Copywriting generation method, device, electronic device and storage medium

Abstract

The present disclosure provides a copywriting generation method, device, electronic device and storage medium, which relate to the field of artificial intelligence technology, especially the field of deep learning and the field of natural language processing technology, and can be applied to large language models and generative dialogue scenarios. The specific implementation scheme is: in response to receiving input copywriting demand information, according to the copywriting generation operation related to the copywriting demand information, the copywriting prompt information in the copywriting demand information is updated to obtain the first target copywriting demand information, wherein the first target copywriting demand information includes the target copywriting prompt information related to the semantic attributes of the copywriting demand information; and the first target copywriting demand information is processed based on the pre-trained deep learning model to generate a first feedback copy corresponding to the copywriting demand information.

Description

Technical Field

[0001] The present disclosure relates to the field of artificial intelligence technology, in particular to the field of deep learning and the field of natural language processing technology, and can be applied to large language models and generative dialogue scenarios. Background Art

[0002] With the rapid development of artificial intelligence technology, users can use large models to complete more complex data processing tasks such as image processing and text generation. Large models can refer to deep learning models with a large number of model parameters, which can be as high as billions. With a large number of model parameters, deep learning models can complete more complex data processing tasks. Summary of the invention

[0003] The present disclosure provides a copywriting generation method, device, electronic device, storage medium and program product.

[0004] According to one aspect of the present disclosure, a copy generation method is provided, comprising: in response to receiving input copy requirement information, updating copy prompt information in the copy requirement information according to a copy generation operation related to the copy requirement information to obtain first target copy requirement information, wherein the first target copy requirement information includes target copy prompt information related to semantic attributes of the copy requirement information; and processing the first target copy requirement information based on a pre-trained deep learning model to generate a first feedback copy corresponding to the copy requirement information.

[0005] According to another aspect of the present disclosure, a copy generation device is provided, including: a first target copy requirement information acquisition module, used to respond to receiving input copy requirement information and, according to a copy generation operation related to the copy requirement information, update the copy prompt information in the copy requirement information to obtain first target copy requirement information, wherein the first target copy requirement information includes target copy prompt information related to the semantic attributes of the copy requirement information; and a first feedback copy generation module, used to process the first target copy requirement information based on a pre-trained deep learning model to generate a first feedback copy corresponding to the copy requirement information.

[0006] According to another aspect of the present 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, and the instructions are executed by the at least one processor so that the at least one processor can execute the method provided according to an embodiment of the present disclosure.

[0007] According to another aspect of the present disclosure, a non-transitory computer-readable storage medium storing computer instructions is provided, wherein the computer instructions are used to enable the computer to execute the method provided according to an embodiment of the present disclosure.

[0008] According to another aspect of the present disclosure, a computer program product is provided, including a computer program, wherein when the computer program is executed by a processor, the method provided according to the embodiment of the present disclosure is implemented.

[0009] It should be understood that the content described in this section is not intended to identify the key or important features of the embodiments of the present disclosure, nor is it intended to limit the scope of the present disclosure. Other features of the present disclosure will become easily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] The accompanying drawings are used to better understand the present solution and do not constitute a limitation of the present disclosure.

[0011] Figure 1 An exemplary system architecture to which the method and device for generating a text document according to an embodiment of the present disclosure can be applied is schematically shown;

[0012] Figure 2 A flowchart of a method for generating a document according to an embodiment of the present disclosure is schematically shown;

[0013] Figure 3 An example diagram of a copywriting prompt template according to an embodiment of the present disclosure is schematically shown;

[0014] Figure 4 The application scenario diagram of the copywriting generation method according to the embodiment of the present disclosure is schematically shown;

[0015] Figure 5 A diagram schematically shows an application scenario of a method for generating a text according to another embodiment of the present disclosure;

[0016] Figure 6 The following schematically shows an architecture diagram of a large model service platform according to an embodiment of the present disclosure;

[0017] Figure 7 A block diagram schematically shows a text generation device according to an embodiment of the present disclosure; and

[0018] Figure 8 A schematic block diagram of an example electronic device 800 suitable for implementing a method for generating a copy of an embodiment of the present disclosure is shown. DETAILED DESCRIPTION

[0019] The following is a description of exemplary embodiments of the present disclosure in conjunction with the accompanying drawings, including various details of the embodiments of the present disclosure to facilitate understanding, which should be considered as merely exemplary. Therefore, it should be recognized by those of ordinary skill in the art that various changes and modifications may be made to the embodiments described herein without departing from the scope and spirit of the present disclosure. Similarly, for the sake of clarity and conciseness, descriptions of well-known functions and structures are omitted in the following description.

[0020] In the technical solution disclosed in the present invention, the acquisition, storage and application of user personal information involved are in compliance with the provisions of relevant laws and regulations, necessary confidentiality measures are taken, and do not violate public order and good morals.

[0021] With the development of artificial intelligence technology, more and more companies are using big models to improve office efficiency and business development speed. For example, relevant big model service providers can open model application service interfaces to enterprises, so that relevant enterprises can easily complete office work such as text editing, image generation, personalized copywriting, script development, etc. through service interfaces.

[0022] The embodiments of the present disclosure provide a copywriting generation method, device, electronic device, storage medium and program product. The copywriting generation method includes: in response to receiving input copywriting requirement information, updating the copywriting prompt information in the copywriting requirement information according to the copywriting generation operation related to the copywriting requirement information, and obtaining the first target copywriting requirement information, wherein the first target copywriting requirement information includes the target copywriting prompt information related to the semantic attribute of the copywriting requirement information; and processing the first target copywriting requirement information based on the pre-trained deep learning model to generate a first feedback copy corresponding to the copywriting requirement information.

[0023] According to an embodiment of the present disclosure, by updating the copy requirement information input by the user according to the copy generation operation, the first target copy requirement information obtained can include target copy prompt information that can indicate the semantic attributes in the copy requirement information, so that the first target copy requirement information is input into the deep learning model, which can enable the deep learning model to understand the semantic attributes of the copy requirements more accurately, and then the first feedback copy obtained can more accurately meet the user's copy requirements, improve the accuracy of copy generation, and achieve improved office efficiency.

[0024] Figure 1 An exemplary system architecture to which the method and apparatus for generating texts according to an embodiment of the present disclosure can be applied is schematically shown.

[0025] It should be noted that Figure 1The examples shown are only examples of system architectures to which the embodiments of the present disclosure can be applied, in order to help those skilled in the art understand the technical content of the present disclosure, but do not mean that the embodiments of the present disclosure cannot be used in other devices, systems, environments or scenarios. For example, in another embodiment, the exemplary system architecture to which the copywriting generation method and apparatus can be applied may include a terminal device, but the terminal device may implement the copywriting generation method and apparatus provided by the embodiments of the present disclosure without interacting with the server.

