style presented on a user device;

[0015] FIG. 3 shows a schematic flowchart of a method for generating copy in an advertisement style according to an embodiment of the present disclosure;

[0016] FIG. 4 shows a schematic flowchart of a method of creating a generative model according to an embodiment of the present disclosure;

[0017] FIG. 5 shows a schematic flowchart of a method of post-processing a candidate script according to an embodiment of the present disclosure;

[0018] FIG. 6 shows a schematic block diagram of an apparatus for content creation according to an embodiment of the present disclosure; and

[0019] FIG. 7 shows a block diagram of a device capable of implementing a number of embodiments of the present disclosure. DETAILED DESCRIPTION

[0020] It can be understood that, before using the technical solutions disclosed in the embodiments of the present disclosure, the type of personal information involved in the present disclosure, the scope of use, the scenario of use, etc. should be informed to the user and the authorization of the user should be obtained according to relevant laws and regulations.

[0021] The present disclosure will now be discussed with reference to a number of example implementations. It should be appreciated that these implementations are discussed solely for the purposes of exemplifying the present disclosure and to enable those of ordinary skill in the art to better understand and thus implement the present disclosure, and are not intended to limit the scope of the present disclosure in any way.

[0022] As used herein, the term “includes” and its variants are to be read as open-ended terms that mean “includes, but is not limited to.” The term “based on” is to be construed as “based at least in part on.” The terms “one implementation” and “an implementation” are to be construed as “at least one implementation.” The term “another implementation” is to be construed as “at least one other implementation.” The terms “first,” “second,” etc. can refer to different or the same objects. Other explicitly and implicitly recited definitions come with the context of the following claims. It is also noted that any numeral or number used in the present disclosure is exemplary only, and is by no means intended to limit the scope of the present disclosure.

[0023] An advertisement format refers to the presentation form or layout of an advertisement, aiming to attract the attention of the target audience and effectively convey the information of the advertiser. The elements of an advertisement format include the advertisement creative (e.g., image or video) and its script, such as a call-to-action (CTA), a selling point information, an advertisement description, etc. The CTA is the core script carrying the conversion of the advertisement, and its position is on the CTA button of the advertisement format. The selling point information can be considered as incremental information, which is usually a refined short route, playing a role of helping the audience to highlight. The position of the selling point information can be diversified, and multiple selling points can be included in the advertisement format. The advertisement description can also be considered as incremental information, providing more abundant and complete information than the selling point information.

[0024] It has been recognized that the effective information provided by the copy of the advertisement style can more effectively attract the attention of the user and help the user make a decision than the layout of the advertisement. However, the existing advertisement copy is difficult to meet the needs of the advertiser. First, most of the advertisement styles have only one CTA button, and the effective information is insufficient and lacks personalization. For example, the common CTA button copy is "click to buy", "learn more", "view", etc., which is difficult to help the user make a decision. Moreover, the advertisement style and its copy have the problem of effectiveness decay, and the magnitude of the available copy needs to be improved to fully exert the personalized optimization capability. In addition, the existing advertisement production cost is high, and many advertisers do not have a professional copywriting team, resulting in low advertisement conversion rate.

[0025] Therefore, embodiments of the present disclosure provide a technical solution of generating the copy of the advertisement style by using a generative model. In an example method, the information of the advertisement to be launched (for example, the landing page, the advertisement video, the advertisement title, the data of the advertisement object) can be input as source data into the generative model to generate more rich and high-quality copy and apply it to various types of advertisement styles. The corresponding copy can be generated for different elements of the advertisement style, for example, the CTA copy, the selling point copy, or the description information of the advertisement. In some embodiments, in order to ensure the compliance and quality of the copy, the copy generated by the model can be post-processed to exclude non-compliant and low-quality copies. In this way, the information quantity and effectiveness of the advertisement style are improved, and the workload of the advertiser is reduced. The implementation details of the embodiments of the present disclosure are described in detail below with reference to FIGS. 1-7.

