Short message copywriting generation method and device, equipment, medium and computer program product

Abstract

The invention discloses a short message copywriting generation method and device, equipment, a medium and a computer program product, and relates to the technical field of data processing, the method comprises the following steps: obtaining externally input generation demand information for generating a short message copywriting, the generation demand information comprising at least one of an activity theme, a target customer group and an activity keyword; the generation demand information is input into a preset large language model for prediction, multiple short message copywriting is obtained, the large language model comprises text preference strategies of different customer groups, multiple first activity texts generated according to the activity theme and the activity keyword are adjusted according to the text preference strategy corresponding to the target customer group, and multiple short message copywriting information is obtained. Obtaining a short message copywriting corresponding to each first activity text; and selecting a target short message copywriting from the plurality of short message copywriting and sending the target short message copywriting to target equipment, wherein the target equipment is equipment used by customers in the target customer group. According to the method, the short message copywriting meeting the personalized requirements of clients can be quickly generated, and the marketing effect is improved.

Description

Technical Field

[0001] The present application relates to a specific field of image processing technology, and in particular to a shooting method, device, equipment, medium and computer program product based on a machine learning model. Background Art

[0002] In today's highly competitive market environment, enterprises, banks and other organizations are increasingly relying on various forms of marketing activities to attract new customers and maintain existing customer relationships, such as SMS marketing. However, SMS marketing has certain defects, that is, SMS marketing texts are generally written manually, which is inefficient and difficult to quickly adapt to market demand and customer personalized needs, resulting in poor marketing results. Therefore, how to quickly generate SMS texts that meet customer personalized needs and improve marketing results has become an urgent problem to be solved.

[0003] The above contents are only used to assist in understanding the technical solution of the present application and do not constitute an admission that the above contents are prior art. Summary of the invention

[0004] The main purpose of this application is to provide a method, device, equipment, medium and computer program product for generating text messages, aiming to solve the technical problem of how to quickly generate text messages that meet the personalized needs of customers and improve marketing effectiveness.

[0005] To achieve the above purpose, the present application proposes a method for generating text messages, the method comprising:

[0006] Acquire externally input generation requirement information for generating SMS copy, wherein the generation requirement information includes at least one of an activity theme, a target customer group, and an activity keyword;

[0007] Inputting the generated demand information into a preset large language model for prediction to obtain a plurality of SMS texts, wherein the large language model includes text preference strategies of different customer groups, and adjusting a plurality of first activity texts generated according to the activity theme and the activity keywords according to the text preference strategies corresponding to the target customer groups to obtain SMS texts corresponding to each first activity text;

[0008] A target SMS copy is selected from the multiple SMS copies and sent to a target device, wherein the target device is a device used by a customer in the target customer group.

[0009] In one embodiment, the method for generating a short message copy further includes:

[0010] Acquire a training data set including at least one task data, wherein the task data includes a training SMS copy and label information, and the label information includes a training activity theme corresponding to the training SMS copy, a training activity keyword, and a training customer group corresponding to the training SMS copy;

[0011] Inputting the label information into the large language model to be trained for prediction, and obtaining a plurality of predicted SMS texts, wherein the large language model to be trained adjusts a plurality of second activity texts generated according to the training activity theme and the training activity keywords according to a text preference strategy corresponding to the training customer group, and obtains a predicted SMS text corresponding to each second activity text;

[0012] A total loss function value is calculated based on the training SMS text and the plurality of predicted SMS texts, and the large language model to be trained is optimized based on the total loss function value until a preset training end condition is met, thereby obtaining a preset large language model.

[0013] In one embodiment, before the step of inputting the label information into the large language model to be trained for prediction, the step includes:

[0014] For any customer group, determine each training SMS copy belonging to the customer group in the training data set, and determine the copy preference information according to the preset copy preference model and each training SMS copy belonging to the customer group;

[0015] Constructing a copywriting preference strategy for the customer group based on the copywriting preference information;

[0016] The copywriting preference strategy of each of the customer groups is updated to the large language model to be trained.

[0017] In one embodiment, the large language model includes an adapter and a model body, and the step of optimizing the large language model to be trained according to the total loss function value includes:

[0018] The model parameters in the adapter are optimized according to the total loss function value.

[0019] In one embodiment, after the step of selecting a target SMS text from the plurality of first SMS texts and sending the target SMS text to the target device, the step includes:

[0020] If feedback information regarding the target SMS copy sent by the target device is received, updating the training data set according to the feedback information;

[0021] The large language model is optimized according to the updated training data set to obtain an optimized large language model.

[0022] In one embodiment, the step of selecting a target SMS text from the plurality of SMS texts and sending it to a target device includes:

[0023] Sorting the multiple SMS texts according to a preset sorting rule, and selecting a SMS text from the sorted multiple SMS texts as a target SMS text;

[0024] The target SMS text is sent to the target device via a preset sending channel, wherein the sending channel includes at least one of SMS, email, and social media.