[0026] like Figure 1 As shown, the system architecture 100 according to this embodiment may include terminal devices 101, 102, 103, a network 104 and a server 105. The network 104 is used to provide a medium for communication links between the terminal devices 101, 102, 103 and the server 105. The network 104 may include various connection types, such as wired and / or wireless communication links, etc.

[0027] Users can use terminal devices 101, 102, 103 to interact with server 105 through network 104 to receive or send messages, etc. Various communication client applications can be installed on terminal devices 101, 102, 103, such as knowledge reading applications, web browser applications, search applications, instant messaging tools, email clients and / or social platform software, etc. (only as examples).

[0028] The terminal devices 101 , 102 , and 103 may be various electronic devices having a display screen and supporting web browsing, including but not limited to smart phones, tablet computers, laptop computers, and desktop computers.

[0029] The server 105 may be a server that provides various services, such as a background management server (only an example) that provides support for the content browsed by the user using the terminal devices 101, 102, and 103. The background management server may analyze and process the received data such as user requests, and feed back the processing results (such as web pages, information, or data obtained or generated according to user requests) to the terminal device.

[0030] Server 105 can also be a cloud server, also known as a cloud computing server or cloud host, which is a host product in the cloud computing service system to solve the defects of difficult management and weak business scalability in traditional physical hosts and VPS services ("Virtual Private Server", or "VPS" for short). Server 105 can also be a server of a distributed system, or a server combined with blockchain.

[0031] It should be noted that the text generation method provided in the embodiment of the present disclosure can generally be executed by the server 105. Accordingly, the text generation device provided in the embodiment of the present disclosure can generally be set in the server 105. The text generation method provided in the embodiment of the present disclosure can also be executed by a server or server cluster that is different from the server 105 and can communicate with the terminal devices 101, 102, 103 and / or the server 105. Accordingly, the text generation device provided in the embodiment of the present disclosure can also be set in a server or server cluster that is different from the server 105 and can communicate with the terminal devices 101, 102, 103 and / or the server 105.

[0032] It should be understood that Figure 1 The number of terminal devices, networks and servers in the embodiment is only for illustration. Any number of terminal devices, networks and servers may be provided according to implementation requirements.

[0033] Figure 2 The flowchart of the copywriting generation method according to the embodiment of the present disclosure is schematically shown.

[0034] like Figure 2 As shown, the text generation method includes operations S210 to S220.

[0035] In operation S210, in response to receiving input copy requirement information, the copy prompt information in the copy requirement information is updated according to a copy generation operation related to the copy requirement information to obtain first target copy requirement information, wherein the first target copy requirement information includes target copy prompt information related to the semantic attributes of the copy requirement information.

[0036] In operation S220, the first target copy requirement information is processed based on the pre-trained deep learning model to generate a first feedback copy corresponding to the copy requirement information.

[0037] According to the embodiments of the present disclosure, the copy generation operation may include an operation for instructing to generate a copy according to copy requirement information. The embodiments of the present disclosure do not limit the specific operation type of the copy generation operation, and those skilled in the art may select it according to actual needs.

[0038] According to the embodiments of the present disclosure, the copy requirement information may be the requirements related to the copy that needs to be generated by the target object, such as the format requirements and content requirements of the copy that needs to be generated. The copy prompt information may be the information related to the semantic attribute input by the target object. For example, the copy prompt information may be "Please query the improperly spelled words in the text", and for another example, the copy prompt information may also be the entire copy requirement information. The embodiments of the present disclosure do not limit the specific type of the copy prompt information.

[0039] According to the embodiments of the present disclosure, the target copy prompt information can be obtained by updating the copy prompt information in the copy requirement information. For example, the copy prompt information in the copy requirement information can be updated based on a deep learning algorithm, but it is not limited thereto. The copy prompt information in the copy requirement information can also be updated based on other methods. For example, the copy prompt information can be updated based on a configuration template to obtain the target copy prompt information. The embodiments of the present disclosure do not limit the specific method for updating the copy prompt information in the copy requirement information, and those skilled in the art can select it according to actual needs.

[0040] According to the embodiments of the present disclosure, the target copy prompt information can accurately characterize the semantic attributes of the copy demand information, so that the first target copy demand information can be processed according to the pre-trained deep learning model, so that the deep learning model can generate a first feedback copy while fully understanding the semantic attributes of the copy demand information, thereby improving the degree of matching between the first feedback copy and the copy demand of the target object, avoiding the situation where the generated feedback copy has low copy quality due to the deep learning model's erroneous understanding of the semantics of the copy demand information, improving the overall quality level of the automatically generated copy, and thereby improving the office efficiency of the target object.

[0041] According to the embodiments of the present disclosure, the pre-trained deep learning model can be constructed based on any type of deep learning algorithm. The embodiments of the present disclosure do not limit the specific model parameters of the deep learning model, and those skilled in the art can make choices based on actual needs.

[0042] According to an embodiment of the present disclosure, the text type of the first feedback text may include at least one of the following: a speech text type, an information collection type, a reply information type, and a text rewriting type.

[0043] According to the embodiments of the present disclosure, the first feedback copy of the speech copy type may include a speech draft, a product explanation draft, and other copies that require the target object to express through spoken language. According to the copy generation method provided by the embodiments of the present disclosure, the colloquial attribute of the first feedback copy of the speech type can be improved, and at the same time, the colloquial attribute of the first feedback copy can be matched with the speech scene according to the specific speech scene of the first feedback copy represented by the target copy prompt information, thereby improving the quality of the first feedback copy.

[0044] According to the embodiments of the present disclosure, the first feedback copy of the information collection type may include a copy obtained after collecting any type of information, such as business information collection, image information collection, etc. According to the copy generation method provided by the embodiments of the present disclosure, the degree of matching between the information collected in the first feedback copy and the needs of the target object can be improved according to the semantic attributes represented by the target copy prompt information, thereby avoiding the omission of important information, reducing the amount of redundant information collected, and improving the quality of the first feedback copy.

[0045] According to an embodiment of the present disclosure, the first feedback text of the reply information type may include question information or answer information in a question-and-answer scenario. The first feedback text of the reply information type may be, for example, intelligent response information for a target object in an e-commerce sales scenario.

[0046] According to the embodiments of the present disclosure, the first feedback copy of the text rewriting type may include a copy obtained by rewriting any type of text, such as news text, notification text, novel, etc. According to the copy generation method provided by the embodiments of the present disclosure, the text requirement information can be rewritten in a manner that matches the semantic attributes according to the semantic attributes represented by the target copy prompt information, such as the style attributes and format attributes of the text rewriting, so as to improve the matching degree between the first feedback copy and the rewriting requirements of the target object and improve the quality of the first feedback copy.

