Method and device for generating promotion information copywriting based on large model and electronic equipment

Abstract

The invention provides a promotion information copywriting generation method and device based on a large model, electronic equipment and a readable storage medium, and relates to the technical field of artificial intelligence such as natural language processing, deep learning and large models. The promotion information copywriting generation method based on the large model comprises the following steps: determining a putting type of an obtained query word by using a putting classification model, and classifying the query word according to the determined putting type; obtaining a keyword cluster according to a classification result of the query word, and obtaining a word root of the keyword cluster; and generating a promotion information copywriting by using a large model according to the root of the keyword cluster. According to the invention, the generation efficiency of the promotion information copywriting and the accuracy of the generated promotion information copywriting can be improved.

Description

Technical Field

[0001] This disclosure relates to the field of computer technology, and more particularly to the fields of artificial intelligence technology such as natural language processing, deep learning, and large-scale models. It provides a method, apparatus, electronic device, and readable storage medium for generating promotional text based on large-scale models. Background Technology

[0002] The quality of promotional copy directly impacts click-through rates. Current technology typically relies on manual writing to obtain promotional copy, which suffers from low efficiency and high costs. Furthermore, manually written promotional copy often needs to be adapted to a large number of keywords, resulting in poor semantic coherence and an inability to cater to personalized user needs. Summary of the Invention

[0003] According to a first aspect of this disclosure, a method for generating promotional copy based on a large model is provided, comprising: using a placement classification model to determine the placement type of the acquired query terms, and classifying the query terms according to the determined placement type; obtaining keyword clusters based on the classification results of the query terms, and obtaining the root words of the keyword clusters; and using a large model to generate promotional copy based on the root words of the keyword clusters.

[0004] According to a second aspect of this disclosure, a promotional copy generation apparatus based on a large model is provided, comprising: a classification unit, configured to determine the delivery type of an acquired query term using a delivery classification model, and classify the query term according to the determined delivery type; a processing unit, configured to obtain a keyword cluster based on the classification result of the query term, and obtain the root words of the keyword cluster; and a generation unit, configured to generate promotional copy using the large model and based on the root words of the keyword cluster.

[0005] According to a third aspect of this disclosure, an electronic device is provided, comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to perform the method as described above.

[0006] According to a fourth aspect of this disclosure, a non-transitory computer-readable storage medium is provided storing computer instructions, wherein the computer instructions are used to cause the computer to perform the method described above.

[0007] According to a fifth aspect of this disclosure, a computer program product is provided, comprising a computer program that, when executed by a processor, implements the method described above.

[0008] It should be understood that the description in this section is not intended to identify key or essential features of the embodiments of this disclosure, nor is it intended to limit the scope of this disclosure. Other features of this disclosure will become readily apparent from the following description. Attached Figure Description

[0009] The accompanying drawings are provided to better understand this solution and do not constitute a limitation of this disclosure. Wherein:

[0010] Figure 1 This is a schematic diagram based on the first embodiment of the present disclosure;

[0011] Figure 2 This is a schematic diagram according to the second embodiment of the present disclosure;

[0012] Figure 3 This is a schematic diagram according to the third embodiment of the present disclosure;

[0013] Figure 4 This is a schematic diagram according to the fourth embodiment of the present disclosure;

[0014] Figure 5 This is a block diagram of an electronic device used to implement the large-model-based promotional copy generation method of the present disclosure embodiments. Detailed Implementation

[0015] The exemplary embodiments of this disclosure are described below with reference to the accompanying drawings, including various details of the embodiments to aid understanding, and should be considered merely exemplary. Therefore, those skilled in the art will recognize that various changes and modifications can be made to the embodiments described herein without departing from the scope and spirit of this disclosure. Similarly, for clarity and brevity, descriptions of well-known functions and mechanisms are omitted in the following description.

[0016] Figure 1 This is a schematic diagram based on the first embodiment of this disclosure. (See diagram below.) Figure 1 As shown, the promotional copy generation method based on a large model in this embodiment specifically includes the following steps:

[0017] S101. Use the delivery classification model to determine the delivery type of the obtained query terms, and classify the query terms according to the determined delivery type;

[0018] S102. Obtain keyword clusters based on the classification results of the query terms, and obtain the root words of the keyword clusters;

[0019] S103. Using a large model, generate promotional copy based on the word roots of the keyword cluster.