[0026] FIG. 1 shows a schematic diagram of an example environment 100 capable of implementing embodiments of the present disclosure. As shown in FIG. 1, the environment 100 includes a source data repository 110, a computing device 120 for generating the copy, and a user device 150 receiving the advertisement launch. The source data repository 110 stores the source data obtained from various data sources and processed, including the landing page of the advertisement (for example, obtained by a crawler), the speech recognition result or the text recognition result of the advertisement material (for example, a video or an image), the title of the advertisement material, and the description information of the advertisement object. In some embodiments, the data in the source data repository 110 can be classified according to the type of the source of the advertisement material (also referred to as "vertical category"). The raw data from various data sources can be pre-processed, processed, useful information extracted, and audited and filtered, etc. The processed data is stored in the source data repository 110 as source data.

[0027] The computing device 120 can be a standalone physical network server, a network server cluster or distributed system composed of multiple physical network servers, a cloud network server providing cloud services, cloud database, cloud computing, cloud function, cloud storage, network service, cloud communication, middleware service, domain name service, security service, CDN, and basic cloud computing services such as big data and artificial intelligence platform, etc.

[0028] The computing device 120 can include or run a generative model 130. The generative model 130 can be, for example, a large language model (LLM), and is trained to receive source data associated with a target advertisement to be served as input, and output an advertisement copy. The source data input to the generative model can include a landing page of the target advertisement, a speech recognition result or a text recognition result of an advertisement creative content (e.g., a video or an image), a title of a broadcast creative content, description information of an advertisement object. The generative model 130 can produce respective copies for different elements in an advertisement style, including a CTA copy, a selling point copy, an advertisement title, description information of an advertisement object, etc. In some embodiments, the generative model 130 can be trained based on source data in a source data repository 130,

[0029] The computing device 120 includes or runs a post-processing module 140 for processing the copy output by the generative model 130, from which to filter out copies that can be used in an advertisement style. In this context, the copy output by the generative model 130 can be referred to as a candidate copy. In some embodiments, the post-processing module 140 can exclude a portion of the candidate copies based on compliance requirements and quality requirements. The filtered copies can be stored in a library (not shown), and can be served to the user device 150 as content of the target advertisement. In some embodiments, the computing device 120 can provide the generated copies to the advertiser for selection. Alternatively, the computing device 120 can not provide the generated copies to the advertiser to reduce the burden on the advertiser. This setup can depend on the elements of the advertisement style involved in the advertisement copy.

[0030] The user device 150 can be any electronic device with display capability, including but not limited to: a smartphone, a tablet computer, a portable computer, a smart television, a vehicle-mounted computer, a wearable device (e.g., a smart bracelet, a smart watch), etc. When a target advertisement conforming to an advertisement style is displayed on the user device 150, in addition to displaying an advertisement creative, a copy customized for the target advertisement is presented on an element of the advertisement style.

[0031] It should be appreciated that the environment 100 illustrated in FIG. 1 is merely exemplary and should not be construed as limiting the functionality and scope of the implementations described herein. The environment 100 can also include more or fewer components than those shown. For ease of understanding, exemplary ad styles are introduced as follows.

[0032] FIG. 2 illustrates a schematic diagram of an exemplary ad style presented on a user device. In FIG. 2, the user device is a smartphone, and it should be appreciated that similar ad styles are applicable to other types of user devices as well.

[0033] The user interface 200 includes an ad style 210 being displayed, which includes the following elements: a region 220 displaying a video or image, advertiser information 212, ad description 214, ad selling points 216, and a CTA button 218. For ease of understanding, assume that the ad being displayed is for a car, the advertiser information 212 can include the sales store information, the ad description 214 can include the car model or promotional phrases, the selling points 216 can be multiple, such as “luxury”, “sporty”, “five-year interest-free”, and the CTA button can be, for example, “Learn More”. As an example, the text of “Learn More” can lack personalization or provide insufficient effective information. Embodiments of the present disclosure can utilize the generative model to provide more abundant CTA text, such as “Schedule a Test Drive”, “Estimate Monthly Payment”, etc., thereby more effectively guiding the user to click on the ad style 210 or any element thereof.

[0034] FIG. 3 illustrates a schematic flowchart of a method 300 for generating text of an ad style, according to embodiments of the present disclosure. The method 300 can be implemented by, for example, the computing device 120 shown in FIG. 1. It should be appreciated that the method 200 can also include additional actions not shown and / or can omit actions shown, and the scope of the present disclosure is not limited in this regard. For ease of explanation, the method 300 is described in conjunction with FIGS. 1 and 2.