[0047] It should be noted that the text type of the first feedback text may also include other types, such as code script editing type, comment content type, etc. The embodiment of the present disclosure does not limit the specific text type of the first feedback text, and those skilled in the art may choose according to actual needs.

[0048] According to an embodiment of the present disclosure, the text requirement information is associated with the associated requirement information, and the associated prompt information corresponding to the associated requirement information matches the text prompt information.

[0049] It should be noted that the associated demand information may be information that establishes an association relationship with the copywriting demand information, for example, it may include the copywriting demand information input by the target object in a historical time period, or it may also be the copywriting demand information input by other objects. The embodiments of the present disclosure limit the specific method of obtaining the associated demand information, and those skilled in the art may make a choice based on actual needs.

[0050] According to an embodiment of the present disclosure, updating the copy prompt information in the copy requirement information according to the copy generation operation related to the copy requirement information to obtain the first target copy requirement information may include: when the copy generation operation is of the first operation type, updating the copy prompt information according to the associated requirement information and the copy requirement information to obtain the first target copy requirement information.

[0051] According to an embodiment of the present disclosure, the first operation type may indicate taking the associated demand information and the copy demand information as the whole of the copy demand of the target object, and updating the copy prompt information by processing the associated demand information and the copy demand information. The semantic attributes of the target object's copy generation requirements may be further accurately understood through the associated demand information and the copy demand information, thereby improving the representation accuracy of the semantic attributes of the obtained first target copy demand information, and further improving the copy quality of the first feedback copy.

[0052] In some embodiments of the present disclosure, a prompt information optimizer constructed based on a neural network algorithm can be used to process the associated demand information and the copy demand information to obtain the first target copy demand information. The number of model parameters of the prompt information optimizer can be less than the number of model parameters of the deep learning model.

[0053] According to an embodiment of the present disclosure, updating the copy prompt information based on the associated demand information and the copy demand information may include: extracting features from the associated demand information and the copy demand information respectively to obtain associated demand features and copy demand features; fusing the associated demand features and the copy demand features based on the attention mechanism to obtain fused demand features; and updating the copy prompt information based on the fused demand features.

[0054] According to the embodiments of the present disclosure, feature extraction can be performed on the associated demand information and the copy demand information based on the encoder, thereby obtaining the associated demand features and the copy demand features. The associated demand features and the copy demand features can be fused based on an attention network algorithm, such as a Transformer algorithm, so that the obtained fused demand features can fully fuse the semantic features between the associated demand information and the copy demand information, and enhance the representation capability of the fused demand features for the semantic attributes of the target object.

[0055] According to an embodiment of the present disclosure, updating the text prompt information according to the fused demand feature may be inputting the fused demand feature into a fully connected network layer and outputting the target text prompt information.

[0056] According to an embodiment of the present disclosure, the associated demand information includes historical demand information associated with the target object, and the copy demand information is generated based on a demand information input operation corresponding to the target object.

[0057] In some embodiments of the present disclosure, the associated demand information may include previous demand information of the text demand information, such as previous text related to the current text demand information input by the target object in a historical time period.

[0058] According to an embodiment of the present disclosure, the copy generation method may further include: determining, according to the copy requirement information, a recommended copy prompt template that matches the semantic attribute of the copy requirement information from a preset copy prompt template library.

[0059] According to an embodiment of the present disclosure, the copywriting requirement information can be processed by a neural network algorithm to obtain a classification result for the copywriting requirement information, and a recommended copywriting prompt template can be determined from a preset copywriting prompt template library based on the classification result.

[0060] According to an embodiment of the present disclosure, the preset copy prompt template library may include a pre-set copy prompt template, and the copy prompt template may be determined according to a deep learning model. For example, sample copy demand information including a copy prompt template may be output to a deep learning model, and a sample feedback copy may be output. When the evaluation result of the sample feedback copy meets the evaluation conditions, the copy prompt template may be stored in the copy prompt template library, so that the pre-trained deep learning model can more accurately understand the semantic attributes of the copy prompt template in the copy prompt template library. In this way, by determining a recommended copy prompt template that matches the copy demand information from the copy prompt template library, the target object can further obtain a feedback copy that matches the copy demand information through the recommended copy prompt template.

[0061] It should be understood that the recommended copy prompt template may include target copy prompt information.

[0062] According to an embodiment of the present disclosure, determining, based on the copy requirement information, a recommended copy prompt template that matches the semantic attributes of the copy requirement information from a preset copy prompt template library may include: extracting semantic features of the copy requirement information to obtain copy requirement features; matching copy prompt template features associated with the copy prompt template with the copy requirement features to obtain feature matching results, the copy prompt template being included in the preset copy prompt template library; and determining, based on the feature matching results, a recommended copy prompt template that matches the semantic attributes of the copy requirement information from the preset copy prompt template library.

[0063] According to an embodiment of the present disclosure, the copy prompt template features associated with the copy prompt template are matched with the copy requirement features, which can include calculating the similarity between the copy prompt template features and the copy requirement features to obtain a feature similarity result, and determining the feature matching result based on a comparison result between the feature similarity result and a preset similarity threshold.

[0064] According to an embodiment of the present disclosure, when the feature matching result representation matches, the text prompt template associated with the feature matching result can be determined as a recommended text prompt template that matches the semantic attribute of the text requirement information.

[0065] According to an embodiment of the present disclosure, the copy generation method may further include: generating second target copy requirement information in response to an editing operation on a recommended copy prompt template; and processing the second target copy requirement information according to a deep learning model to generate a second feedback copy.

[0066] According to an embodiment of the present disclosure, the target object can add demand keywords related to the demand attributes to the recommended copy prompt template by performing an editing operation on the recommended copy prompt template. For example, "two" (demand keyword) can be added to the {number} position in the recommended copy prompt template. For another example, "snow scene" (demand keyword) can be added to the {theme} position in the recommended copy prompt template, thereby completing the editing operation and obtaining the second target copy demand information.

[0067] According to an embodiment of the present disclosure, the target object can further generate second target copy requirement information by editing the recommended copy prompt template, thereby satisfying the personalized representation of the copy generation requirement for the target object.

[0068] According to the embodiments of the present disclosure, the pre-trained deep learning model can more accurately understand the semantic attributes in the second target copy demand information, so that the second target copy demand information can be processed by the deep learning model, so that the obtained second feedback copy can further accurately represent the copy generation needs of the target object and improve the quality level of the copy.

[0069] Figure 3 An example diagram of a copywriting prompt template according to an embodiment of the present disclosure is schematically shown.