[0025] In addition, to achieve the above purpose, the present application also proposes a text message generation device, comprising:

[0026] An acquisition module, used to acquire generation requirement information input from the outside for generating SMS copy, wherein the generation requirement information includes at least one of an activity theme, a target customer group, and an activity keyword;

[0027] A prediction module, used for inputting the generated demand information into a preset large language model for prediction, and obtaining a plurality of SMS texts, wherein the large language model includes text preference strategies of different customer groups, and adjusting a plurality of first activity texts generated according to the activity theme and the activity keywords according to the text preference strategies corresponding to the target customer groups, and obtaining SMS texts corresponding to each first activity text;

[0028] The sending module is used to select a target SMS copy from the multiple SMS copies and send it to a target device, wherein the target device is a device used by customers in the target customer group.

[0029] In addition, to achieve the above-mentioned purpose, the present application also proposes a text message generation device, which includes: a memory, a processor, and a computer program stored in the memory and executable on the processor, and the computer program is configured to implement the steps of the text message generation method described above.

[0030] In addition, to achieve the above-mentioned purpose, the present application also proposes a medium, which is a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, the steps of the SMS text generation method described above are implemented.

[0031] In addition, to achieve the above-mentioned purpose, the present application also provides a computer program product, which includes a computer program, and when the computer program is executed by a processor, the steps of the SMS text generation method as described above are implemented.

[0032] In the present application, by obtaining the generation demand information including at least one of the activity theme, target customer group and activity keywords inputted from the outside, and inputting the generation demand information into the large language model for prediction, and the large language model contains the text preference strategies of different customer groups, and according to the text preference strategies corresponding to the target customer group, the multiple first activity texts generated according to the activity theme and the activity keywords are adjusted to obtain multiple SMS copy. And the target SMS copy is selected from the multiple SMS copy to send to the target device used by the customers in the target customer group. It can be realized that the target SMS copy that meets the target customer group is automatically generated by the large language model, which significantly improves the efficiency of SMS copy generation and reduces labor costs compared to the manual writing method. And when the target SMS copy is generated by using the large language model, the SMS copy is generated according to the text preference strategy of the target customer group, and then one of the generated multiple SMS copy is selected as the target SMS copy and sent to the target device used by the customers in the target customer group, so that the generated SMS copy can meet the personalized needs of the target customer group, that is, the SMS copy that meets the personalized needs of customers is quickly generated, and the marketing effect is improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and, together with the description, serve to explain the principles of the present application.

[0034] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative labor.

[0035] Figure 1 A flowchart of the first embodiment of the method for generating text messages of this application is provided;

[0036] Figure 2 A flowchart of the second embodiment of the method for generating text messages of this application is provided;

[0037] Figure 3 This is a schematic diagram of the overall scenario flow in the method for generating SMS text for this application;

[0038] Figure 4 This is a schematic diagram of the module architecture of the device for generating text messages for this application;

[0039] Figure 5 Schematic diagram of the device structure of the hardware operating environment involved in the SMS text generation method in the embodiment of the present application.

[0040] The purpose, features and advantages of this application will be further described in conjunction with the embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0041] It should be understood that the specific embodiments described herein are only used to explain the technical solutions of the present application and are not used to limit the present application.

[0042] In order to better understand the technical solution of the present application, a detailed description will be given below in conjunction with the accompanying drawings and specific implementation methods.

[0043] Since the SMS marketing program of the sensor relies on manual writing, it is difficult to achieve large-scale personalized customization, and different customer groups have different interests and needs, and the unified SMS content is difficult to meet the expectations of all customers. In other words, the SMS marketing program written by humans has the following defects: the cost of manual writing is high. Since the writing of SMS copy usually requires a professional copywriting team, it consumes a lot of manpower and time, and the efficiency is low. The content lacks diversity. The content of manually written SMS often lacks creativity and is difficult to stand out from many homogeneous information, resulting in poor marketing results. And because different customer groups have different interests and needs, the unified SMS content is difficult to meet the expectations of all customers and lacks pertinence and personalization. The response speed is slow. Market demand and customer preferences change rapidly. The content of manually written SMS is difficult to respond to these changes in a timely manner, affecting the flexibility and timeliness of marketing activities.

[0044] Therefore, in this embodiment, advanced natural language processing and artificial intelligence technologies are used to automatically generate diversified, creative and personalized marketing SMS messages, thereby improving the effect and efficiency of SMS marketing. Optionally, in this embodiment, by training the large language model, the SMS copy of diversified marketing SMS messages can be automatically generated according to the trained large language model, reducing the time and cost of manual writing. And through fine-tuning and optimization of the large language model, the generated SMS copy has high creativity and attractiveness, and can stand out from many homogeneous information. It can also quickly respond to changes in market demand and customer preferences according to the large language model, generate timely and relevant SMS copy, and improve the flexibility and effect of marketing activities. It can support the distribution of generated SMS copy through multiple channels, such as SMS, email and social media, to ensure that marketing information can be widely disseminated. By collecting feedback data from marketing activities, the generation effect of the large language model and the quality of SMS content can be continuously optimized, and the effect and efficiency of marketing activities can be continuously improved. Since the SMS copy is generated by the large language model, the labor cost and time cost are significantly reduced compared to the manual writing method.

[0045] It should be noted that the execution subject of this embodiment can be a computing service device with data processing, network communication and program running functions, such as a tablet computer, a personal computer, a mobile phone, etc., or an electronic device capable of realizing the above functions, a text message generation device, etc. The text message generation device is taken as an example to illustrate this embodiment and the following embodiments.