[0035] As shown in FIG. 3, at block 310, the computing device 120 obtains source data associated with a target ad. In some embodiments, the computing device 120 can obtain raw data from a source data repository 120 and pre-process the raw data to obtain the source data. The source data can be in a text form for ease of input to the generative model 130. The pre-processing can include production processing, extracting data usable by the generative model 130, review filtering operations, etc. For example, the computing device 120 can obtain a landing page of the target ad by a crawler and extract text information of the landing page (including text recognition results of images therein), perform automatic speech recognition (ASR) on a creative video of the target ad to obtain text of the speech.

[0036] At block 320, the computing device 120 generates, using the trained generative model 130, candidate scripts for the target elements in the ad style from the source data. In some embodiments, the generative model 130 can generate the candidate scripts from the prompt words that include the source data and the target element information. For example, the target elements can be any one or more of the CTA button 218, the ad selling point 216, or the ad description 214.

[0037] For each target element, the generative model 130 can output a set of candidate scripts that are associated with the characteristics of the ad object of the target ad, thus providing personalized customized scripts. For example, for a car ad, the scripts output by the generative model 130 can be car-related phrases such as “schedule a test drive”; for a clothing ad, the scripts output by the generative model 130 can be phrases associated with the attributes of the clothing such as “try on the outfit”; for a game ad, the scripts output by the generative model 130 can be phrases related to installing the game app such as “download and try it out”. It can be appreciated that the quality or richness of the output candidate scripts can vary depending on the training level of the generative model 130.

[0038] At block 330, the computing device 120 post-processes the candidate scripts to obtain available scripts for the target elements. To meet the compliance requirements and improve the quality of the scripts, the post-processing module 140 of the computing device 120 can select a subset of the candidate scripts as the available scripts. In some embodiments, the post-processing module 140 can include one or more sub-modules, each of which can remove non-compliant or low-quality scripts from the candidate scripts. Details of the post-processing module 140 will be described in connection with FIG. 5 below.

[0039] The post-processed scripts can be saved in a library for subsequent delivery to the user device 150. In some embodiments, the ad owner can be asked which scripts can be saved and used, in other words, the ad owner can select the script information of the ad to be delivered. For example, when the target element is a CTA button, the ad owner can be shown the scripts from the generative model 130, from which the ad owner can select a portion to allow use; or select a portion not to allow use. The scripts allowed to use will be saved to the library. The library can also include other content such as the pictures, videos, selling points, ad owner information, etc. of the ad.

[0040] At block 340, the computing device 120 serves the targeted advertisement to the user device, such that the targeted advertisement is displayed on the user device according to the advertisement style, and the at least one available copy is displayed on the target element of the advertisement style. In some embodiments, for the targeted advertisement, the computing device 120 can determine the advertisement style to be used (e.g., according to the type of the user device), and fetch one or more copies corresponding to the specific target element in the advertisement style from the material library, package the advertisement style information in combination with other materials of the advertisement style, and serve to the user device 150.

[0041] As mentioned above, the generative model 130 can be trained offline to have the ability to generate copies for target elements from source data. An exemplary training method is introduced below.

[0042] FIG. 4 shows a schematic flowchart of a method 400 of creating a generative model according to an embodiment of the present disclosure. Generally, in the method 400, a pre-trained generative model (also referred to as a “base model”) is fine-tuned to obtain a generative model having the ability to generate copies.

[0043] At block 410, the computing device 120 obtains training data. The training data can be from the source data repository 110 of FIG. 1, including landing pages of advertisements, speech recognition results or text recognition results of advertisement materials, titles of advertisement materials, description information of advertisement objects, etc.

[0044] At block 420, the computing device 120 generates a fine-tuning data set based on the obtained training data. In some embodiments, the fine-tuning of the base model can be supervised training. For this purpose, the training data can be labeled using a large language model different from the base model. In some embodiments, the fine-tuning data set can be generated from the training data based on a prompt word for the large language model, where the prompt word can be customized to instruct the large language model to generate copies for elements of the advertisement style. The copies generated by the large language model can be paired with the corresponding training data to form the fine-tuning data set. Based on such a manner, the prompt word can be designed for different scenarios, and then the high-quality training data set is labeled by the large language model and the base model is fine-tuned, finally covering different target elements, such as different downstream tasks of selling points, CTAs, advertisement descriptions, etc.

