Data generation method and electronic equipment

Abstract

The invention provides a data generation method, which comprises the following steps: in response to obtaining first input data, obtaining target cue word data matched with the first input data from a target database, and processing the first input data into first cue word data by utilizing the target cue word data; adjusting the first cue word data based on the target guide data to obtain second cue word data; and calling the target processing model to generate a target file based on the second cue word data.

Description

Technical Field

[0001] The present disclosure relates to the field of data processing, and in particular to a data generation method and an electronic device. Background Art

[0002] With the continuous development of artificial intelligence technology, the data generation capabilities of large models have gradually matured. For example, based on the prompt words input by users, large models can generate data such as images, text, video, and voice. Among them, the accuracy, richness, and professionalism of the prompt words input by users have a great impact on the quality of data generation. Summary of the invention

[0003] In view of this, the present disclosure provides a data generating method and an electronic device.

[0004] One aspect of the present disclosure provides a data generation method, comprising: in response to obtaining first input data, acquiring target prompt word data matching the first input data from a target database, so as to process the first input data into first prompt word data using the target prompt word data; adjusting the first prompt word data based on target guidance data to obtain second prompt word data; and calling a target processing model to generate a target file based on the second prompt word data.

[0005] According to an embodiment of the present disclosure, obtaining target prompt word data matching the first input data from a target database includes at least one of the following: in response to obtaining the first input data, identifying a first user intention represented by the first input data, and obtaining matching target prompt word data from the first database based on the first user intention; in response to obtaining the first input data, obtaining user portrait information of the target user, and obtaining matching target prompt word data from a second database based on the user portrait information; in response to obtaining the first input data, outputting a first response result for the first input data, and obtaining matching target prompt word data from a third database based on the first response result; in response to obtaining the first input data, identifying the type of intention represented by the first input data, and obtaining target prompt word data matching the first input data from a fourth database based on the intention type.

[0006] According to an embodiment of the present disclosure, matching target prompt word data is obtained from a fourth database corresponding to the intent type based on first input data, including: extracting target feature data from the first input data based on the intent type; retrieving a matching prompt word data set from the fourth database based on the target feature data, and determining evaluation information corresponding to the prompt word data set; and determining target prompt word data from the prompt word data set based on the evaluation information.

[0007] According to an embodiment of the present disclosure, matching target prompt word data is obtained from a second database based on user portrait information, including: obtaining historical operation data of the target user, and processing the historical operation data into structured data; using the structured data to construct a knowledge graph to obtain target prompt word data from a corresponding node of the knowledge graph based on the first input data.

[0008] According to an embodiment of the present disclosure, first input data is processed into first prompt word data using target prompt word data, including at least one of the following: performing prompt engineering processing on the first input data to obtain third prompt word data, embedding the target prompt word data into a corresponding position of the third prompt word data to obtain first prompt word data; performing generation processing on the target prompt word data and the first input data to obtain first prompt word data; outputting the target prompt word data for the first input data in a window interface of a target application, and in response to obtaining a selection operation on the target prompt word data, adding the prompt word data determined by the selection operation to the first input data to obtain the first prompt word data.

[0009] According to an embodiment of the present disclosure, the first prompt word data is adjusted based on the target guidance data to obtain the second prompt word data, including: obtaining the type and / or source of the target guidance data, and updating the first prompt word data according to the corresponding update method based on the type and / or source to obtain the second prompt word data.

[0010] According to an embodiment of the present disclosure, adjusting the first prompt word data based on the target guidance data includes at least one of the following: adjusting the first prompt word data based on the first input data; adjusting the first prompt word data based on the second input data input to the target application; adjusting the first prompt word data based on feedback data of the second response result obtained in response to the first prompt word data.