[0046] Based on this, the present application embodiment provides a method for generating a text message. Figure 1 , Figure 1 This is a flow chart of the first embodiment of the method for generating SMS text in the present application.

[0047] In this embodiment, the method for generating a short message copy includes steps S10 to S40.

[0048] Step S10, obtaining generation requirement information input from the outside for generating SMS copy, wherein the generation requirement information includes at least one of an activity theme, a target customer group, and an activity keyword;

[0049] Optionally, the activity theme may be an activity theme for a marketing activity, such as a festival activity, a discount activity, an installment activity, a card application activity, a financial activity, etc. For example, the activity theme may be a summer promotion.

[0050] Optionally, the customer group may be a group that divides multiple customers according to preset rules, such as dividing according to gender, there may be a female customer group and a male customer group. Dividing according to age group, there may be a customer group of 18-30 years old, a customer group of 50-60 years old, etc. Dividing according to occupation, there may be a teacher customer group, a student customer group, etc. Each type of customer group has a corresponding group type, and the group type may be identification information of the customer group, such as teachers, students, etc. The target customer group may be a customer group input by the user.

[0051] Optionally, the activity keywords may be keywords related to the marketing activity, such as activity date information, activity content, etc. Optionally, the activity keywords may include product features, promotion details, and interest points of the target customer group related to the marketing activity.

[0052] Optionally, the SMS copy generation device may receive generation requirement information for generating SMS copy input from the outside. For example, the generation requirement information may be received from a user, or from another terminal device. Optionally, the generation requirement information may not be limited to the activity theme, target customer group, and activity keywords, but may also include other information, such as the format of the SMS copy to be generated.

[0053] Optionally, when the SMS copy generation device only receives the activity theme input from the outside, the activity keyword can be determined according to a preset rule. For example, the SMS copy generation device obtains the current date information and determines the holiday activity according to the current date information. The holiday activity is used as the activity keyword. Optionally, after determining the activity keyword, the activity theme, the determined activity keyword, and each customer group can be used as the generated demand information.

[0054] Optionally, when the SMS copywriting generation device only receives activity keywords input from the outside, the activity theme corresponding to the activity keyword can be predicted according to preset rules, such as performing cluster analysis on each input activity keyword to obtain the activity theme. Optionally, after determining the activity theme, the activity keyword, the determined activity theme and each customer group can be used as the generated demand information.

[0055] Optionally, after the SMS copy generation device receives only the customer group input from the outside, it can determine the activity theme and activity keywords used by the SMS copy generation device when generating SMS copy last time, and use the determined activity theme, activity keywords and customer group input from the outside as generation demand information.

[0056] Step S20, inputting the generated demand information into a preset large language model for prediction, and obtaining a plurality of SMS texts, wherein the large language model includes text preference strategies of different customer groups, and adjusting a plurality of first activity texts generated according to the activity theme and activity keywords according to the text preference strategies corresponding to the target customer groups, and obtaining a first SMS text corresponding to each first activity text;

[0057] Optionally, the large language model is a model that can be used to perform natural language processing tasks, such as LLaMa2 13B, LLaMa2 7B, OPT 7B, etc. Optionally, the SMS copy generation device can be a device with computing power that can support the training and operation of the target large language model, and can be deployed in enterprises such as banks and insurance companies.

[0058] Optionally, in the large language model, each customer group corresponds to a text preference strategy. The text preference strategy may be a method for indicating the customer group's preference for the form of expression of the text elements in the generated SMS text. The text elements may include at least one of sentence polishing, emotional style, personal title, text description, specific rights, action encouragement, and presentation location. Optionally, sentence polishing may be to enhance the effect of the SMS text by using expressive language. Emotional style may be the emotional color conveyed by the SMS text, including motivation, seriousness, etc. Personal title may be a personalized language, such as using the title of "you". Text description may be a clear and engaging text description that can help customers better understand the characteristics of the product or service. Specific rights may be to clearly list the benefits or advantages that the product or service brings to customers. Action encouragement may be a statement or button that encourages customers to take the next step. The presentation location may be the layout and position of the SMS text in the page.

[0059] Optionally, after receiving the generation demand information, the SMS copy generation device can pre-process the generation demand information, such as converting the format of the generation demand information into a format that can be recognized by the large language model, and then inputting it into the large language model for prediction to obtain multiple SMS copies corresponding to the group type of the customer group.

[0060] In the large language model, after receiving the generation requirement information, a context environment will be created based on the activity theme and activity keywords. This context environment contains all relevant information about the activity theme to help the large language model better understand what kind of copy should be generated. The large language model will create sentences that conform to the context according to the context environment. Optionally, each activity keyword can correspond to at least one sentence. The large language model can combine at least one sentence corresponding to each activity keyword to generate multiple texts and use it as the first activity text.

[0061] Optionally, in the large language model, the target customer group can be matched with each customer group stored in the large language model to obtain a customer group that matches the target customer group, and determine the text preference strategy corresponding to the matched customer group, which is the text preference strategy corresponding to the target customer group. Then, each first activity text is adjusted according to the text preference strategy to obtain the SMS copy corresponding to each first activity text. For example, if the text preference strategy corresponding to the target customer group includes a personalized title, and the title is "you", then the titles used to refer to customers in the first activity text can be adjusted to "you".