[0045] At block 430, the base model is trained offline using the fine-tuning data set. In this way, the base model can be trained as a generative model 130 having the ability to generate target copies for the advertisement style from the source data.

[0046] Considering that the inference process of the generative model 130 is computationally expensive, at block 440, quantization acceleration and batch inference can be introduced for the fine-tuned generative model 130 to improve the query per second (QPS) performance of the model. Based on such a manner, the amount of copywriting output of the computing device 120 is significantly improved, thereby providing more high-quality copywriting.

[0047] FIG. 5 shows a schematic flowchart of a method 500 of post-processing candidate copywriting according to an embodiment of the present disclosure. Generally, the method 500 is used to obtain available copywriting 504 from candidate copywriting 502 output by the screening generative model 120. The method 500 can be an exemplary implementation of block 330 in FIG. 3. It should be understood that the method 500 can include more steps, or some steps can be omitted, and the steps shown in FIG. 5 can be performed in different orders.

[0048] At block 510, blacklist filtering is performed on the candidate copywriting 502. In some embodiments, the post-processing module 130 can access a pre-configured word blacklist, and if a word in the word blacklist exists in a copywriting, the copywriting is excluded from the candidate copywriting.

[0049] At block 520, quality evaluation is performed on the candidate copywriting filtered by the blacklist. In some embodiments, the post-processing module can include a quality evaluation model for scoring each of the candidate copywritings. If the score of a copywriting is lower than a certain quality threshold, the corresponding low-quality copywriting can be excluded from the candidate copywriting. The quality evaluation model can be trained based on human annotation.

[0050] At block 530, further risk detection is performed on the candidate copywriting. In some embodiments, the post-processing module 140 can have one or more risk detection models, each of which detects whether a copywriting has compliance risk from a respective risk factor.

[0051] The candidate copywritings that pass the blacklist filtering 510, the quality evaluation 520, and the risk evaluation 530 form the available copywriting 504.

[0052] FIG. 6 shows a schematic block diagram of an apparatus 600 for generating copywriting of an advertisement style according to an embodiment of the present disclosure. The apparatus 600 can be implemented by the computing device 100 shown in FIG. 1. As shown in FIG. 6, the apparatus 600 includes a source data acquisition unit 610, a copywriting generation unit 620, a post-processing unit 630, and an advertisement delivery unit 640.

[0053] The source data acquisition unit 610 is configured to acquire source data associated with a target advertisement. The script generation unit 620 is configured to generate, by using a trained generative model, candidate scripts of target elements in an advertisement style from the source data. The post-processing unit 630 is configured to post-process the candidate scripts to obtain available scripts for the target elements. The advertisement delivery unit 640 is configured to deliver the target advertisement to a user device, wherein the target advertisement is displayed on the user device according to the advertisement style, and at least one script of the available scripts is displayed on the target elements of the advertisement style.

[0054] It should be noted that more actions or steps described with reference to FIGS. 1-5 can be implemented by the apparatus 600 shown in FIG. 6. For example, the apparatus 600 can include more modules or units to implement the above-described actions or steps, or some units or modules shown in FIG. 6 can be further configured to implement the above-described actions or steps. Here, no further elaboration is repeated.

[0055] The above describes the technical solutions of the embodiments of the present disclosure for generating scripts of advertisement styles by using generative models with reference to FIGS. 1-6. Compared with the traditional method, the proposed technical solutions have the advantages that by inputting the information of the advertisement to be delivered as source data into the generative model, more rich and high-quality scripts are generated and applied to various advertisement styles, thereby improving the information quantity and effectiveness of the advertisement style, and reducing the workload of the advertiser. Some example implementation manners of the present disclosure are listed as follows.