[0020] The large-scale model-based promotional copy generation method in this embodiment first uses a targeting classification model to classify the acquired query terms. Then, based on the classification results of the query terms, keyword clusters are obtained and their corresponding root words are acquired. Finally, the large-scale model is used to generate promotional copy based on the acquired root words. This embodiment, based on the generation capabilities of the large-scale model, automatically generates promotional copy based on the acquired root words, which can improve the generation efficiency of promotional copy, reduce the acquisition cost of promotional copy, and since the promotional copy is closely related to the keyword clusters, it can also improve the accuracy of the acquired promotional copy, thereby improving the delivery efficiency and effectiveness of the promotional copy.

[0021] In this embodiment, the promotional information is an advertisement, and the promotional information copy is the advertising copy. The promotional information copy can be advertising copy in a narrow sense (i.e., only including text) or advertising copy in a broad sense (i.e., including text and images). The large model in this embodiment can be a large language model (LLM) specifically used to perform natural language processing tasks, or it can be a multimodal large model.

[0022] In this embodiment, when executing S101, multiple query terms are first obtained, then the delivery classification model is used to determine the delivery type of each query term, and finally, based on the determined delivery type, the multiple query terms are divided into different classification results to complete the classification.

[0023] In this embodiment, the delivery type includes reading page delivery type and tool page delivery type, and may also include other delivery types. This embodiment does not limit this type. Different delivery types correspond to different delivery scenarios.

[0024] In this embodiment, when executing S101, the text content currently entered by the input terminal can be obtained as a query term; or multiple query terms entered by different users within a preset time period (e.g., one day) can be obtained based on search logs from different channels, i.e., query terms can be obtained in batches.

[0025] In this embodiment, obtaining query terms and using them to generate promotional copy is done with the user's prior consent or permission, meaning the user is aware that the query terms they input will be used to generate promotional copy.

[0026] After obtaining the query terms in S101, this embodiment can also perform risk control filtering on the query terms, thereby strictly screening the query terms used to generate promotional copy.

[0027] The targeting classification model used in S101 of this embodiment is pre-trained. This targeting classification model can output the targeting type corresponding to the input query term, such as the reading page targeting type, the tool page targeting type, or other targeting types.

[0028] In this embodiment, after executing S101 to determine the delivery type of the query term using the delivery classification model, the query term can be classified according to the determined delivery type, that is, the query term is assigned to the classification results corresponding to its delivery type; in this embodiment, different delivery types correspond to different classification results.

[0029] For example, if the targeting type includes reading page targeting type and tool page targeting type, this embodiment will include category result 1 corresponding to reading page targeting type and category result 2 corresponding to tool page targeting type. Category result 1 includes query terms for targeting type reading page, and category result 2 includes query terms for targeting type tool page.

[0030] It is understandable that, since this embodiment will acquire query terms in batches and the query terms may also include redundant information, if the query terms are directly put into the corresponding classification results, the classification results will occupy a lot of storage resources. Therefore, after executing S101 to determine the delivery type of the query terms, this embodiment can also extract keywords from the query terms and then put only the extracted keywords into the corresponding classification results.

[0031] In this embodiment, after S101 classifies the query terms according to the determined delivery type, S102 is executed to obtain keyword clusters based on the classification results of the query terms and to obtain the root words of the keyword clusters.

[0032] Specifically, in this embodiment, when executing S101 to obtain keyword clusters based on the classification results of query terms, the implementation method can be as follows: for each classification result, obtain the semantics of each query term in the classification result; based on the obtained semantics, cluster the query terms in the classification result; and obtain keyword clusters based on the clustering results of the query terms; wherein, the query terms in the classification result can be the query terms themselves or keywords extracted from the query terms.