[0062] Step S30 , selecting a target SMS copy from a plurality of SMS copies and sending it to a target device, wherein the target device is a device used by a customer in the target customer group.

[0063] Optionally, after acquiring multiple SMS texts output by the large language model, the SMS text generation device can select one SMS text from each SMS text as the target SMS text. The target SMS text can be sent to the target device used by each customer in the target customer group in the form of a marketing SMS. The customer can view the marketing SMS on his device, and when the customer is interested in the content of the marketing SMS, he can participate in the marketing activity of the target SMS text by clicking on the target SMS text.

[0064] Further, in a feasible embodiment, in step S30, the step of selecting a target SMS text from a plurality of SMS texts and sending the same to a target device includes steps a10-a20.

[0065] Step a10, sorting the multiple SMS texts according to a preset sorting rule, and selecting a SMS text from the sorted multiple SMS texts as a target SMS text;

[0066] Step a20, sending the target SMS text to the target device via a preset sending channel, wherein the sending channel includes at least one of SMS, email, and social media.

[0067] Optionally, the preset arrangement rule may be an arrangement rule set in advance according to user needs, for example, the text messages may be arranged in order from small to large according to the number of characters in each text message. Optionally, after obtaining multiple text messages, the text message generation device may sort the multiple text messages according to the pre-set sorting rule to obtain a sorted text message sequence. The text message sequence includes the sorted multiple text messages.

[0068] A text message can be selected from the text message sequence as the target text message. Optionally, when selecting, the text message can be selected according to a selection instruction input by the user. The selection can also be made according to a pre-set rule, such as selecting the first text message in the text message sequence as the target text message.

[0069] Optionally, after determining the target SMS copy, the target SMS copy can be sent to the target devices used by each customer in the target customer group through a predetermined sending channel, such as sending in the form of SMS, sending in the form of email, sending in the form of social media, etc. This can ensure that the relevant information of the marketing activity can be widely disseminated.

[0070] In this embodiment, by obtaining the generation demand information including at least one of the activity theme, target customer group and activity keywords input from the outside, and inputting the generation demand information into the large language model for prediction, and the large language model contains the text preference strategies of different customer groups, and according to the text preference strategies corresponding to the target customer group, the multiple first activity texts generated according to the activity theme and the activity keywords are adjusted to obtain multiple SMS copy. And the target SMS copy is selected from the multiple SMS copy to send to the target device used by the customers in the target customer group. It can be realized that the target SMS copy that meets the target customer group is automatically generated by the large language model, which significantly improves the efficiency of SMS copy generation and reduces labor costs compared to the manual writing method. And when the target SMS copy is generated by using the large language model, the SMS copy is generated according to the text preference strategy of the target customer group, and then one of the generated multiple SMS copy is selected as the target SMS copy and sent to the target device used by the customers in the target customer group, so that the generated SMS copy can meet the personalized needs of the target customer group, that is, the SMS copy that meets the personalized needs of customers is quickly generated, and the marketing effect is improved.

[0071] Based on the first embodiment of the present application, in the second embodiment of the present application, the same or similar contents as those in the first embodiment can be referred to the above introduction, and will not be described in detail later. Figure 2 , a method for generating SMS text includes steps S100-S300.

[0072] Step S100, obtaining a training data set including at least one task data, wherein the task data includes a training SMS copy and label information, and the label information includes a training activity theme corresponding to the training SMS copy, a training activity keyword, and a training customer group corresponding to the training SMS copy;

[0073] Optionally, during the model training phase, the SMS copy generation device may perform model training on a large language model to be trained according to a pre-set training data set.

[0074] Optionally, the SMS copywriting device can collect and organize historical marketing SMS copywriting and related data in advance, and perform cleaning, labeling and classification processing to generate a training data set containing at least one task data. For example, if the content of a historical marketing SMS copywriting is: xx, don’t recharge your phone bill at the original price anymore, your digital virtual currency phone bill recharge discount is waiting to be used, 10-30 yuan off for purchases over 30, more than 13,000 people have participated, click t.cn / a / abc to enjoy it on mobile banking. The historical marketing SMS copywriting can be labeled to obtain activity keywords: digital virtual currency recharge discount, 10-30 yuan off for purchases over 30, etc. The theme of the activity can be a phone bill recharge discount activity. Optionally, the data related to the historical marketing SMS copywriting can include the customer groups targeted by the historical marketing SMS copywriting.

[0075] Optionally, you can collect various historical marketing SMS copywriting and use it as training SMS copywriting, then use the activity theme of the historical marketing SMS copywriting as the training activity theme of the corresponding training SMS copywriting, use the activity keywords of the historical marketing SMS copywriting as the training activity keywords of the corresponding training SMS copywriting, and use the customer groups of the historical marketing SMS copywriting as the training customer groups of the corresponding training SMS copywriting.

[0076] Optionally, task data is constructed according to the training SMS copy and the corresponding training activity theme, training keywords and training customer groups, each task data corresponds to a training SMS copy, and each training SMS copy is aggregated to obtain a training data set.