[0056] In a first aspect, a method for generating a script of an advertisement style is provided, comprising: acquiring source data associated with a target advertisement; generating, by using a trained generative model, candidate scripts of target elements in an advertisement style from the source data; post-processing the candidate scripts to obtain available scripts for the target elements; and delivering the target advertisement to a user device, wherein the target advertisement is displayed on the user device according to the advertisement style, and at least one script of the available scripts is displayed on the target elements of the advertisement style.

[0057] In some embodiments, the method can further include training the generative model, the training including: acquiring original data from a source data repository, the original data including at least one of a landing page, a speech recognition result or a text recognition result of an advertisement material, a title of the advertisement material, and description information of an advertisement object; generating, based on a prompt word for a large language model, a fine-tuning data set from the original data, wherein the prompt word is customized to instruct the large language model to generate scripts for at least one element of the advertisement style; and fine-tuning the generative model using the fine-tuning data set to obtain the trained generative model.

[0058] In some embodiments, the source data associated with the target advertisement comprises at least one of: a landing page of the target advertisement; a speech recognition result or a text recognition result of the target advertisement; a title of the target advertisement; or description information of an ad object of the target advertisement.

[0059] In some embodiments, the post-processing of the candidate scripts comprises: excluding a portion of the candidate scripts from the candidate scripts based on a word blacklist.

[0060] In some embodiments, the post-processing of the candidate scripts comprises: using a trained quality evaluation model to score the candidate scripts in terms of quality; and excluding a portion of the candidate scripts with a quality score lower than a threshold from the candidate scripts.

[0061] In some embodiments, the post-processing of the candidate scripts comprises: excluding a portion of the candidate scripts from the candidate scripts based on at least one risk detection model.

[0062] In some embodiments, the method further comprises: presenting the available scripts to a provider of the target advertisement; and determining to use a selected script for the target element based on a selection of at least one of the presented available scripts by the provider.

[0063] In some embodiments, the target element comprises at least one of: a call-to-action (CTA) button; an ad selling point; or an ad description.

[0064] In some embodiments, the script displayed on the target element is associated with a characteristic of an ad object of the target advertisement.

[0065] In a second aspect, an apparatus for generating scripts in an ad style is provided, comprising: a source data obtaining unit configured to obtain source data associated with a target advertisement; a script generating unit configured to generate, by using a trained generative model, candidate scripts for a target element in an ad style from the source data; a post-processing unit configured to post-process the candidate scripts to obtain available scripts for the target element; and an ad serving unit configured to serve the target advertisement to a user device, wherein the target advertisement is displayed on the user device according to the ad style, and one of the available scripts is displayed on the target element of the ad style.

[0066] In a third aspect, an electronic device is provided, the electronic device comprising: one or more processors; and memory storing one or more programs for execution by the one or more processors, the one or more programs, when executed by the one or more processors, cause the one or more processors to implement the method according to the first aspect.

[0067] In a fourth aspect, the present disclosure provides a computer-readable storage medium comprising machine executable instructions that, when executed by a device, cause the device to perform the method of the first aspect described above.

[0068] In a fifth aspect, the present disclosure provides a computer program product tangibly stored in a non-transitory computer storage medium and comprising machine executable instructions that, when executed by a device, cause the device to perform the method of the first aspect described above.

[0069] FIG. 7 shows a schematic block diagram of an example device 700 that can be used to implement embodiments of the present disclosure. As shown, the device 700 includes a computing unit 701 that can perform various appropriate actions and processes in accordance with computer program instructions stored in a read-only memory (ROM) 702 or computer program instructions loaded into a random access memory (RAM) 703 from a storage unit 706. Various programs and data required by the device 700 to operate can also be stored in the RAM 703. The computing unit 701, the ROM 702, and the RAM 703 are connected to each other by a bus 704. An input / output (I / O) interface 705 is also connected to the bus 704.

[0070] A plurality of components in the device 700 are connected to the I / O interface 705, including: an input unit 706, such as a keyboard, a mouse, etc.; an output unit 707, such as various types of displays, a speaker, etc.; a storage unit 708, such as a magnetic disk, an optical disk, etc.; and a communication unit 709, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 709 allows the device 700 to exchange information / data with other devices through a computer network, such as the Internet, and / or various telecommunication networks.