[0033] In other words, after the query terms are first segmented according to the targeting type, this embodiment will further segment the query terms in different classification results according to semantics, thereby obtaining keyword clusters based on the second segmentation, so as to continue the subsequent promotional copy generation process. This embodiment refines the granularity of generating promotional copy by dividing query terms into different keyword clusters, so that the generated promotional copy corresponds to different keyword clusters, which can improve the targeting of promotional copy and thus improve the conversion rate, click-through rate, etc. of promotional information.

[0034] In this embodiment, after obtaining the keyword cluster by executing S102, the root word of each keyword cluster can be obtained; wherein, only one root word will be obtained for each keyword cluster.

[0035] In this embodiment, the root of a keyword cluster is a word or phrase that can carry the main meaning of the query term (or the keyword of the query term) in the keyword cluster.

[0036] In this embodiment, when executing S102 to obtain the word roots of the keyword cluster, the following implementation method can be adopted: input the query terms in the keyword cluster into the large model; obtain the word roots of the keyword cluster based on the first text content output by the large model.

[0037] In other words, this embodiment leverages the text processing capabilities of a large model to generate word roots for keyword clusters based on the input query terms, thereby improving the efficiency and accuracy of word root acquisition.

[0038] In this embodiment, when executing S102, a first preset prompt text (i.e., prompt) can also be obtained. The first preset prompt text is used to prompt the large model to generate word roots based on the input query words. The query words in the keyword cluster and the first preset prompt text are input together into the large model so that the large model can generate the first text content.

[0039] In this embodiment, after executing S102 to obtain the word roots of the keyword cluster, S103 is executed to use the large model to generate promotional copy based on the word roots of the keyword cluster.

[0040] In this embodiment, the promotional copy generated by the large model corresponds to the keyword cluster. Since each keyword cluster can determine the corresponding placement type based on its classification results, the promotional copy generated by executing S103 in this embodiment can also correspond to the corresponding placement type, thereby achieving the purpose of generating promotional copy corresponding to different placement types.

[0041] In this embodiment, when executing S103 using the large model to generate promotional copy based on the word roots of the keyword cluster, the implementation method can be as follows: input the word roots of the keyword cluster into the large model; obtain the promotional copy based on the second text content output by the large model; in this embodiment, the large model can generate one promotional copy or multiple promotional copy based on the input word roots, and this embodiment does not limit this.

[0042] In this embodiment, when executing S103, a second preset prompt text can also be obtained. The second preset prompt text is used to prompt the large model to generate promotional information copy based on the input word roots. Then, the word roots of the keyword cluster and the second preset prompt text are input into the large model together so that the large model can generate the second text content.

[0043] In this embodiment, when executing S103 to obtain the promotional information copy based on the second text content output by the large model, the following content may also be included: inputting the second text content into the large model to obtain the target image output by the large model; and obtaining the promotional information copy based on the second text content and the target image.

[0044] In other words, this embodiment can also use a large model to further generate a target image based on the generated second text content, and then combine the second text content and the target image to obtain promotional copy, thereby improving the richness of the content included in the promotional copy.

[0045] In addition, when executing S103 to input the second text content into the large model and obtain the target image output by the large model, this embodiment can also obtain target information according to the delivery type corresponding to the keyword cluster, such as preset manufacturer information, spokesperson information, delivery scenario information, etc., and then input the target information and the second text content into the large model together so that the large model can generate the target image.

[0046] In addition, when executing S103 to input the word roots of the keyword cluster into the large model, this embodiment may also include the following: obtaining the target generation strategy according to the delivery type corresponding to the keyword cluster; inputting the word roots of the keyword cluster and the target generation strategy into the large model so that the large model can combine the target generation strategy and the word roots of the keyword cluster to generate the second text content.

[0047] In this embodiment, the generation strategy for the reading page delivery type can be RAG (Retrieval Augmented Generation), which enables the large model to integrate external knowledge (such as resources within the library) into the generation process of promotional copy; the generation strategy for the tool page delivery type can be CoT (Chain of Thought), which enables the large model to first perform corresponding reasoning (such as the corresponding AI function, AI function application scenarios, etc.), and then combine the reasoning results to generate promotional copy.

[0048] In other words, this embodiment can pre-set generation strategies corresponding to different campaign types, enabling the large model to combine target generation strategies to generate promotional copy, further improving the accuracy of promotional copy generation and significantly enhancing the creativity and richness of the generated promotional copy.