[0011] According to an embodiment of the present disclosure, the first input data is processed into the first prompt word data using the target prompt word data, and further includes: obtaining the scoring data of each prompt word data under the first dimension to determine the fourth prompt word data under the first dimension; integrating the fourth prompt word data with the first input data to obtain the fifth prompt word data; obtaining the scoring data of each prompt word data under the second dimension to determine the sixth prompt word data under the second dimension; integrating the sixth prompt word data with the first input data to obtain the first prompt word data.

[0012] According to an embodiment of the present disclosure, the method further includes: in response to obtaining a configuration operation on the target file, updating configuration parameters of the target file and generating model parameters of a processing model of the target file.

[0013] The second aspect of the present disclosure also provides an electronic device, including at least one processor and at least one processing model capable of running on the processor, wherein the processing model can be called by a target application to perform at least one of the following: in response to obtaining first input data, obtaining target prompt word data matching the first input data from a target database, so as to process the first input data into first prompt word data using the target prompt word data; adjusting the first prompt word data based on target guide data to obtain second prompt word data; and generating a target file based on the second prompt word data.

[0014] 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

[0015] The above and other objects, features and advantages of the present disclosure will become more apparent through the following description of the embodiments of the present disclosure with reference to the accompanying drawings, in which:

[0016] Figure 1 A diagram schematically shows an application scenario of a data generation method according to an embodiment of the present disclosure;

[0017] Figure 2 A flow chart of a data generation method according to an embodiment of the present disclosure is schematically shown;

[0018] Figure 3 A schematic diagram of an application interface of a target application of a data generation method according to an embodiment of the present disclosure is schematically shown;

[0019] Figure 4 A schematic diagram of an application interface of a target application of a data generation method according to an embodiment of the present disclosure is schematically shown;

[0020] Figure 5 A flowchart of determining target prompt word data of a data generation method according to an embodiment of the present disclosure is schematically shown;

[0021] Figure 6 A schematic diagram of an application interface of a target application of a data generation method according to an embodiment of the present disclosure is schematically shown;

[0022] Figure 7 A schematic diagram of an application interface of a target application of a data generation method according to an embodiment of the present disclosure is schematically shown;

[0023] Figure 8 A schematic diagram of an application interface of a target application of a data generation method according to an embodiment of the present disclosure is schematically shown;

[0024] Fig. 9 A schematic diagram of an application interface of a target application of a data generation method according to an embodiment of the present disclosure is schematically shown;

[0025] Fig.10 A schematic diagram of an application interface of a target application of a data generation method according to an embodiment of the present disclosure is schematically shown;

[0026] Fig.11 Schematically shows a flow chart of determining fifth prompt word data of a data generation method according to an embodiment of the present disclosure;

[0027] Fig.12 A block diagram of an electronic device suitable for implementing a data generating method according to an embodiment of the present disclosure is schematically shown. DETAILED DESCRIPTION

[0028] Hereinafter, embodiments of the present disclosure will be described with reference to the accompanying drawings. Various details of the embodiments of the present disclosure are included to facilitate understanding, and they should be considered as merely exemplary. Therefore, it should be appreciated 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. Likewise, for the sake of clarity and conciseness, descriptions of well-known functions and structures are omitted in the following description.

[0029] In the technical solution of the present disclosure, the collection, storage, use, processing, transmission, provision, disclosure and application of the data involved (including but not limited to user personal information) comply with the provisions of relevant laws and regulations, take necessary confidentiality measures, and do not violate public order and good morals.

[0030] With the continuous development of artificial intelligence technology, the data generation capabilities of large models have gradually matured. For example, based on the prompt words input by users, large models can generate data such as images, text, video, and voice. Among them, the accuracy, richness, and professionalism of the prompt words input by users have a great impact on the quality of data generation.

[0031] In the related art, most of the prompt words entered by users are not sufficiently described, and the input prompt words are too few, resulting in poor quality of data generation. Or the input prompt words are not accurately described, and lack professional prompt words. For example, in the model that generates images using the input prompt words, some professional descriptions of image parameters are missing, resulting in unclear generated images and failing to achieve the desired effect of users.