[0070] like Figure 3 As shown, the first copy prompt template 310 can be related to "image material collection", and the first copy prompt template 310 can include template copy prompt information: "Collect an image album related to {theme}, including image sources and authorization information for image collection", where "{theme}" can be an embedded position corresponding to the required keywords of the theme attributes.

[0071] like Figure 3 As shown, the second copy prompt template 320 may be related to "movie review", and the third copy prompt template 330 may be related to "script rewriting". A preset copy prompt template library may be constructed based on the first copy prompt template 310, the second copy prompt template 320 and the third copy prompt template 330.

[0072] According to an embodiment of the present disclosure, the copywriting generation method may further include: extracting demand keywords related to demand attributes from the copywriting demand information.

[0073] According to the embodiments of the present disclosure, the demand keywords can be extracted based on the keyword extraction method. For example, the text words in the copy demand information can be tagged with parts of speech, and the demand keywords can be extracted from the copy demand information according to the parts of speech of the text words and the grammatical relationship between the text words. However, this is not limited to this. The demand keywords can also be extracted from the copy demand information based on other types of keyword extraction methods. The embodiments of the present disclosure do not limit the specific method of extracting the demand keywords.

[0074] According to an embodiment of the present disclosure, based on the copy generation operation related to the copy requirement information, the copy prompt information in the copy requirement information is updated to obtain the first target copy requirement information, including: when the copy generation operation is the second operation type, determining the embedding position corresponding to the requirement keyword from the recommended copy prompt template; and based on the embedding position, embedding the requirement keyword into the recommended copy prompt template to obtain the first target copy requirement information.

[0075] According to an embodiment of the present disclosure, the text generation operation of the second operation type can indicate that the first target text requirement information is generated by recommending a text prompt template, and then the first feedback text is automatically generated according to the first target text requirement information. A first text generation operation object indicating the first operation type and a second text generation operation object indicating the second operation type can be generated in an interactive interface related to the target object, so that the target object can conveniently select a specific method for obtaining the first feedback text.

[0076] According to an embodiment of the present disclosure, the embedding position in the recommendation copy prompt template can be represented by a prompt symbol, for example, it can be represented by symbols such as "{}", ":". The embedding position corresponding to the demand keyword can be determined by the correspondence between the part-of-speech attribute identifier related to the embedding position and the part-of-speech of the demand keyword, so as to realize automatic embedding of the demand keyword into the recommendation copy prompt template.

[0077] According to the embodiments of the present disclosure, it is also possible to train a neural network model through a copy prompt template, and process the copy demand information based on the neural network model obtained after training, so as to determine the embedding position corresponding to the demand keyword from the recommended copy prompt template according to the neural network model; and based on the embedding position, embed the demand keyword into the recommended copy prompt template to obtain the first target copy demand information.

[0078] It should be noted that the neural network model can be constructed based on the attention network algorithm. For example, the neural network model can be constructed based on the BERT (Bidirectional Encoder Representations from Transformers) algorithm, thereby improving the timeliness of obtaining the first target copy requirement information and reducing the delay in obtaining the feedback copy.

[0079] According to an embodiment of the present disclosure, the copy generation method may further include: updating the custom template library corresponding to the target object according to the recommended copy prompt template to obtain an updated custom template library, wherein the custom template library is suitable for displaying the custom copy prompt template to the target object when the operation authority verification of the target object is passed.

[0080] According to the embodiments of the present disclosure, the target object can be enabled to access and operate the custom template library in a timely manner by setting corresponding operation permissions for the target object. The custom template library can be a text prompt template adapted to the text requirement attributes of the target object. The target object can edit the custom text prompt template in the custom template library, such as filling in the required keywords, to conveniently generate the second target text requirement information for the target object, thereby saving the computational overhead of determining the recommended text prompt template, and facilitating the target object to quickly obtain the second feedback text according to the actual text generation requirements.

[0081] According to the embodiments of the present disclosure, by updating the custom template library corresponding to the target object according to the recommended copy prompt template, the custom template library can be updated in a timely manner according to the copy generation requirements of the target object, so that the target object can quickly obtain the second target copy requirement information according to the latest custom copy prompt template in the custom template library during the subsequent copy generation operation, thereby improving the copy quality of the second feedback copy.

[0082] According to an embodiment of the present disclosure, the copy generation method may further include: in response to a template optimization operation corresponding to a target object, processing a copy prompt template to be optimized corresponding to the template optimization operation according to an attention mechanism algorithm to obtain an optimized custom copy prompt template; and updating a custom template library corresponding to the target object according to the optimized custom copy prompt template to obtain an updated custom template library.

[0083] Figure 4 The application scenario diagram of the copywriting generation method according to the embodiment of the present disclosure is schematically shown.

[0084] like Figure 4As shown, the application scenario may include a client 410 and a server 420. The server 420 may include a prompt information optimizer 421 constructed based on an attention network algorithm and a custom template library 422 associated with the target object.

[0085] The target object can perform the template optimization operation through the client 410, and send the text prompt template to be optimized to the prompt information optimizer 421. The prompt information optimizer 421 can process the text prompt template to be optimized corresponding to the template optimization operation according to the attention mechanism algorithm to obtain the optimized custom text prompt template 4221. The custom text prompt template 4221 can be stored in the custom template library 422 to update the custom template library.

[0086] The target object can access the custom template library 422 through the client 410 , and can call any custom prompt information template from the custom template library 422 .

[0087] According to an embodiment of the present disclosure, the text prompt template to be optimized is generated based on the prompt information editing operation corresponding to the target object.

[0088] According to the copywriting generation method provided by the embodiment of the present disclosure, the prompt information edited by the target object can also be used as the copywriting prompt template to be optimized, so that the prompt information edited by the target object can be input into the prompt information optimizer 421 to obtain an optimized custom copywriting prompt template to meet the personalized copywriting generation needs of the target object.

[0089] According to an embodiment of the present disclosure, the text prompt template to be optimized may also be determined from a custom template library corresponding to the target object.

[0090] The text generation method provided according to the embodiment of the present disclosure may also optimize the custom text prompt template generated in the historical time period to ensure that the optimized custom text prompt template can meet the current text generation needs of the target object.

[0091] Figure 5 The following schematically shows an application scenario diagram of a method for generating text according to another embodiment of the present disclosure.

[0092] like Figure 5 As shown, the application scenario may include a prompt information optimizer 510, a custom template library 520, and a deep learning model 530. The deep learning model 530 may be a natural language processing model.

[0093] When the text requirement information 501 input by the target object includes the requirement keyword and the text generation operation is detected to be the first operation type, the text requirement information 501 can be input to the prompt information optimizer 510, and the prompt information optimizer 510 can update the text prompt information in the text requirement information 501 to obtain the first target text requirement information including the target text prompt information. The first target text requirement information can be input to the deep learning model 530 to output the first feedback text 502.