[0049] In this embodiment, after obtaining the promotional information copy in step S103, the promotional information copy can be delivered according to the corresponding delivery type. After clicking on the corresponding promotional information copy, the user can enter the corresponding page. For example, after clicking on the promotional information copy for the corresponding reading page delivery type, the user can enter the reading page to read the relevant content; or after clicking on the promotional information copy for the corresponding tool page delivery type, the user can enter the tool page to use relevant AI tools (such as AI-based PPT creation tools, AI-based chart creation tools, or AI-based resume creation tools, etc.).

[0050] Figure 2 This is a schematic diagram according to the second embodiment of this disclosure. (See diagram below.) Figure 2 As shown in the figure, this embodiment illustrates the training process of the classification model, which specifically includes the following steps:

[0051] S201. Input the sample query terms into multiple candidate large models to obtain the candidate delivery types output by the multiple candidate large models respectively;

[0052] S202. Based on multiple candidate delivery types, determine the label delivery type corresponding to the sample query term;

[0053] S203. Using the sample query terms and their corresponding labeled delivery types, train the neural network model to obtain the delivery classification model.

[0054] In other words, this embodiment fully leverages the semantic understanding capabilities of large models, uses multiple candidate large models to automatically label sample query terms, and then obtains training data for training the distribution classification model based on the sample query terms and their corresponding labeling and delivery types. This simplifies the steps for obtaining training data, improves the efficiency of obtaining training data, and reduces the cost of obtaining training data.

[0055] In this embodiment, when executing S201, sample query terms can be obtained in batches based on the search log; similarly, obtaining the user's query terms and using them accordingly has been pre-approved by the user.

[0056] In this embodiment, when executing S201, a third preset prompt text can also be obtained. The third preset prompt text is used to prompt the candidate big model to generate candidate delivery types based on the input sample query terms. Then, the sample query terms and the third preset prompt text are input together into the candidate big model so that the candidate big model can generate candidate delivery types.

[0057] This embodiment does not limit the type of candidate large model; it can be any common open-source large language model available on the market.

[0058] In this embodiment, when executing S202 to determine the labeled delivery type of the corresponding sample query term based on multiple candidate delivery types, a voting method can be used to determine the number of votes for the reading page delivery type and the tool page delivery type based on multiple candidate delivery types, and then the delivery type with more votes is taken as the labeled delivery type.

[0059] In addition, when there is a need to add or delete delivery types, it is only necessary to modify the third preset prompt text according to the added or deleted delivery type, thereby effectively performing data mining and realizing flexible expansion of the delivery classification model.

[0060] In this embodiment, the neural network model trained using sample query terms and their corresponding annotation delivery types in step S203 can be a classification model with any network structure, or a lightweight version of a specific large model (such as ERNIE).

[0061] Figure 3 This is a schematic diagram according to the third embodiment of this disclosure. (See diagram below.) Figure 3As shown in the figure, this embodiment illustrates a flowchart for delivering promotional copy: S301, obtain multiple query terms based on search logs, i.e., perform batch query term acquisition; S302, perform risk control filtering on the obtained multiple query terms. It can be understood that this embodiment can also filter out completely identical query terms among multiple query terms, retaining only one; S303, use a delivery classification model to classify the multiple query terms; S304, based on the semantics of the query terms in the classification results, cluster the query terms to obtain keyword clusters; S305, use a large model to generate the root words of the keyword clusters based on the keywords in the keyword clusters; S306, use the large model to generate promotional copy based on the root words of the keyword clusters. In this embodiment, the large model can generate promotional text in batches; S307, establish the correspondence between the generated promotional text and different placement types, for example, store the promotional text in the account corresponding to the placement type; S308, according to the detected placement demand, obtain the promotional text corresponding to the corresponding placement type for display, for example, if the user enters the query term "primary school students' spring outing composition", the placement type corresponding to this placement demand is "reading page placement type", then obtain the promotional text from the account corresponding to "reading page placement type" for display. After the user clicks on the displayed promotional text, they can jump to the reading page, which includes documents related to "primary school students' spring outing composition".