[0071] The computing unit 701 can be various general and / or special purpose processing components with processing and computing capabilities. Some examples of the computing unit 701 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various specialized artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, a digital signal processor (DSP), and any suitable processor, controller, microcontroller, etc. The computing unit 701 performs various methods and processes described above, such as the method 300. For example, in some embodiments, the method 300 can be implemented as a computer software program tangibly embodied in a machine-readable medium, such as the storage unit 708. In some embodiments, part or all of the computer program can be loaded and / or installed onto the device 700 via the ROM 702 and / or the communication unit 709. When the computer program is loaded onto the RAM 703 and executed by the computing unit 701, one or more steps of the method 300 described above can be performed. Alternatively, in other embodiments, the computing unit 701 can be configured to perform the method 300 by any other suitable means, such as by means of firmware.

[0072] In some embodiments, the methods and processes described above can be implemented as computer program products. Computer program products can include computer readable storage media having computer readable program instructions thereon for performing various aspects of the present disclosure.

[0073] Computer readable storage media can be, for example, but not limited to, electronic, magnetic, optical, electromagnetic, infrared, semiconductor system, or any suitable combination of the foregoing. More specific examples (a non-exhaustive list) of computer readable storage media include the following: an electrical connection having one or more wires, a portable computer diskette, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), a static random access memory (SRAM), a portable compact disc read-only memory (CD-ROM), a digital versatile disk (DVD), a memory stick, a floppy disk, a mechanically encoded device such as punch-cards or punched tape, a magnetically encoded device such as tape, an optically encoded device such as a compact disc (CD) or DVD, and / or any suitable combination of the foregoing. Computer readable program instructions can be downloaded to or from computer readable storage media 708. The computer readable program instructions can cause the computing unit 701 to perform various ones of the methods and processes described above. The computer program instructions can be executed, for example, by a processor such as the computing unit 701. The computer program instructions can be stored in a computer readable storage medium, such as the storage unit 708. The computer program instructions can cause the computing unit 701 to perform the steps of a method described above. The computer program instructions can be executed by the computing unit 701 to cause the computing unit 701 to perform the steps of a method described above.

[0074] Computer readable program instructions described herein can be downloaded to respective computing / processing devices from a computer readable storage medium or to an external computer or external storage device via a network, for example, the Internet, a local area network, a wide area network and / or a wireless network. The network can comprise copper transmission cables, optical transmission fibers, wireless transmission, routers, firewalls, switches, gateway computers and / or edge servers. A network adapter card or network interface in each computing / processing device receives computer readable program instructions from the network and forwards the computer readable program instructions for storage in a computer readable storage medium within the respective computing / processing device.

[0075] Computer readable program instructions for carrying out operations of the present disclosure can be assembly-level instructions, instructions set architecture (ISA) instructions, machine instructions, machine dependent instructions, microcode, firmware instructions, state-setting data, or either source code or object code written in any combination of one or more programming languages, including object oriented programming languages and conventional procedural programming languages. The computer readable program instructions can execute entirely on the user's computer, partly on the user's computer, as a stand-alone software package, partly on the user's computer and partly on a remote computer or entirely on the remote computer or server. In the latter scenario, the remote computer can be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or the connection can be made to an external computer (for example, through the Internet using an Internet Service Provider). In some embodiments, electronic circuitry including, for example, programmable logic circuitry, field-programmable gate array (FPGA), or programmable logic array (PLA) can execute the computer readable program instructions by utilizing state information of the computer readable program instructions to personalize the electronic circuitry, in order to perform aspects of the present disclosure.

[0076] These computer readable program instructions can be provided to a processor of a general purpose computer, special purpose computer, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions / acts specified in the flowchart and / or block diagram block or blocks. These computer readable program instructions can also be stored in a computer readable storage medium that can include a non-transitory computer readable storage medium that can direct a computer, a programmable data processing apparatus, and / or other devices to function in a particular manner, such that the computer readable storage medium having instructions for causing an apparatus to implement one or more functions / acts specified in the flowchart and / or block diagram block or blocks is provided. The instructions of the computer

[0077] The computer readable program instructions can also be loaded onto a computer, other programmable data processing apparatus, or other device to cause a series of operational steps to be performed on the computer, other programmable data processing apparatus or other device to produce a computer implemented process such that the instructions which execute on the computer, other programmable data processing apparatus, or other device implement the functions / acts specified in the flowchart and / or block diagram block or blocks.