[0094] like Figure 5 As shown, the target copy prompt information can also be used as an optimized custom copy prompt template related to the target object, and the custom copy prompt template is input into the custom template library 520 related to the target object, so that the target object can call it at any time.

[0095] According to an embodiment of the present disclosure, the copy generation method may further include: responding to a model update request related to the target object, determining a training sample based on an updated custom template library; and training a deep learning model based on the training sample to obtain an updated target deep learning model.

[0096] According to the embodiments of the present disclosure, training samples can be obtained by adding sample keywords to the custom copy prompt template in the custom template library, or by masking the embedded positions corresponding to the sample keywords in the custom copy prompt template. The embodiments of the present disclosure do not limit the specific method of obtaining training samples, and those skilled in the art can make a choice according to actual needs.

[0097] According to the embodiments of the present disclosure, the deep learning model can be trained by a self-supervised training method. For example, the deep learning model can be trained by setting learning tasks for the deep learning model, or the training samples can be annotated to obtain training samples associated with labels, thereby training the updated target deep learning model through a supervised training method. The embodiments of the present disclosure do not limit the specific training method.

[0098] According to the embodiments of the present disclosure, a model training interface can be encapsulated for a client related to a target object, so that the target object can timely train the deep learning model according to a custom template library that is highly relevant to business needs or text generation needs, so as to obtain an updated deep learning model that can accurately generate feedback text in real time. By providing an interface to the client of the target object, the target object can conveniently fine-tune the model parameters of the deep learning model to improve the adaptability of the updated target deep learning model to the business needs or text generation needs of the target object, and improve the matching degree of the feedback file generated by the target deep learning model with the actual needs of the target object.

[0099] According to an embodiment of the present disclosure, training a deep learning model according to training samples to obtain an updated target deep learning model may include: performing an i-th training task on an i-1th candidate deep learning model according to the training samples to obtain an i-th candidate model parameter of an i-th candidate deep learning model; in the case of I=i, obtaining candidate model parameters of I candidate deep learning models, wherein I≥i>1, and the first candidate deep learning model is a deep learning model; determining a target model parameter from 1 candidate model parameter according to an obtained model selection operation related to the target object; and obtaining an updated target deep learning model according to the target model parameter.

[0100] According to an embodiment of the present disclosure, the candidate model parameters of the candidate deep learning model and the sample candidate feedback text output by the candidate deep learning model can be recorded after each training task is completed, so that the target object can perform a model selection operation on the candidate model parameters by viewing the sample candidate feedback text, select the optimal candidate model parameters as the target model parameters, and then obtain the updated target deep learning model according to the target model parameters.

[0101] According to the embodiments of the present disclosure, by retraining a deep learning model with a larger amount of model parameters according to a model update request of a target object, it is possible to adjust the model parameters of the deep learning model according to the business needs of the target object, thereby enabling the target deep learning model to meet the personalized requirements of the target object and improve the quality level of copy generation.

[0102] Figure 6 The schematic diagram shows the architecture of the large model service platform according to the embodiment of the present disclosure.

[0103] like Figure 6As shown, the big model service platform 600 can include a data management module, a model training module, an evaluation and optimization module, a prediction service deployment module and a plug-in service module, a total of five service modules. These five service modules constitute a big model tool chain to provide comprehensive service support such as big model training, model application, and model evaluation for the target object.

[0104] The data management module may include service units with data management functions such as data collection and cleaning, data generation, data import / task labeling, prompt information manager, data reflow, etc. The data management module supports the processing and management of massive large-scale data, and also provides efficient labeling capabilities for data to help target objects build high-quality data, so that high-quality data can be used to better drive model training and improve the application function of the model.

[0105] The model training module can include functional service units that assist model training, such as model pre-training, reinforcement learning mechanism, plug-in generation unit, copywriting prompt template library, and end-to-end training efficiency improvement. The model training module can provide commonly used training service functions such as retraining, fine-tuning, and reinforcement learning, and the target object can train a deep learning model that meets business needs.

[0106] The prediction service deployment module may include service deployment and hosting, copy prompt template optimization, copy prompt template recommendation, basic information storage and other service units. The prediction service deployment module may provide feedback copy for the target object based on the copy generation method provided in the embodiment of the present disclosure. In addition, it may also provide a custom template library update service, so that the target object can improve the quality of the feedback copy through high-quality prompt information templates. At the same time, it may also provide prompt information template arrangement, rewriting, preset and other service functions.

[0107] The evaluation and optimization module can include service function units such as large model evaluation, large model compression, large model security, and large model interpretability. The evaluation and optimization module can provide efficient model evaluation and optimization functions, thereby improving the training efficiency of deep learning models.

[0108] The plug-in service module can include functional service units such as plug-in library, service orchestration, FaaS (Function-as-a-Service) mechanism, online effect tester, etc. Through plug-in library, service orchestration and other methods, the capability boundary of the big model is further expanded, so that related enterprises can flexibly adapt business rules in the process of business innovation and improve business innovation efficiency.

[0109] In addition, the large model service platform 600 can also include artificial intelligence basic service systems such as heterogeneous computing power management systems, high-performance file systems, high-speed network systems, and artificial intelligence scheduling enhancement systems.

[0110] The heterogeneous computing power management system can efficiently manage various types of computing power units, as well as computing power devices of different specifications such as GPU (Graphic Processing Unit), NPU (Neural network Processing Unit), XPU (eXtreme Processing Unit), etc., so that they can be used for different tasks, or can also cooperate to complete different parts of the same job.

[0111] High-performance file systems can significantly improve the efficiency of file reading and writing during large model training and inference.

[0112] High-speed network systems can provide distributed training or reasoning service functions during the training and reasoning of large models. Frequent updates and synchronization of model parameters between different nodes require extremely high network bandwidth. High-speed Internet networks based on the RDMA (Remote Direct Memory Access) protocol can greatly improve the efficiency of distributed training and reasoning.

[0113] The artificial intelligence scheduling enhancement system can provide intelligent task and job scheduling, rationally utilize existing service resources, improve service resource utilization efficiency, increase job throughput, and at the same time reduce service resource costs.

[0114] Figure 7 The block diagram of the text generation device according to the embodiment of the present disclosure is schematically shown.

[0115] like Figure 7 As shown, the text generation device 700 may include: a first target text requirement information obtaining module 710 and a first feedback text generation module 720 .

[0116] The first target copy requirement information obtaining module 710 is used to respond to the received input copy requirement information, update the copy prompt information in the copy requirement information according to the copy generation operation related to the copy requirement information, and obtain the first target copy requirement information, wherein the first target copy requirement information includes the target copy prompt information related to the semantic attributes of the copy requirement information.