[0062] In other words, this embodiment can process batch-acquired query terms to generate promotional copy in batches, generating a large number of promotional copy in a short time, effectively improving the generation and delivery efficiency of promotional copy.

[0063] Figure 4 This is a schematic diagram according to the fourth embodiment of this disclosure. (See diagram below.) Figure 4 As shown, the large-model-based promotional copywriting generation device 400 of this embodiment includes:

[0064] Classification unit 401 is used to determine the delivery type of the obtained query terms using a delivery classification model, and to classify the query terms according to the determined delivery type.

[0065] Processing unit 402 is used to obtain keyword clusters based on the classification results of the query terms, and to obtain the root words of the keyword clusters;

[0066] The generation unit 403 is used to generate promotional copy based on the word roots of the keyword cluster using a large model.

[0067] In this embodiment, the promotional information is an advertisement, and the promotional information copy is the advertising copy. The promotional information copy can be advertising copy in a narrow sense (i.e., only including text) or advertising copy in a broad sense (i.e., including text and images). The large model in this embodiment can be a large language model (LLM) specifically used to perform natural language processing tasks, or it can be a multimodal large model.

[0068] Classification unit 401 first obtains multiple query terms, then uses the delivery classification model to determine the delivery type of each query term, and finally classifies the multiple query terms into different classification results according to the determined delivery type to complete the classification.

[0069] In this embodiment, the delivery type includes reading page delivery type and tool page delivery type, and may also include other delivery types. This embodiment does not limit this type. Different delivery types correspond to different delivery scenarios.

[0070] The classification unit 401 can obtain the text content currently entered by the input terminal as a query term; it can also obtain multiple query terms entered by different users within a preset time period (e.g., one day) based on search logs from different channels, i.e., batch obtain query terms.

[0071] In this embodiment, obtaining query terms and using them to generate promotional copy is done with the user's prior consent or permission, meaning the user is aware that the query terms they input will be used to generate promotional copy.

[0072] After obtaining the query terms, classification unit 401 can also perform risk control filtering on the query terms, thereby strictly screening the query terms used to generate promotional copy.

[0073] The targeting classification model used by classification unit 401 is pre-trained. This targeting classification model can output the targeting type corresponding to the input query term, such as the reading page targeting type, the tool page targeting type, or other targeting types.

[0074] After the classification unit 401 determines the delivery type of the query term using the delivery classification model, it can classify the query term according to the determined delivery type, that is, classify the query term into the classification results corresponding to its delivery type; in this embodiment, different delivery types correspond to different classification results.

[0075] It is understandable that, since this embodiment will acquire query terms in batches and the query terms may also include redundant information, if the query terms are directly put into the corresponding classification results, the classification results will occupy more storage resources. Therefore, after determining the delivery type of the query terms, the classification unit 401 can also extract keywords from the query terms and then put only the extracted keywords into the corresponding classification results.

[0076] In this embodiment, after the classification unit 401 classifies the query terms according to the determined delivery type, the processing unit 402 obtains the keyword clusters based on the classification results of the query terms and obtains the root words of the keyword clusters.

[0077] Specifically, when processing unit 402 obtains keyword clusters based on the classification results of query terms, it can do so by: obtaining the semantics of each query term in each classification result; clustering the query terms in the classification result based on the obtained semantics; and obtaining keyword clusters based on the clustering results of the query terms. The query terms in the classification results can be the query terms themselves or keywords extracted from the query terms.

[0078] In other words, after the query terms are first divided according to the delivery type, the processing unit 402 will further divide the query terms into different categories based on semantics, thereby obtaining keyword clusters based on the second division, so as to continue the subsequent promotional copy generation process. This embodiment refines the granularity of generating promotional copy by dividing query terms into different keyword clusters, so that the generated promotional copy corresponds to different keyword clusters, which can improve the targeting of promotional copy and thus improve the conversion rate, click-through rate, etc. of promotional information.

[0079] After obtaining the keyword clusters, the processing unit 402 can obtain the root word of each keyword cluster; however, only one root word will be obtained for each keyword cluster.