[0078] The flow diagrams and the block diagrams in the drawings are presented to illustrate the architecture, functionality, and operation of possible implementations of devices, methods and computer program products according to various embodiments of the present disclosure. In this regard, each block in the flow diagrams and the block diagrams can represent a module, segment, or portion of instructions, which comprises one or more executable instructions for implementing the specified logic functions. In some alternative implementations, the functions noted in the blocks can occur out of the order noted in the figures. For example, two blocks shown in succession may, in fact, be executed substantially concurrently, or the blocks can sometimes be executed in the reverse order, depending upon the functionality involved. It will also be noted that each block of the block diagrams and / or flow diagrams, and combinations thereof, can be implemented by special purpose hardware-based systems that perform the specified functions or acts, or combinations of special purpose hardware and

[0079] Embodiments of the present disclosure have been described above, and the description is intended to be illustrative of the embodiments and not restrictive. Many modifications and variations of the described embodiments are possible and are within the scope of the disclosure. The selection of terms is intended to best describe the principles of the embodiments, practical application, or technical improvements over the technology found in the market, or to enable others skilled in the art to understand the embodiments disclosed herein.

Claims

1. A method for generating copy for an ad style, comprising: obtaining source data associated with a target ad; generating, from the source data, candidate copy for a target element in an ad style by using a trained generative model; post-processing the candidate copy to obtain available copy for the target element; and serving the target ad to a user device, wherein the target ad is displayed on the user device according to the ad style, and one of the available copy is displayed on the target element of the ad style.

2. The method of claim 1, further comprising training the generative model, the training comprising: obtaining training data from a source data repository, the training data comprising at least one of a landing page, speech recognition results or text recognition results of ad materials, titles of ad materials, description information of ad objects; generating, from the training data, a fine-tuning data set based on a prompt word for a large language model, wherein the prompt word is customized to instruct the large language model to generate copy for at least one element of the ad style; and fine-tuning the generative model using the fine-tuning data set to obtain the trained generative model.

3. The method of claim 1, wherein the source data associated with the target ad comprises at least one of: a landing page of the target ad; speech recognition results or text recognition results of the target ad; a title of the target ad; or description information of ad objects of the target ad.

4. The method of claim 1, wherein the post-processing the candidate copy comprises: excluding a portion of the candidate copy from the candidate copy based on a word blacklist.

5. The method of claim 1, wherein the post-processing the candidate copy comprises: quality scoring the candidate copy using a trained quality assessment model; and excluding a portion of the candidate copy based on a quality score lower than a threshold.

6. The method of claim 1, wherein the post-processing the candidate copy comprises: excluding a portion of the candidate copy from the candidate copy based on at least one risk detection model.

7. The method of claim 1, further comprising: presenting the available copy to a provider of the target ad; and determining to use a selected copy for the target element based on a selection of at least a portion of the presented available copy by the provider.

8. The method of claim 1, wherein the target element comprises at least one of: a call-to-action (CTA) button; an ad selling point; or an ad description.

9. The method of any one of claims 1-8, wherein the copy displayed on the target element is associated with a characteristic of an ad object of the target ad.

10. An apparatus for generating copy in an ad style, comprising: a source data obtaining unit configured to obtain source data associated with a target ad; a copy generating unit configured to generate, from the source data, candidate copy for a target element in an ad style by using a trained generative model; ​ ​ ​ ​ a post-processing unit configured to post-process the candidate scripts to obtain available scripts for the target element; and an advertisement delivery unit configured to deliver the target advertisement to a user device, wherein the target advertisement is displayed on the user device according to the advertisement style, and at least one script of the available scripts is displayed on the target element of the advertisement style. 11.An electronic device, comprising: one or more processors; and memory storing one or more programs, the one or more programs, when executed by the one or more processors, cause the one or more processors to implement the method according to any one of claims 1-9. 12.A computer-readable storage medium having stored thereon a computer program, the program, when executed by a processor, implements the method according to any one of claims 1-9. 13.A computer program product comprising a computer program, the computer program, when executed by a processor, implements the method according to any one of claims 1-9.

Images