[0032] Figure 1 An application scenario diagram of a data generation method according to an embodiment of the present disclosure is schematically shown.

[0033] like Figure 1As shown, the application scenario 100 according to this embodiment may include a first terminal device 101, a second terminal device 102, a third terminal device 103, a network 104, and a server 105. The network 104 is used to provide a medium for a communication link between the first terminal device 101, the second terminal device 102, the third terminal device 103, and the server 105. The network 104 may include various connection types, such as wired, wireless communication links, or optical fiber cables, etc.

[0034] The user may use at least one of the first terminal device 101, the second terminal device 102, and the third terminal device 103 to interact with the server 105 through the network 104 to receive or send messages, etc. Various communication systems and client applications may be installed on the first terminal device 101, the second terminal device 102, and the third terminal device 103, such as web browser applications, search applications, instant messaging tools, email clients, social platform software, etc.

[0035] The first terminal device 101, the second terminal device 102, and the third terminal device 103 may be various electronic devices having display screens and supporting web browsing, including but not limited to smart phones, tablet computers, laptop computers, desktop computers, and the like.

[0036] The server 105 may be a server that provides various services, such as a background management server (only as an example) that provides support for websites browsed by users using the first terminal device 101, the second terminal device 102, and the third terminal device 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.

[0037] It should be noted that the data generation method provided in the embodiment of the present disclosure can generally be executed by the server 105. Accordingly, the message processing device provided in the embodiment of the present disclosure can generally be set in the server 105. The data 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 first terminal device 101, the second terminal device 102, the third terminal device 103 and / or the server 105. Accordingly, the message processing 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 first terminal device 101, the second terminal device 102, the third terminal device 103 and / or the server 105.

[0038] 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.

[0039] The following will be based on Figure 1 The scene described by Figure 2 to Figure 12 A data generating method and an electronic device according to a disclosed embodiment are described in detail.

[0040] Figure 2 A flow chart of a data generation method according to an embodiment of the present disclosure is schematically shown. Figure 2 As shown, a data generating method according to an embodiment of the present disclosure includes operations S201 to S203.

[0041] Operation S201: in response to obtaining first input data, acquiring target prompt word data matching the first input data from a target database.

[0042] Among them, the first input data is the data input by the user into the target application, target model or target system, or the data received from other systems, such as: text, voice, uploaded pictures or user demand descriptions input by the user. The target application, target model or target system generates a target file according to the first input data. The target database is used to store data of various modes, such as text information such as historical question and answer records, historical task execution logs, image data, video data or other structured or unstructured data, or the database can also store dimensional feature data that can be used as prompt word data, and the prompt word data includes at least one prompt word. The prompt word data can be used to adjust the first input data and provide reference information for the generation of the target file to optimize the generation effect of the target file. The target database can be a vector database, a text database, an image database, a knowledge graph database, etc. The key features or elements included in the first input data are matched with each data record in the target database, and a data record whose matching degree with the first input data is greater than a preset threshold is obtained, and the words in the data record can be used as target prompt words. The set formed by all target prompt words is the target prompt word data.

[0043] For example, a target database stores multiple reference texts, each of which includes one or more prompt words. The first input data is text data: "Draw a little girl, 8 years old". The first input data can be matched with each reference text for keywords, and the reference texts with a keyword matching degree greater than a preset threshold are determined to be similar texts to the first input data. The prompt words in all similar texts in the target database can be determined as target prompt words, and the set of all target prompt words is used as target prompt word data.