[0080] In this embodiment, the root of a keyword cluster is a word or phrase that can carry the main meaning of the query term (or the keyword of the query term) in the keyword cluster.

[0081] When processing unit 402 obtains the root words of keyword clusters, it can do so by: inputting the query terms in the keyword clusters into the large model; and obtaining the root words of the keyword clusters based on the first text content output by the large model.

[0082] In other words, the processing unit 402 uses the text processing capabilities of the large model to generate the word roots of the keyword cluster based on the input query terms, which can improve the efficiency and accuracy of word root acquisition.

[0083] The processing unit 402 can also obtain a first preset prompt text (i.e., prompt). The first preset prompt text is used to prompt the large model to generate word roots based on the input query words. The query words in the keyword cluster and the first preset prompt text are input into the large model together so that the large model can generate the first text content.

[0084] In this embodiment, after the processing unit 402 obtains the word roots of the keyword cluster, the generation unit 403 uses a large model to generate promotional copy based on the word roots of the keyword cluster.

[0085] In this embodiment, the promotional copy generated by the large model corresponds to the keyword cluster. Since each keyword cluster can determine the corresponding placement type based on its classification results, the promotional copy generated by the generation unit 403 can also correspond to the corresponding placement type, thereby achieving the purpose of generating promotional copy corresponding to different placement types.

[0086] When the generation unit 403 uses the large model to generate promotional copy based on the word roots of the keyword cluster, the following implementation method can be adopted: input the word roots of the keyword cluster into the large model; obtain the promotional copy based on the second text content output by the large model; in this embodiment, the large model can generate one promotional copy or multiple promotional copy based on the input word roots, and this embodiment does not limit this.

[0087] The generation unit 403 can also obtain a second preset prompt text. The second preset prompt text is used to prompt the large model to generate promotional information copy based on the input word roots. Then, the word roots of the keyword cluster and the second preset prompt text are input into the large model together so that the large model can generate the second text content.

[0088] When generating unit 403 obtains promotional copy based on the second text content output by the large model, it may also include the following: inputting the second text content into the large model to obtain the target image output by the large model; and obtaining the promotional copy based on the second text content and the target image.

[0089] In other words, the generation unit 403 can also generate a target image based on the generated second text content by the large model, and then combine the second text content and the target image to obtain the promotional copy, thereby improving the richness of the content included in the promotional copy.

[0090] In addition, when the generation unit 403 inputs the second text content into the large model and obtains the target image output by the large model, it can also obtain target information according to the delivery type corresponding to the keyword cluster, such as preset manufacturer information, spokesperson information, delivery scenario information, etc., and then input the target information together with the second text content into the large model so that the large model can generate the target image.

[0091] In addition, when the generation unit 403 inputs the word roots of the keyword cluster into the large model, it may also include the following: obtaining the target generation strategy according to the delivery type corresponding to the keyword cluster; inputting the word roots of the keyword cluster and the target generation strategy into the large model so that the large model can combine the target generation strategy and the word roots of the keyword cluster to generate the second text content.

[0092] In this embodiment, the generation strategy for the reading page delivery type can be RAG (Retrieval Augmented Generation), which enables the large model to integrate external knowledge (such as resources within the library) into the generation process of promotional copy; the generation strategy for the tool page delivery type can be CoT (Chain of Thought), which enables the large model to first perform corresponding reasoning (such as the corresponding AI function, AI function application scenarios, etc.), and then combine the reasoning results to generate promotional copy.

[0093] In other words, the generation unit 403 can pre-set generation strategies corresponding to different delivery types, enabling the large model to combine the target generation strategy to complete the generation of promotional copy, further improving the accuracy of promotional copy generation, and significantly enhancing the creativity and richness of the generated promotional copy.

[0094] After obtaining the promotional copy, the generation unit 403 can deliver the promotional copy according to the corresponding delivery type. After clicking on the corresponding promotional copy, users can enter the corresponding page. For example, clicking on the promotional copy for the reading page delivery type will take them to a reading page to read related content, or clicking on the promotional copy for the tool page delivery type will take them to a tool page to use related AI tools (such as AI-based PPT creation tools, AI-based chart creation tools, or AI-based resume creation tools, etc.).