[0077] Step S200, inputting the label information into the large language model to be trained for prediction, and obtaining a plurality of predicted SMS texts, wherein the large language model to be trained adjusts a plurality of second activity texts generated according to the training activity theme and the training activity keywords according to the text preference strategy corresponding to the training group type, and obtains a predicted SMS text corresponding to each second activity text;

[0078] Optionally, for any set of task data, the label information in the task data can be input into the large language model to be trained. In the large language model, a context environment will be created based on the training activity theme, and multiple texts will be generated based on the training activity keywords in the context environment and used as the second activity text.

[0079] Optionally, in the large language model, the training customer group can be matched with each customer group stored in the large language model to obtain a customer group that matches the training customer group, and the text preference strategy corresponding to the matched customer group is determined as the text preference strategy corresponding to the training customer group. Then, each second activity text is adjusted according to the text preference strategy to obtain the predicted SMS copy corresponding to each second activity text. For example, if the text preference strategy corresponding to the training customer group includes a personalized title, and the title is "you", then the titles used to refer to customers in the first activity text can be adjusted to "you".

[0080] Step S300, calculating a total loss function value based on the training SMS text and multiple predicted SMS texts, optimizing the large language model to be trained based on the total loss function value until a preset training end condition is met, thereby obtaining a preset large language model.

[0081] Optionally, the preset training end condition may be a condition that needs to be satisfied for the end of the pre-training, and may be set as needed, and is not limited in this embodiment. For example, in a specific implementation, it may be set to the convergence of the loss function of the large language model, or to the number of training rounds reaching a set number of rounds, or to the duration of training reaching a set duration, etc.

[0082] Optionally, the total loss function corresponding to the large language model can be determined, and the training SMS text and each predicted SMS text are input into the total loss function for calculation to obtain the total loss function value, and then the large language model is reversely parameter updated and optimized according to the total loss function value until the preset training end conditions are met to obtain a trained large language model, that is, the large language model preset in the first embodiment.

[0083] In this embodiment, a training data set including at least one task data is obtained, and the label information in the task data is input into the large language model to be trained, multiple predicted SMS texts are predicted, and the total loss function value is calculated according to the training SMS texts and the multiple predicted SMS texts, so as to optimize the large language model to be trained until the preset training end condition is met, and the preset large language model is obtained. It can be achieved that the trained large language model can generate SMS texts that meet the needs of different customer groups.

[0084] Further, in a feasible embodiment, in step S200, before the step of inputting the label information into the large language model to be trained for prediction, steps b10-b30 are included.

[0085] Step b10, for any customer group, determine each training SMS copy belonging to the customer group in the training data set, and determine the copy preference information according to the preset copy preference model and each training SMS copy belonging to the customer group;

[0086] Step b20, constructing a copywriting preference strategy for the customer group based on the copywriting preference information;

[0087] Step b30, updating the copywriting preference strategy of each customer group to the large language model to be trained.

[0088] Optionally, the copy preference model can be a large model of user-generated copy preferences, such as the ChatGPT large model.

[0089] Optionally, each training SMS copy in the training data set may be classified to obtain training SMS copies sent to different customer groups. Optionally, the training SMS copy corresponding to each customer group is used as the first copy.

[0090] Optionally, for each customer group, the first copy belonging to the customer group can be determined, and after being sent to the devices of customers in the customer group, the click rate of customers clicking on the first copy can be determined. All click rates are summarized to obtain a total click rate. The first copy and the total click rate corresponding to the first copy are input into the copy preference strategy model, and whether the customer group prefers the first copy is determined based on the output result of the copy preference strategy model. If the customer group prefers the first copy, the copy elements contained in the first copy are determined, such as sentence polishing, emotional style, personality title, text description, specific rights, action encouragement and presentation position. At least one of them. And it is used as the copy preference information corresponding to the customer group.

[0091] Optionally, after the text preference information is determined, a text preference strategy corresponding to the customer group may be constructed according to the text preference information, that is, used to indicate the customer group's preference for the presentation form of text elements in the generated SMS text.

[0092] Optionally, the copywriting preference strategies corresponding to each customer group may be updated to the large language model to be trained, so that the large language model includes the copywriting preference strategies corresponding to each customer group.

[0093] In this embodiment, for any customer group, the copy preference information is determined according to the copy preference model and the training SMS copy belonging to the customer group, and a copy preference strategy is constructed according to the copy preference information. The copy preference strategy of each customer group is updated to the large language model to be trained, so that the large language model has the ability to generate SMS copy that meets the personalized needs of different customer groups.

[0094] Further, in a feasible embodiment, the large language model includes an adapter and a model ontology.

[0095] Optionally, in step S300, the step of optimizing the large language model to be trained according to the total loss function value includes step c10.

[0096] Step c10, optimizing the model parameters in the adapter according to the total loss function value.

[0097] Optionally, the large language model includes an adapter and a model body, and the adapter may include an adapter located above the model body and an adapter located below the model body.

[0098] Optionally, when training a large language model, the model parameters of the model body can be frozen so that the model parameters of the model body remain unchanged during the training process. The model parameters in the adapter are reversely optimized according to the loss function value.

[0099] In this embodiment, when training a large language model, the model parameters of the model body are frozen, and the model parameters of the adapter are optimized according to the loss function value, thereby improving the model training efficiency.

[0100] Furthermore, in a feasible embodiment, after the step of selecting a target SMS text from a plurality of first SMS texts and sending the text to a target device, steps d10 to d20 are included.