[0044] When the first input data is voice data, the voice data can be converted into text, and then keyword matching is performed with the prompt word data in the target database to obtain the target prompt word data. When the first input data is an image, the features of the image can be identified first, and the key elements can be extracted according to the image features. Then, the prompt words whose relevance to the key elements meets the preset conditions are retrieved from the target database as the target prompt word data. For example: when a user uploads an image of a pet, the type of pet in the image can be identified, and then the prompts describing similar pet types can be matched in the target database as the target prompt word data. When the first input data is data related to the user's demand description, the prompt words related to the demand description can be retrieved in the target database. For example: when the user inputs: "Please design a system interaction interface including a send button", the prompt words related to the interface design and the send button can be searched in the target database as the target prompt word data.

[0045] The first input data can be processed into the first prompt word data using the target prompt word data. Specifically, the prompt words in the target prompt word data can be embedded into the first input data to obtain the first prompt word data, for example, the target prompt word data can be added to the corresponding position of the input text. Certain keywords in the first input data can be replaced with the prompt words in the target prompt word data to obtain the first prompt word data. A user demand description can be generated based on the target prompt word data, and the user demand description can be added to the first input data to obtain the first prompt word data. Alternatively, a data generation model can be used to generate the first prompt word data based on the target prompt word data and the first input data.

[0046] Operation S202: adjusting the first prompt word data based on the target guidance data to obtain second prompt word data.

[0047] The target guidance data may come from multiple sources, and may be the first input data, or the guidance recommendation options generated by the target application or system based on the first prompt word data, or may be the data subsequently input by the user, or feedback information for the generated target file. The feedback information may include the user's selection operation on the first prompt word data or the target prompt word data, or the selection operation on the guidance recommendation options, or the user's input evaluation information on the first prompt word data or the target prompt word data, or the optimization suggestions for the display interface design of the target application, etc. The first prompt word data is adjusted using the target guidance data to obtain the second prompt word data, so that the second prompt word data can better meet the user's generation requirements and improve the accuracy of the target file generated by the target model or the target application according to the second prompt word.

[0048] Specifically, the target guidance data can be used to update the first prompt word data to obtain the second prompt word data, for example: deleting words in the first prompt word data that do not match the description of the first input data, or adding or modifying the first prompt word data based on the second input data subsequently input by the user or the user's feedback information on the generated target file.

[0049] Operation S203: calling the target processing model to generate a target file based on the second prompt word data.

[0050] The target processing model is a model with data generation function, which can generate a target file including the content described by the second prompt word data according to the content described by the second prompt word data. For example: a generation model for data such as text, video, voice, and image. The target processing model can be located on a local device as a functional dependency of a target application or system on the local device, or it can be a cloud processing model stored in an edge device or a cloud device.

[0051] Before calling the target processing model, the intent recognition model can be used to identify the generation intent type of the first input data, and the corresponding type of target processing model can be selected according to the generation intent type. For example: if the user inputs "Please generate a landscape picture", the image generation model is called, or the image generation model of the landscape type is called; if the user inputs "Please write a product recommendation article", the text generation model is called. After determining the target processing model to be called, the second prompt word data is parsed, and the second prompt word data is parsed into a corresponding format and then input into the target processing model. The target processing model generates a target file based on the second prompt word data. After the target file is provided to the user, the user can input feedback information on the target file, and use the feedback information to adjust the model parameters or the second prompt word data to make the target file more in line with user needs.

[0052] For example: The first input data is "Please draw a little girl, 8 years old, reading under a big tree, with a high ponytail."

[0053] In the target database, the target prompt word data matched with the first input data is {girl, woman, 8 years old, 18 years old, big tree, rest, reading, dress, white shirt, two-dimensional style}.

[0054] After using the dimension retrieval model to perform dimension retrieval on the first input data, the dimension extraction model performs dimension extraction on the first input data, and the dimension extraction result is: {girl, 8 years old, big tree, reading}.

[0055] The target prompt word data is embedded into the dimension extraction result of the first input data, and the first prompt word data is obtained as follows: {girl, woman, 8 years old, 18 years old, big tree, rest, reading, dress, white shirt, high ponytail, two-dimensional style}.