[0095] The large-model-based promotional copy generation device 400 in this embodiment further includes a training unit 404, which is used to train a delivery classification model in the following manner: inputting sample query terms into multiple candidate large models to obtain candidate delivery types output by the multiple candidate large models respectively; determining the labeled delivery type of the corresponding sample query terms based on the multiple candidate delivery types; and training the neural network model using the sample query terms and their corresponding labeled delivery types to obtain the delivery classification model.

[0096] In other words, training unit 404 fully leverages the semantic understanding capabilities of large models, using multiple candidate large models to automatically label sample query terms, and then obtains training data for training the distribution classification model based on the sample query terms and their corresponding labeling and delivery types. This simplifies the steps for obtaining training data, improves the efficiency of obtaining training data, and reduces the cost of obtaining training data.

[0097] The acquisition, storage, and application of user personal information involved in the technical solution disclosed herein comply with the provisions of relevant laws and regulations and do not violate public order and good morals.

[0098] According to embodiments of this disclosure, this disclosure also provides an electronic device, a readable storage medium, and a computer program product.

[0099] like Figure 5 The diagram shown is a block diagram of an electronic device for a large-model-based promotional copy generation method according to an embodiment of the present disclosure. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device may also represent various forms of mobile devices, such as personal digital processors, cellular phones, smartphones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely illustrative and are not intended to limit the implementation of the present disclosure described and / or claimed herein.

[0100] like Figure 5 As shown, device 500 includes a computing unit 501, which can perform various appropriate actions and processes based on a computer program stored in read-only memory (ROM) 502 or a computer program loaded from storage unit 508 into random access memory (RAM) 503. RAM 503 may also store various programs and data required for the operation of device 500. The computing unit 501, ROM 502, and RAM 503 are interconnected via bus 504. Input / output (I / O) interface 505 is also connected to bus 504.

[0101] Multiple components in device 500 are connected to I / O interface 505, including: input unit 506, such as keyboard, mouse, etc.; output unit 507, such as various types of displays, speakers, etc.; storage unit 508, such as disk, optical disk, etc.; and communication unit 509, such as network card, modem, wireless transceiver, etc. Communication unit 509 allows device 500 to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks.

[0102] The computing unit 501 can be a variety of general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of the computing unit 501 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various special-purpose 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 501 performs the various methods and processes described above, such as a large-model-based promotional copy generation method. For example, in some embodiments, the large-model-based promotional copy generation method can be implemented as a computer software program tangibly contained in a machine-readable medium, such as storage unit 508.

[0103] In some embodiments, part or all of the computer program may be loaded and / or installed on the device 500 via ROM 502 and / or communication unit 509. When the computer program is loaded into RAM 503 and executed by computing unit 501, one or more steps of the large-model-based promotional copy generation method described above may be performed. Alternatively, in other embodiments, computing unit 501 may be configured to perform the large-model-based promotional copy generation method by any other suitable means (e.g., by means of firmware).

[0104] Various implementations of the systems and techniques described herein can be implemented in digital electronic circuit systems, integrated circuit systems, field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), application-specific standard products (ASSPs), systems-on-a-chip (SoCs), complex programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various implementations may include implementations in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which may be a dedicated or general-purpose programmable processor, capable of receiving data and instructions from a storage system, at least one input device, and at least one output device, and transferring data and instructions to the storage system, the at least one input device, and the at least one output device.

[0105] The program code used to implement the methods of this disclosure may be written in any combination of one or more programming languages. This program code may be provided to the processor or controller of a general-purpose computer, special-purpose computer, or other programmable model-based promotional copy generation device, such that when executed by the processor or controller, the functions / operations specified in the flowcharts and / or block diagrams are implemented. The program code may be executed entirely on a machine, partially on a machine, as a standalone software package partially on a machine and partially on a remote machine, or entirely on a remote machine or server.

[0106] In the context of this disclosure, a machine-readable medium can be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, apparatus, or device. A machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. Machine-readable media can be, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

[0107] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device for showing information to the user (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor); and a keyboard and pointing device (e.g., a mouse or trackball) through which the user provides input to the computer. Other types of devices can also be used to provide interaction with the user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including sound input, voice input, or tactile input).