[0101] Step d10: if feedback information regarding the target SMS copy sent by the target device is received, the training data set is updated according to the feedback information;

[0102] Step d20, optimizing the large language model according to the updated training data set to obtain an optimized large language model.

[0103] Optionally, the SMS copy generation device can receive feedback information sent by the target device, and the feedback information may include click-through rate and conversion rate, etc. The marketing strategy can be adjusted according to the feedback information, and the training data set can be updated according to the adjusted marketing strategy, such as adding new task data, and then the large language model is trained according to the updated training data set to complete the optimization of the large language model until the preset training end conditions are met to obtain the optimized large language model. The SMS copy generation device can generate multiple SMS copies suitable for each customer group based on the optimized large language model.

[0104] In this embodiment, after receiving feedback information, the training data set is updated according to the feedback information, and the large language model is optimized according to the updated training data set to obtain an optimized large language model, so that the SMS copy generated by the subsequent SMS copy generation device using the optimized large language model can better meet the needs of users.

[0105] In addition, to assist in understanding the overall process of the method for generating a text message in this embodiment, an example is given below.

[0106] For example, Figure 3 As shown, it can be divided into three stages. In the model training stage, the copy preference strategies of different customer groups can be determined first, and then updated to the large language model to be trained. Then perform data preparation operations to obtain a training data set containing at least one task data, the task data includes training SMS copy and label information, and the label information includes training activity theme, training activity keywords and training customer groups. When training the model, the large language model to be trained is optimized according to the training data set to obtain a preset large language model. In the model application stage, the generation demand information input from the outside is obtained, and the generation demand information is input into the large language model. The large language model uses the text preference strategy corresponding to the target customer group to adjust the first activity text generated according to the activity theme and activity keywords to obtain the SMS copy, and select the target SMS copy from multiple SMS copy to perform SMS generation operations to obtain marketing SMS, which includes the target SMS copy, and the marketing SMS is sent to the target device used by customers in the target customer group through multi-channel distribution. In the feedback update stage, if effect evaluation feedback, i.e. feedback information, is received from the target device, the marketing activities can be adjusted according to the feedback information, and the training data set can be adjusted according to the feedback information. The large language model is retrained according to the adjusted training data set to obtain an optimized large language model, so that the SMS generation device can perform subsequent SMS copy generation operations according to the optimized large language model.

[0107] Furthermore, the present application provides a device for generating text messages, referring to Figure 4 , the SMS copy generating device comprises:

[0108] The acquisition module A10 is used to acquire generation requirement information input from the outside for generating SMS copy, wherein the generation requirement information includes at least one of the activity theme, target customer group and activity keyword;

[0109] Prediction module A20, used for inputting the generated demand information into a preset large language model for prediction, and obtaining a plurality of SMS texts, wherein the large language model includes text preference strategies of different customer groups, and adjusting a plurality of first activity texts generated according to the activity theme and activity keywords according to the text preference strategies corresponding to the target customer groups, and obtaining SMS texts corresponding to each first activity text;

[0110] The sending module A30 is used to select a target SMS copy from multiple SMS copies and send it to a target device, wherein the target device is a device used by a customer in a target customer group.

[0111] Optionally, the SMS copy generation device further includes a training module (not shown in the figure), the training module being used to:

[0112] Acquire a training data set including at least one task data, wherein the task data includes a training SMS copy and label information, and the label information includes a training activity theme corresponding to the training SMS copy, a training activity keyword, and a training customer group corresponding to the training SMS copy;

[0113] Input the label information into the large language model to be trained for prediction, and obtain multiple predicted SMS texts, wherein the large language model to be trained adjusts multiple second activity texts generated according to the training activity theme and the training activity keywords according to the text preference strategy corresponding to the training customer group, and obtains the predicted SMS text corresponding to each second activity text;

[0114] The total loss function value is calculated based on the training SMS text and multiple predicted SMS texts, and the large language model to be trained is optimized based on the total loss function value until the preset training end condition is met to obtain the preset large language model.

[0115] Optionally, a training module is used to:

[0116] For any customer group, determine each training SMS copy belonging to the customer group in the training data set, and determine the copy preference information based on the preset copy preference model and each training SMS copy belonging to the customer group;

[0117] Build copywriting preference strategies for customer groups based on copywriting preference information;

[0118] Update the copywriting preference strategy for each customer group to the large language model to be trained.

[0119] Optionally, the large language model includes an adapter and a model ontology, a training module, for:

[0120] The model parameters in the adapter are optimized according to the total loss function value.

[0121] Optionally, the sending module A30 is used to:

[0122] If feedback information regarding the target SMS copy is received from the target device, the training data set is updated according to the feedback information;

[0123] The large language model is optimized according to the updated training data set to obtain an optimized large language model.

[0124] Optionally, the sending module A30 is used to:

[0125] Sorting multiple SMS texts according to a preset sorting rule, and selecting a SMS text from the sorted multiple SMS texts as a target SMS text;

[0126] The target SMS text is sent to the target device via a preset sending channel, wherein the sending channel includes at least one of SMS, email, and social media.

[0127] The SMS copy generation device provided by the present application can quickly generate SMS copy that meets the personalized needs of customers and improve marketing effects. Compared with the prior art, the beneficial effects of the SMS copy generation device provided by the present application are the same as the beneficial effects of the SMS copy generation method provided by the above embodiment, which will not be repeated here.