[0056] Then, the first input data is used as the target guidance data, and the first prompt word data is adjusted, and the prompt words in the first prompt word data that do not conform to the dimensional features described in the first input data are deleted, or the prompt words in the first prompt word data that already exist in the first input data are deleted, and the second prompt word data is obtained as follows: {girl, 8 years old, big tree, reading, dress, high ponytail, two-dimensional style}. Or, the image generation model is called to generate an image according to the first prompt word data. The user wants to adjust the style of the image, so the feedback information is input: please adjust the style of the image to the movie style, and the feedback information is used as the target guidance data to update the first prompt word data, and the second prompt word data is obtained as follows: {girl, 8 years old, big tree, reading, dress, high ponytail, movie style}. The image generation model is called to generate an image according to the second prompt word data.

[0057] According to a data generation method provided by an embodiment of the present disclosure, in response to obtaining first input data, target prompt word data matching the first input data is obtained from a target database, and the first input data can be expanded to make the first input data richer, further improving the accuracy of target file generation. The first prompt word data is adjusted based on the target guidance data to obtain second prompt word data, which can make the second prompt word data more in line with the user's generation needs, improve the effect of target file generation, and improve the accuracy of the target file generated by the target processing model according to the second prompt word, so that the target file is more in line with user needs.

[0058] based on Figure 2 In the embodiments of the present disclosure shown, in some embodiments, target prompt word data matching first input data is obtained from a target database, including: in response to obtaining the first input data, identifying a first user intention represented by the first input data, and obtaining matching target prompt word data from the first database based on the first user intention; in response to obtaining the first input data, obtaining user portrait information of the target user, and obtaining matching target prompt word data from a second database based on the user portrait information; in response to obtaining the first input data, outputting a first response result for the first input data, and obtaining matching target prompt word data from a third database based on the first response result; in response to obtaining the first input data, identifying the type of intention represented by the first input data, and obtaining target prompt word data matching the first input data from a fourth database based on the intention type.

[0059] In response to obtaining the first input data, a first user intention represented by the first input data is identified, and matching target prompt word data is acquired from a first database based on the first user intention.

[0060] The first user intention is used to characterize the type of the generated target file and the content category of the target file. The type of the target file may include image, text, video, voice, etc.

[0061] For example: When the target file is an image, the content category of the target file can be divided into scenery, people, animals, objects, etc. When the target file is a text, the content category of the target file can be divided into science and technology, prose, news, novels, etc. When the target file is a video, the content category of the target file can be divided into animation, action, documentary, etc. When the target file is a voice, the content category of the target file can be divided into instruction, dialogue, etc. The first database is used to store target prompt word data corresponding to various user intentions. For example: Use the intention recognition model to identify that the first user's intention is to generate images of the person type. According to the first user's intention, prompt word data related to the generation of person-type images is retrieved in the first database as target prompt word data.

[0062] In response to obtaining the first input data, user portrait information of the target user is obtained, and matching target prompt word data is obtained from the second database based on the user portrait information.

[0063] The user profile information of the target user includes the user's historical behavior, browsing history, preference information, and other information that can characterize the user's personalized needs for the target file to be generated. The second database stores the user profile information of all users and the prompt word data related to the user profile information. For example, the second database can be a knowledge graph, which can determine the type of target file that the user needs to generate, as well as the features and elements included in the target file based on the user's behavior habits or preference information. Furthermore, the prompt words related to the type of target file that the user needs to generate, as well as the features and elements included in the target file, are determined as target prompt word data.

[0064] For example: The user inputs: "Please generate a picture of the sunset at the seaside." Using the knowledge graph, based on the user's portrait information, it can be predicted that the user likes natural scenery and warm tones, so the target prompt word data is: {natural, warm, cozy, soft}.

[0065] In response to obtaining the first input data, a first response result for the first input data is output, and matching target prompt word data is acquired from a third database based on the first response result.