[0108] The systems and technologies described herein can be implemented in computing systems that include backend components (e.g., as a data server), or computing systems that include middleware components (e.g., an application server), or computing systems that include frontend components (e.g., a user computer with a graphical user interface or web browser through which a user can interact with embodiments of the systems and technologies described herein), or any combination of such backend, middleware, or frontend components. The components of the system can be interconnected via digital data communication of any form or medium (e.g., a communication network). Examples of communication networks include local area networks (LANs), wide area networks (WANs), and the Internet.

[0109] Computer systems can include clients and servers. Clients and servers are generally geographically separated and typically interact via communication networks. The client-server relationship is created by computer programs running on the respective computers and having a client-server relationship with each other. A server can be a cloud server, also known as a cloud computing server or cloud host, a hosting product within the cloud computing service system that addresses the management difficulties and weak business scalability inherent in traditional physical hosts and VPS (Virtual Private Server) services. Servers can also be servers for distributed systems or servers incorporating blockchain technology.

[0110] It should be understood that the various forms of processes shown above can be used to rearrange, add, or delete steps. For example, the steps described in this disclosure can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution disclosed in this disclosure can be achieved, and this is not limited herein.

[0111] The specific embodiments described above do not constitute a limitation on the scope of protection of this disclosure. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this disclosure should be included within the scope of protection of this disclosure.

Claims

1. A method for generating promotional copy based on a large model, comprising: The target audience classification model is used to determine the target audience type of the obtained query terms, and the query terms are classified according to the determined target audience type. Based on the classification results of the query terms, keyword clusters are obtained, and the root words of the keyword clusters are acquired. Using a large model, promotional copy is generated based on the word roots of the keyword cluster.

2. The method according to claim 1, further comprising: Input the sample query terms into multiple candidate large models to obtain the candidate delivery types output by the multiple candidate large models respectively; Based on multiple candidate delivery types, determine the label delivery type corresponding to the sample query term; The neural network model is trained using the sample query terms and their corresponding labeled delivery types to obtain the delivery classification model.

3. The method according to claim 1, wherein, The keyword clusters obtained based on the classification results of the query terms include: For each classification result, obtain the semantics of each query term in that classification result; Based on the obtained semantics, cluster the query terms in the classification results; The keyword clusters are obtained based on the clustering results of the query terms.

4. The method according to claim 1, wherein, The process of obtaining the root words of the keyword cluster includes: Input the query terms from the keyword cluster into the large model; Based on the first text content output by the large model, the root words of the keyword cluster are obtained.

5. The method according to claim 1, wherein, The use of a large model to generate promotional copy based on the root words of the keyword cluster includes: Input the root words of the keyword cluster into the large model; Based on the second text content output by the large model, the promotional information copy is obtained.

6. The method according to claim 5, wherein, The step of obtaining the promotional copy based on the second text content output by the large model includes: Input the second text content into the large model, and obtain the target image output by the large model; Based on the second text content and the target image, the promotional information copy is obtained.

7. The method according to claim 5, wherein, The step of inputting the root words of the keyword cluster into the large model includes: Based on the targeting type corresponding to the keyword cluster, obtain the target generation strategy; The root words of the keyword cluster and the target generation strategy are input into the large model.

8. A device for generating promotional copy based on a large model, comprising: The classification unit is used to determine the delivery type of the acquired query terms using a delivery classification model, and to classify the query terms according to the determined delivery type. The processing unit is used to obtain keyword clusters based on the classification results of the query terms, and to obtain the root words of the keyword clusters; The generation unit is used to generate promotional copy based on the word roots of the keyword cluster using a large model.

9. An electronic device, comprising: At least one processor; as well as A memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor to enable the at least one processor to perform the method of any one of claims 1-7.

10. A non-transitory computer-readable storage medium storing computer instructions, wherein, The computer instructions are used to cause the computer to perform the method according to any one of claims 1-7.

11. A computer program product comprising a computer program that, when executed by a processor, implements the method according to any one of claims 1-7.

Images