[0117] The first feedback copy generating module 720 is used to process the first target copy requirement information based on the pre-trained deep learning model, and generate a first feedback copy corresponding to the copy requirement information.

[0118] According to an embodiment of the present disclosure, the text requirement information is associated with the associated requirement information, and the associated prompt information corresponding to the associated requirement information matches the text prompt information.

[0119] According to an embodiment of the present disclosure, the first target copy requirement information obtaining module includes a copy prompt information updating submodule.

[0120] The copy prompt information updating submodule is used to update the copy prompt information according to the associated requirement information and the copy requirement information to obtain the first target copy requirement information when the copy generation operation is of the first operation type.

[0121] According to an embodiment of the present disclosure, the text prompt information updating submodule includes a feature extraction unit, a fusion requirement feature obtaining unit and a text prompt information updating unit.

[0122] The feature extraction unit is used to extract features from the associated demand information and the copywriting demand information respectively to obtain associated demand features and copywriting demand features.

[0123] The fused demand feature obtaining unit is used to fuse the associated demand features and the copywriting demand features based on the attention mechanism to obtain the fused demand features.

[0124] The copy prompt information updating unit is used to update the copy prompt information according to the fusion requirement characteristics.

[0125] According to an embodiment of the present disclosure, the associated demand information includes historical demand information associated with the target object, and the copy demand information is generated based on a demand information input operation corresponding to the target object.

[0126] According to an embodiment of the present disclosure, the text generation device further includes a recommended text prompt template determination module.

[0127] The recommended copy prompt template determination module is used to determine, according to the copy requirement information, a recommended copy prompt template that matches the semantic attributes of the copy requirement information from a preset copy prompt template library.

[0128] According to an embodiment of the present disclosure, the recommended copy prompt template determination module includes a copy requirement feature acquisition submodule, a feature matching result acquisition submodule and.

[0129] The copywriting requirement feature acquisition submodule is used to extract semantic features from the copywriting requirement information to obtain the copywriting requirement features.

[0130] The feature matching result obtaining submodule is used to match the copy prompt template features associated with the copy prompt template with the copy requirement features to obtain feature matching results. The copy prompt template is included in a preset copy prompt template library.

[0131] The recommended copy prompt template determination submodule is used to determine a recommended copy prompt template that matches the semantic attributes of the copy requirement information from a preset copy prompt template library according to the feature matching result.

[0132] According to an embodiment of the present disclosure, the text generation device further includes: a second target text requirement information generation module and a second feedback text generation module.

[0133] The second target copy requirement information generating module is used to generate the second target copy requirement information in response to the editing operation on the recommended copy prompt template.

[0134] The second feedback copy generation module is used to process the second target copy requirement information according to the deep learning model to generate the second feedback copy.

[0135] According to an embodiment of the present disclosure, the copywriting generation device further includes a demand keyword obtaining module.

[0136] The demand keyword acquisition module is used to extract demand keywords related to demand attributes from the copywriting demand information.

[0137] According to an embodiment of the present disclosure, the first target copy requirement information obtaining module includes: an embedding position determining submodule and a requirement keyword embedding submodule.

[0138] The embedding position determination submodule is used to determine the embedding position corresponding to the required keyword from the recommended copy prompt template when the copy generation operation is the second operation type.

[0139] The demand keyword embedding submodule is used to embed the demand keyword into the recommended copy prompt template based on the embedding position to obtain the first target copy demand information.

[0140] According to an embodiment of the present disclosure, the copywriting generation device further includes a first custom template library updating module.

[0141] The first custom template library update module is used to update the custom template library corresponding to the target object according to the recommended copy prompt template to obtain an updated custom template library, wherein the custom template library is suitable for displaying the custom copy prompt template to the target object when the operation authority verification of the target object is passed.

[0142] According to an embodiment of the present disclosure, the copy generation device further includes a custom copy prompt template optimization module and a second custom template library update module.

[0143] The custom copy prompt template optimization module is used to respond to the template optimization operation corresponding to the target object, process the to-be-optimized copy prompt template corresponding to the template optimization operation according to the attention mechanism algorithm, and obtain the optimized custom copy prompt template.

[0144] The second custom template library updating module is used to update the custom template library corresponding to the target object according to the optimized custom copy prompt template to obtain an updated custom template library.

[0145] According to an embodiment of the present disclosure, the text prompt template to be optimized is generated based on the prompt information editing operation corresponding to the target object.

[0146] According to an embodiment of the present disclosure, the copy prompt template to be optimized is determined from a custom template library corresponding to the target object.

[0147] According to an embodiment of the present disclosure, the text generation device further includes a training sample determination module and a training module.

[0148] The training sample determination module is used to determine the training sample according to the updated custom template library in response to the model update request related to the target object.

[0149] The training module is used to train the deep learning model according to the training samples to obtain an updated target deep learning model.

[0150] According to an embodiment of the present disclosure, the training module includes a training task execution submodule, a candidate model parameter acquisition submodule, a target model parameter determination submodule and a target deep learning model determination submodule.

[0151] The training task execution submodule is used to execute the i-th training task on the i-1-th candidate deep learning model according to the training sample, and obtain the i-th candidate model parameters of the i-th candidate deep learning model.

[0152] The candidate model parameter acquisition submodule is used to obtain the candidate model parameters of each of I candidate deep learning models when I=i, where I≥i>1, and the first candidate deep learning model is the deep learning model.

[0153] The target model parameter determination submodule is used to determine the target model parameters from I candidate model parameters according to the obtained model selection operation related to the target object.

[0154] The target deep learning model determination submodule is used to obtain an updated target deep learning model according to the target model parameters.

[0155] According to an embodiment of the present disclosure, there are speech copy type, information collection type, reply information type, and text rewriting type.

[0156] According to an embodiment of the present disclosure, the present disclosure also provides an electronic device, a readable storage medium and a computer program product.

[0157] According to an embodiment of the present disclosure, an electronic device includes: 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, and the instructions are executed by the at least one processor so that the at least one processor can execute the method as described above.

[0158] According to an embodiment of the present disclosure, a non-transitory computer-readable storage medium storing computer instructions is provided, wherein the computer instructions are used to cause a computer to execute the method as described above.

[0159] According to an embodiment of the present disclosure, a computer program product includes a computer program, and when the computer program is executed by a processor, the computer program implements the method as described above.

[0160] Figure 8 A schematic block diagram of an example electronic device 800 suitable for implementing a copywriting generation method that can be used to implement an embodiment of the present disclosure is shown. 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 can also represent various forms of mobile devices, such as personal digital processing, cellular phones, smart phones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present disclosure described and / or required herein.