[0066] Among them, the third database can be the same as the first database, and can also include a verification database or a standard database corresponding to the type of the generated target file or the first user's intention, for example: the first user's intention is to generate a picture of a person. The third database can store prompt word data of the person type and a target prompt word data template. The target prompt word data template specifies the dimensions and specific prompt words that the target prompt word data should include. For example: the target prompt word data template of the person class is: the person's gender is xx, the scene is xx, the expression is xx, the clothing is xx, the occupation is xx, the action is xx, and the image style is xx. The first response result of the first input data is to call the target processing model and generate the first target file based on the first input data. According to the target prompt word data template and the prompt word data in the third database, the target prompt word data not included in the first response result can be queried in the prompt word data.

[0067] For example, see Figure 3 . Figure 3 FIG. 1 is a schematic diagram of an interface for displaying target prompt word data according to a data generation method of an embodiment of the present disclosure. Figure 3 As shown, on the application interface of the target application, the first input data input by the user is: draw a little girl, and then the target processing model is called according to the first input data, and the first response result of the first input data is generated as picture 1. If picture 1 does not include the scene, image style, expression and other features of the little girl, then the target prompt word data about the scene, image style and expression is retrieved from the third database: {blue sky, two-dimensional, smile, movie light effect}. The target prompt word data is displayed on the application interface as a guide recommendation option.

[0068] In response to obtaining the first input data, the intention type represented by the first input data is identified, and target prompt word data matching the first input data is acquired from a fourth database based on the intention type.

[0069] Among them, the intent type represented by the first input data is used to represent the refined type under the category of the generated target file. The first database is used to store the prompt word data corresponding to each intent type. For example: use the intent recognition model to identify the intent type represented by the first input data. The first input data is: please draw a picture of a teacher teaching a class, then the identified intent type is to generate a picture of a character scene. Based on the identified intent type, the matching target prompt word data is retrieved from the first database.

[0070] For example: Table 1 is a first database corresponding to the intent of the character scene class. The first database stores multiple texts describing character scenes and prompt words under multiple dimensions included in each text. If the first input data is text and the intent type is a picture of the character scene class, then the text similar to the first input data is searched in the first database shown in Table 1, and the prompt word corresponding to the text similar to the first input data is determined as the target prompt word data.

[0071] For example, see Figure 4 . Figure 4 FIG. 1 is a schematic diagram of an interface for displaying target prompt word data according to a data generation method of an embodiment of the present disclosure. Figure 4 As shown, on the application interface of the target application, the first input data input by the user is: draw a little girl. The intent type of the first input data is identified as a portrait, and then the target prompt word data related to the portrait is matched in the target database according to the first input data: {big eyes, high ponytail, dress, two-dimensional, movie light effect, ultra-high definition}. The target prompt word data is displayed on the application interface as a guide recommendation option.

[0072] According to the method provided in the above embodiment, matching target prompt word data can be obtained from multiple target databases based on user intention, user portrait, response results of the target processing model and intention type, so as to optimize the first input data, significantly improve the quality of the generated target file, ensure that the generated target file better meets the actual needs of the user, and significantly improve the generated target file in terms of content richness, relevance and accuracy.

[0073] Table 1

[0074]

[0075] According to the method provided in the above embodiment, matching target prompt word data can be obtained from multiple target databases based on user intention, user portrait, response results of the target processing model and intention type, so as to optimize the first input data, significantly improve the quality of the generated target file, ensure that the generated target file better meets the actual needs of the user, and significantly improve the generated target file in terms of content richness, relevance and accuracy.

[0076] On the basis of the above embodiments, in some embodiments, matching target prompt word data is obtained from a fourth database corresponding to the intent type based on the first input data, including: extracting target feature data from the first input data based on the intent type; retrieving a matching prompt word data set from the fourth database based on the target feature data, and determining evaluation information corresponding to the prompt word data set; and determining the target prompt word data from the prompt word data set based on the evaluation information.