[0128] The present application provides a text message generation device, which includes: at least one processor; and a memory that is communicatively connected to the at least one processor; wherein the memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the text message generation method in the above-mentioned embodiment 1.

[0129] Reference below Figure 5 , which shows a schematic diagram of the structure of a text message generation device suitable for implementing the embodiment of the present application. The text message generation device in the embodiment of the present application may include but is not limited to mobile terminals such as mobile phones, laptop computers, digital broadcast receivers, PDAs (Personal Digital Assistants), PADs (Portable Application Descriptions), PMPs (Portable Media Players), vehicle-mounted terminals (such as vehicle-mounted navigation terminals), etc., and fixed terminals such as digital TVs, desktop computers, etc. The text message generation device shown in the figure is only an example and should not bring any limitations to the functions and scope of use of the embodiments of the present application.

[0130] The SMS copywriting generation device may include a processing device 1001 (e.g., a central processing unit, a graphics processor, etc.), which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM: Read Only Memory) 1002 or a program loaded from a storage device 1003 to a random access memory (RAM: Random Access Memory) 1004. In RAM1004, various programs and data required for the operation of the device are also stored. The processing device 1001, ROM1002, and RAM1004 are connected to each other through a bus 1005. An input / output (I / O) interface 1006 is also connected to the bus. Generally, the following systems can be connected to the I / O interface 1006: an input device 1007 including, for example, a touch screen, a touch pad, a keyboard, a mouse, an image sensor, a microphone, an accelerometer, a gyroscope, etc.; an output device 1008 including, for example, a liquid crystal display (LCD: Liquid Crystal Display), a speaker, a vibrator, etc.; a storage device 1003 including, for example, a magnetic tape, a hard disk, etc.; and a communication device 1009. The communication device 1009 can allow the short message copy generation device to communicate wirelessly or wired with other devices to exchange data. Although the short message copy generation device with various systems is shown in the figure, it should be understood that it is not required to implement or have all the systems shown. More or fewer systems can be implemented or have alternatively.

[0131] In particular, according to the embodiments disclosed in the present application, the process described above with reference to the flowchart can be implemented as a computer software program. For example, the embodiments disclosed in the present application include a computer program product, which includes a computer program carried on a computer-readable medium, and the computer program includes a program code for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from a network through a communication device, or installed from a storage device 1003, or installed from a ROM 1002. When the computer program is executed by the processing device 1001, the above-mentioned functions defined in the method of the embodiment disclosed in the present application are executed.

[0132] The SMS copy generation device provided by the present application adopts the SMS copy generation method in the above embodiment, and can quickly generate SMS copy that meets the personalized needs of customers, thereby improving marketing effects. Compared with the prior art, the beneficial effects of the SMS copy generation device provided by the present application are the same as the beneficial effects of the SMS copy generation method provided by the above embodiment, and other technical features in the SMS copy generation device are the same as the features disclosed in the method of the previous embodiment, which will not be repeated here.

[0133] It should be understood that the various parts disclosed in this application can be implemented by hardware, software, firmware or a combination thereof. In the description of the above embodiments, specific features, structures, materials or characteristics can be combined in any one or more embodiments or examples in a suitable manner.

[0134] The above are only specific implementations of the present application, but the protection scope of the present application is not limited thereto. Any technician familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

[0135] The present application provides a medium, which is a computer-readable storage medium having computer-readable program instructions (ie, computer programs) stored thereon, and the computer-readable program instructions are used to execute the SMS text generation method in the above-mentioned embodiment.

[0136] The computer-readable storage medium provided in the present application may be, for example, a USB flash drive, but is not limited to electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, systems or devices, or any combination of the above. More specific examples of computer-readable storage media may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In this embodiment, the computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in combination with an instruction execution system, system or device. The program code contained on the computer-readable storage medium may be transmitted using any appropriate medium, including but not limited to: wires, optical cables, RF (Radio Frequency), etc., or any suitable combination of the above.

[0137] The computer-readable storage medium may be included in the SMS text generation device; or may exist independently without being assembled into the SMS text generation device.

[0138] The computer-readable storage medium carries one or more programs. When the one or more programs are executed by the SMS text generation device, the SMS text generation device can execute the steps of the SMS text generation method.

[0139] Computer program code for performing the operations of the present application may be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, C++, and conventional procedural programming languages ​​such as "C" or similar programming languages. The program code may be executed entirely on the user's computer, partially on the user's computer, as a separate software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In the case of a remote computer, the remote computer may 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 may be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0140] The flow chart and block diagram in the accompanying drawings illustrate the possible architecture, function and operation of the system, method and computer program product according to various embodiments of the present application. In this regard, each square box in the flow chart or block diagram can represent a module, a program segment or a part of a code, and the module, the program segment or a part of the code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the square box can also occur in a sequence different from that marked in the accompanying drawings. For example, two square boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each square box in the block diagram and / or flow chart, and the combination of the square boxes in the block diagram and / or flow chart can be implemented with a dedicated hardware-based system that performs a specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0141] The modules involved in the embodiments described in this application may be implemented by software or hardware, wherein the name of the module does not constitute a limitation on the unit itself in some cases.