[0161] like Figure 8 As shown, the device 800 includes a computing unit 801, which can perform various appropriate actions and processes according to a computer program stored in a read-only memory (ROM) 802 or a computer program loaded from a storage unit 808 into a random access memory (RAM) 803. In the RAM 803, various programs and data required for the operation of the device 800 can also be stored. The computing unit 801, the ROM 802, and the RAM 803 are connected to each other via a bus 804. An input / output (I / O) interface 805 is also connected to the bus 804.

[0162] A number of components in the device 800 are connected to the I / O interface 805, including: an input unit 806, such as a keyboard, a mouse, etc.; an output unit 807, such as various types of displays, speakers, etc.; a storage unit 808, such as a disk, an optical disk, etc.; and a communication unit 809, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 809 allows the device 800 to exchange information / data with other devices through a computer network such as the Internet and / or various telecommunication networks.

[0163] The computing unit 801 may be a variety of general and / or special processing components with processing and computing capabilities. Some examples of the computing unit 801 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various dedicated artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, digital signal processors (DSPs), and any appropriate processors, controllers, microcontrollers, etc. The computing unit 801 performs the various methods and processes described above, such as the copywriting generation method. For example, in some embodiments, the copywriting generation method may be implemented as a computer software program, which is tangibly contained in a machine-readable medium, such as a storage unit 808. In some embodiments, part or all of the computer program may be loaded and / or installed on the device 800 via ROM 802 and / or communication unit 809. When the computer program is loaded into RAM 803 and executed by the computing unit 801, one or more steps of the copywriting generation method described above may be performed. Alternatively, in other embodiments, the computing unit 801 may be configured to perform the copywriting generation method in any other appropriate manner (e.g., by means of firmware).

[0164] Various implementations of the systems and techniques described above 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 chips (SOCs), load programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various implementations can include: being implemented in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which can be a special purpose or general purpose programmable processor that can receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit data and instructions to the storage system, the at least one input device, and the at least one output device.

[0165] The program code for implementing the method of the present disclosure may be written in any combination of one or more programming languages. These program codes may be provided to a processor or controller of a general-purpose computer, a special-purpose computer, or other programmable data processing device, so that the program code, when executed by the processor or controller, enables the functions / operations specified in the flow chart and / or block diagram to be implemented. The program code may be executed entirely on the machine, partially on the machine, partially on the machine and partially on a remote machine as a stand-alone software package, or entirely on a remote machine or server.

[0166] In the context of the present disclosure, a machine-readable medium may be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, device, or equipment. A machine-readable medium may be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium may include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, device, or equipment, or any suitable combination of the foregoing. A more specific example of a machine-readable storage medium may include an electrical connection based on one or more lines, 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 foregoing.

[0167] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user can provide input to the computer. Other types of devices can also be used to provide interaction with the user; for example, the 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 acoustic input, voice input, or tactile input).

[0168] The systems and techniques described herein may be implemented in a computing system that includes back-end components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes front-end components (e.g., a user computer with a graphical user interface or a web browser through which a user can interact with implementations of the systems and techniques described herein), or a computing system that includes any combination of such back-end components, middleware components, or front-end components. The components of the system may be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include: a local area network (LAN), a wide area network (WAN), and the Internet.

[0169] A computer system may include a client and a server. The client and the server are generally remote from each other and typically interact through a communication network. The relationship of client and server arises through computer programs running on respective computers and having a client-server relationship to each other. The server may be a cloud server, a server in a distributed system, or a server combined with a blockchain.

[0170] It should be understood that the various forms of processes shown above can be used to reorder, add or delete steps. For example, the steps recorded in this disclosure can be executed in parallel, sequentially or in different orders, as long as the desired results of the technical solutions disclosed in this disclosure can be achieved, and this document does not limit this.

[0171] The above specific implementations do not constitute a limitation on the protection scope of the present disclosure. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions can be made according to design requirements and other factors. Any modification, equivalent substitution and improvement made within the spirit and principle of the present disclosure shall be included in the protection scope of the present disclosure.

Claims

1. A copywriting generation method, comprising: In response to receiving the input copy requirement information, when the copy generation operation related to the copy requirement information is of the first operation type, feature extraction is performed on the copy requirement information and the associated requirement information associated with the copy requirement information to obtain associated requirement features and copy requirement features, wherein the copy requirement information is generated based on the requirement information input operation corresponding to the target object, the associated requirement information includes information input by other objects, and the associated prompt information corresponding to the associated requirement information matches the copy prompt information; Based on the attention mechanism, the associated demand feature and the copywriting demand feature are fused to obtain a fused demand feature; and updating the text prompt information in the text requirement information according to the fusion requirement feature to obtain first target text requirement information, wherein the first target text requirement information includes target text prompt information representing the semantic attributes of the text requirement information; and The first target copy requirement information is processed based on the pre-trained deep learning model to generate a first feedback copy corresponding to the copy requirement information.

2. The method according to claim 1, wherein: The associated demand information also includes historical demand information associated with the target object.

3. The method according to claim 1, further comprising: According to the text requirement information, a recommended text prompt template matching the semantic attribute of the text requirement information is determined from a preset text prompt template library.

4. The method according to claim 3, wherein determining, based on the text requirement information, a recommended text prompt template that matches the semantic attribute of the text requirement information from a preset text prompt template library comprises: Extracting semantic features from the copywriting requirement information to obtain copywriting requirement features; Matching the copy prompt template feature associated with the copy prompt template with the copy requirement feature to obtain a feature matching result, wherein the copy prompt template is included in the preset copy prompt template library; as well as According to the feature matching result, a recommended copy prompt template matching the semantic attribute of the copy requirement information is determined from the preset copy prompt template library.

5. The method according to claim 4, further comprising: In response to an editing operation on the recommended copy prompt template, generating second target copy requirement information; as well as The second target copy requirement information is processed according to the deep learning model to generate a second feedback copy.

6. The method according to claim 4, further comprising: Extracting demand keywords related to demand attributes from the copywriting demand information; Wherein, updating the copy prompt information in the copy demand information according to the copy generation operation related to the copy demand information to obtain the first target copy demand information includes: In the case where the text generation operation is of the second operation type, determining an embedding position corresponding to the demand keyword from the recommended text prompt template; and Based on the embedding position, the demand keyword is embedded into the recommended copy prompt template to obtain the first target copy demand information.

7. The method according to claim 4, further comprising: The custom template library corresponding to the target object is updated according to the recommended text prompt template to obtain an updated custom template library, wherein the custom template library is suitable for displaying the custom text prompt template to the target object when the operation authority verification of the target object is passed.