[0077] The target feature data may be a feature vector corresponding to the first input data. According to the intent type, all features related to the intent type are extracted from the first input data, and the first input data is vectorized according to the extracted features to obtain the target feature data.

[0078] The fourth database includes multiple vector databases and multiple prompt word databases, as well as a mapping relationship between the vector database and the prompt word database. Among them, the vector database is used to store feature vectors corresponding to the intent type. The target vector database matching the intent type is queried in the fourth database. The target feature vector whose similarity with the target feature data is greater than the first preset threshold is queried in the target vector database. According to the index of the target feature vector, the corresponding prompt word data is queried in the prompt word database corresponding to the target vector library. All prompt word data constitute a prompt word data set. Among them, the evaluation information corresponding to the prompt word data set can be the scoring information corresponding to each prompt word in the prompt word data set, and the scoring information can be determined based on the user's feedback information and the generation effect of the model for the dimension represented by the prompt word.

[0079] Determine the target prompt word data from the prompt word data set based on the evaluation information. For example: using the dimensional recommendation model, determine the target prompt word data according to the scoring information corresponding to each prompt word data in the prompt word data set. For example: for each prompt word stored in the prompt word data set, determine the score of each prompt word according to the generation effect of the target processing model on the dimension represented by each prompt word. Or determine the score of the prompt word used to generate the historical target file based on the user's feedback information on the generated historical target file. According to the score of each prompt word, divide each prompt word into three types: strongly recommended, not recommended, and custom recommended. Prompt words of the strongly recommended and custom recommended types can be used as target prompt word data.

[0080] For example, see Figure 5 . Figure 5 FIG. 1 is a flow chart of determining target prompt word data according to a data generation method according to an embodiment of the present disclosure. Figure 5 As shown, the first input data is "draw a little girl, 8 years old". The intent type of the first input data is determined to be a portrait by using the intent recognition model. Then, the features of the first input data are extracted according to the intent type to generate a feature vector of the first input data. According to the feature vector of the first input data, in the vector database of the portrait type included in the fourth database, feature vectors with a matching similarity greater than a preset threshold are matched: vector 1, vector 2. According to vector 1 and vector 2, in the prompt word database of the portrait type included in the fourth database, that is, Figure 5In the text database of the character portrait type, the corresponding text 1 and text 2 are searched. The prompt words included in the text 1 and text 2 are put into the prompt word set. Then, each prompt word in the prompt word set performs the following steps: the prompt word is added to the first input data, and the first input data with the prompt word added is input into the scoring system for scoring to obtain the score corresponding to the prompt word. The prompt words with scores greater than the preset threshold are screened out as the target prompt word data. Figure 5 If the scores 2 and 3 in the example are greater than the preset threshold, and the score 1 is less than the preset threshold, the prompt word "big tree" corresponding to the score 2 and the prompt word "rest" corresponding to the score 3 are used as the target prompt word data. Alternatively, the scores are sorted from small to large, and the prompt words corresponding to M scores are taken as the target prompt word data.

[0081] According to the method provided by the above embodiment, the target prompt word data can be determined in combination with the evaluation information of the prompt word data set, and the target prompt word data can be optimized, thereby improving the quality of the generated target file.

[0082] In some embodiments, matching target prompt word data is obtained from a second database based on user portrait information, including: obtaining historical operation data of the target user, and processing the historical operation data into structured data; and constructing a knowledge graph using the structured data.

[0083] The target user's historical operation data includes the user's operation records, web browsing data, generated historical target files, and user-uploaded document libraries, image libraries, or third-party application behavior records. The historical operation data is processed into structured data in a certain format to facilitate computer processing and analysis. For example, a user has repeatedly generated images containing elements such as "landscape", "sunset", and "forest" on an image generation application. These elements are organized into a table that contains fields such as user ID, generation time, and key elements.