[0142] The medium provided in this application is a computer-readable storage medium, which stores computer-readable program instructions (i.e., computer programs) for executing the above-mentioned SMS copy generation method, and can quickly generate SMS copy that meets the personalized needs of customers, thereby improving marketing effects. Compared with the prior art, the beneficial effects of the computer-readable storage medium provided in this application are the same as the beneficial effects of the SMS copy generation method provided in the above-mentioned embodiment, and will not be elaborated here.

[0143] The present application also provides a computer program product, including a computer program, which implements the steps of the above-mentioned method for generating text messages when the computer program is executed by a processor.

[0144] The computer program product provided by this application can quickly generate SMS texts that meet the personalized needs of customers and improve marketing effects. Compared with the prior art, the beneficial effects of the computer program product provided by this application are the same as the beneficial effects of the SMS text generation method provided by the above embodiment, which will not be repeated here.

[0145] The above are only some embodiments of the present application, and are not intended to limit the patent scope of the present application. All equivalent structural changes made using the contents of the present application specification and drawings under the technical concept of the present application, or direct / indirect application in other related technical fields are included in the patent protection scope of the present application.

Claims

1. A method for generating a short message copy, characterized in that: The method comprises: Acquire externally input generation requirement information for generating SMS copy, wherein the generation requirement information includes at least one of an activity theme, a target customer group, and an activity keyword; Inputting the generated demand information into a preset large language model for prediction to obtain a plurality of SMS texts, wherein the large language model includes text preference strategies of different customer groups, and adjusting a plurality of first activity texts generated according to the activity theme and the activity keywords according to the text preference strategies corresponding to the target customer groups to obtain SMS texts corresponding to each first activity text; A target SMS copy is selected from the multiple SMS copies and sent to a target device, wherein the target device is a device used by a customer in the target customer group.

2. The method for generating a short message copy according to claim 1, wherein: The method further comprises: Acquire a training data set including at least one task data, wherein the task data includes a training SMS copy and label information, and the label information includes a training activity theme corresponding to the training SMS copy, a training activity keyword, and a training customer group corresponding to the training SMS copy; Inputting the label information into the large language model to be trained for prediction, and obtaining a plurality of predicted SMS texts, wherein the large language model to be trained adjusts a plurality of second activity texts generated according to the training activity theme and the training activity keywords according to a text preference strategy corresponding to the training customer group, and obtains a predicted SMS text corresponding to each second activity text; A total loss function value is calculated based on the training SMS text and the plurality of predicted SMS texts, and the large language model to be trained is optimized based on the total loss function value until a preset training end condition is met, thereby obtaining a preset large language model.

3. The method for generating a short message copy according to claim 2, wherein: Before the step of inputting the label information into the large language model to be trained for prediction, the method includes: For any customer group, determine each training SMS copy belonging to the customer group in the training data set, and determine the copy preference information according to the preset copy preference model and each training SMS copy belonging to the customer group; Constructing a copywriting preference strategy for the customer group based on the copywriting preference information; The copywriting preference strategy of each of the customer groups is updated to the large language model to be trained.

4. The method for generating a short message copy according to claim 2, wherein: The large language model includes an adapter and a model body, The step of optimizing the large language model to be trained according to the total loss function value comprises: The model parameters in the adapter are optimized according to the total loss function value.

5. The method for generating a short message copy according to claim 2, wherein: After the step of selecting a target SMS text from the plurality of first SMS texts and sending it to the target device, the following steps are included: If feedback information regarding the target SMS copy sent by the target device is received, updating the training data set according to the feedback information; The large language model is optimized according to the updated training data set to obtain an optimized large language model.

6. The method for generating a short message copy according to any one of claims 1 to 5, characterized in that: The step of selecting a target SMS text from the multiple SMS texts and sending it to the target device includes: Sorting the multiple SMS texts according to a preset sorting rule, and selecting a SMS text from the sorted multiple SMS texts as a target SMS text; The target SMS text is sent to the target device via a preset sending channel, wherein the sending channel includes at least one of SMS, email, and social media.

7. A device for generating text messages, characterized in that: The short message copy generating device comprises: An acquisition module, used to acquire generation requirement information input from the outside for generating SMS copy, wherein the generation requirement information includes at least one of an activity theme, a target customer group, and an activity keyword; A prediction module, used for inputting the generated demand information into a preset large language model for prediction, and obtaining a plurality of SMS texts, wherein the large language model includes text preference strategies of different customer groups, and adjusting a plurality of first activity texts generated according to the activity theme and the activity keywords according to the text preference strategies corresponding to the target customer groups, and obtaining SMS texts corresponding to each first activity text; The sending module is used to select a target SMS copy from the multiple SMS copies and send it to a target device, wherein the target device is a device used by customers in the target customer group.

8. A device for generating text messages, characterized in that: The device comprises: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the computer program is configured to implement the steps of the method for generating text for a short message according to any one of claims 1 to 6.

9. A medium, characterized in that The medium is a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the steps of the method for generating a text message as described in any one of claims 1 to 6 are implemented.

10. A computer program product, characterized in that The computer program product comprises a computer program, and when the computer program is executed by a processor, the steps of the method for generating text for a short message according to any one of claims 1 to 6 are implemented.

Images