8. The method according to claim 1, further comprising: In response to a template optimization operation corresponding to a target object, processing the to-be-optimized copy prompt template corresponding to the template optimization operation according to an attention mechanism algorithm to obtain an optimized custom copy prompt template; as well as According to the optimized custom copy prompt template, the custom template library corresponding to the target object is updated to obtain an updated custom template library.

9. The method according to claim 8, wherein: The to-be-optimized text prompt template is generated based on an editing operation of prompt information corresponding to the target object.

10. The method according to claim 8, wherein: The text prompt template to be optimized is determined from the custom template library corresponding to the target object.

11. The method according to claim 7 or 8, further comprising: In response to a model update request related to the target object, determining a training sample according to the updated custom template library; as well as The deep learning model is trained according to the training samples to obtain an updated target deep learning model.

12. The method according to claim 11, wherein: The step of training the deep learning model according to the training samples to obtain an updated target deep learning model comprises: Performing an i-th training task on the i-1th candidate deep learning model according to the training sample to obtain an i-th candidate model parameter of the i-th candidate deep learning model; In the case of I=i, obtaining candidate model parameters of each of I candidate deep learning models, wherein I≥i>1, and the first candidate deep learning model is the deep learning model; Determining a target model parameter from the I candidate model parameters according to the obtained model selection operation related to the target object; and The updated target deep learning model is obtained according to the target model parameters.

13. The method according to claim 1, wherein: The type of the first feedback copy includes at least one of the following: Speech copy type, information collection type, reply information type, text rewriting type.

14. A copywriting generation device, comprising: a first target copy requirement information obtaining module, configured to, in response to receiving input copy requirement information, update the copy prompt information in the copy requirement information according to a copy generation operation related to the copy requirement information to obtain first target copy requirement information, wherein the first target copy requirement information includes target copy prompt information representing semantic attributes of the copy requirement information; and A first feedback copy generating module, configured to process the first target copy requirement information based on a pre-trained deep learning model, and generate a first feedback copy corresponding to the copy requirement information; The copy requirement information is generated based on the requirement information input operation corresponding to the target object, the copy requirement information is associated with associated requirement information, the associated requirement information includes information input by other objects, and the associated prompt information corresponding to the associated requirement information matches the copy prompt information; Wherein, the first target copy requirement information obtaining module includes: A copy prompt information updating submodule, configured to update the copy prompt information according to the associated requirement information and the copy requirement information to obtain the first target copy requirement information when the copy generation operation is of the first operation type; The text prompt information update submodule includes: A feature extraction unit, used to extract features from the associated demand information and the copy demand information respectively to obtain associated demand features and copy demand features; A fusion demand feature obtaining unit, used for fusing the associated demand feature and the text demand feature based on an attention mechanism to obtain a fusion demand feature; and The copy prompt information updating unit is used to update the copy prompt information according to the fusion requirement characteristics.

15. The device according to claim 14, wherein: The associated demand information includes historical demand information associated with the target object.

16. The apparatus according to claim 14, further comprising: The recommended copy prompt template determination module is used to determine, according to the copy requirement information, a recommended copy prompt template that matches the semantic attributes of the copy requirement information from a preset copy prompt template library.

17. The device according to claim 16, wherein the recommended copy prompt template determination module comprises: A copywriting requirement feature acquisition submodule is used to extract semantic features from the copywriting requirement information to obtain copywriting requirement features; A feature matching result obtaining submodule, for matching the feature of the copy prompt template associated with the copy prompt template with the copy requirement feature to obtain a feature matching result, wherein the copy prompt template is included in the preset copy prompt template library; as well as The recommended text prompt template determination submodule is used to determine, according to the feature matching result, a recommended text prompt template that matches the semantic attribute of the text requirement information from the preset text prompt template library.

18. The apparatus according to claim 17, further comprising: A second target copy requirement information generating module, configured to generate second target copy requirement information in response to an editing operation on the recommended copy prompt template; as well as The second feedback copy generating module is used to process the second target copy requirement information according to the deep learning model to generate a second feedback copy.

19. The apparatus according to claim 17, further comprising: A demand keyword acquisition module is used to extract demand keywords related to demand attributes from the copywriting demand information; Wherein, the first target copy requirement information obtaining module includes: an embedding position determination submodule, configured to determine, when the text generation operation is of the second operation type, an embedding position corresponding to the demand keyword from the recommended text prompt template; and The demand keyword embedding submodule is used to embed the demand keyword into the recommended copy prompt template based on the embedding position to obtain the first target copy demand information.

20. The apparatus of claim 17, further comprising: The first custom template library update module is used to update the custom template library corresponding to the target object according to the recommended text prompt template to obtain an updated custom template library, wherein the custom template library is suitable for displaying the custom text prompt template to the target object when the operation authority verification of the target object is passed.

21. The apparatus of claim 14, further comprising: A custom copy prompt template optimization module, used to respond to a template optimization operation corresponding to a target object, process the to-be-optimized copy prompt template corresponding to the template optimization operation according to an attention mechanism algorithm, and obtain an optimized custom copy prompt template; as well as The second custom template library updating module is used to update the custom template library corresponding to the target object according to the optimized custom copy prompt template to obtain an updated custom template library.

22. The device according to claim 21, wherein The to-be-optimized text prompt template is generated based on an editing operation of prompt information corresponding to the target object.

23. The device according to claim 21, wherein The text prompt template to be optimized is determined from the custom template library corresponding to the target object.

24. The apparatus according to claim 20 or 21, further comprising: A training sample determination module, configured to determine a training sample according to the updated custom template library in response to a model update request related to the target object; as well as A training module is used to train the deep learning model according to the training samples to obtain an updated target deep learning model.

25. The device according to claim 24, wherein: The training module includes: A training task execution submodule, used to execute an i-th training task on the i-1th candidate deep learning model according to the training sample, to obtain an i-th candidate model parameter of the i-th candidate deep learning model; A candidate model parameter obtaining submodule, for obtaining candidate model parameters of each of I candidate deep learning models when I=i, wherein I≥i>1, and the first candidate deep learning model is the deep learning model; a target model parameter determination submodule, configured to determine a target model parameter from the I candidate model parameters according to the obtained model selection operation related to the target object; and The target deep learning model determination submodule is used to obtain the updated target deep learning model according to the target model parameters.

26. The device according to claim 14, wherein: The type of the first feedback copy includes at least one of the following: Speech copy type, information collection type, reply information type, text rewriting type.

27. 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 executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the method according to any one of claims 1 to 13.

28. A non-transitory computer-readable storage medium storing computer instructions, wherein: The computer instructions are used to cause the computer to execute the method according to any one of claims 1 to 13.

29. A computer program product comprising a computer program which, when executed by a processor, implements the method according to any one of claims 1 to 13.

Images