[0084] The knowledge graph is constructed using structured data to obtain target prompt word data from a corresponding node of the knowledge graph based on the first input data.

[0085] Among them, the knowledge graph is a graphical data structure used to represent entities and their relationships, and is used to store and query information. For example: map the information in the structured data to the knowledge graph, and create nodes and edges of the knowledge graph. For example: use the user ID, key elements, image style, etc. in the table as nodes, and use the generation time and the number of generated images as edges to represent the user's preferences. For example: create a node for "landscape" and another node for the user name, and connect the two with an edge to represent the user's preference to generate an image containing landscape elements.

[0086] Based on the first input data, the target prompt word data is obtained from the corresponding node of the knowledge graph. For example, the node related to the first input data is searched in the knowledge graph, and other prompt words associated with the related nodes are extracted, and the extracted other prompt words are used as the target prompt word data. For example, the user's first input data is "generate a beautiful landscape painting". According to the first input data, the knowledge graph is queried and it is found that the user prefers images in the styles of "landscape" and "sunset", so "landscape" and "sunset" are used as target prompt word data.

[0087] According to the method of the above embodiment, the target prompt word data is determined according to the historical operation data of the user, and then the target file is generated according to the target prompt word data, so that the generated target file better meets the personalized needs of the user.

[0088] In some embodiments, the first input data is processed into the first prompt word data using the target prompt word data, including at least one of the following: performing prompt engineering processing on the first input data to obtain third prompt word data, embedding the target prompt word data into the corresponding position of the third prompt word data to obtain the first prompt word data; performing generation processing on the target prompt word data and the first input data to obtain the first prompt word data; outputting the target prompt word data for the first input data in the window interface of the target application, and in response to obtaining a selection operation on the target prompt word data, adding the prompt word data determined by the selection operation to the first input data to obtain the first prompt word data.

[0089] The first input data is processed by prompt engineering to obtain third prompt word data, and the target prompt word data is embedded into a corresponding position of the third prompt word data to obtain first prompt word data.

[0090] The prompt engineering processing of the first input data may include: extracting keywords from the first input data, or extracting keywords describing a specific dimension from the first input data, and the extracted keywords are the third prompt word data. For example, the first input data is: "Generate a landscape picture, including blue sky, sea, sunset." Extract the keywords from the first input data, and obtain the third prompt word data as: {landscape, blue sky, sea, sunset}.

[0091] The target prompt word data matching the first input data is obtained in the target database, and then the target prompt word data is embedded into the corresponding position of the third prompt word data to obtain the first prompt word data. For example, the target prompt word data obtained is: {beach, waves, trees, birds}. The target prompt word data is embedded into the third prompt word data to obtain the first prompt word data, for example: {scenery, blue sky, sea, sunset, beach, waves, trees, birds}.

[0092] The target prompt word data and the first input data are generated to obtain the first prompt word data.

[0093] The prompt word generation model or the large language model can be used to generate and process the target prompt word data and the first input data to obtain the first prompt word data. For example, the first input data is: "Draw a little girl, drawing under a big tree." The target prompt word data is: {dress, 8 years old, straw hat, blue sky}. The first prompt word data generated by the prompt word generation model or the large language model is: "Draw a little girl wearing a dress and a straw hat, who is 8 years old, and there is a big tree under the blue sky, and the little girl is drawing under the tree."

[0094] Target prompt word data for the first input data is output in the window interface of the target application, and in response to obtaining a selection operation on the target prompt word data, the prompt word data determined by the selection operation is added to the first input data to obtain the first prompt word data.

[0095] For example, see Figure 6 . Figure 6 FIG. 1 is a schematic diagram of a window interface of a target application of a data generation method according to an embodiment of the present disclosure. Figure 6 As shown, the first input data is: "Draw a little girl, age 8". The target prompt word data matched in the target database is displayed in a way of guiding the recommended options. The target prompt word data is: {big eyes, high ponytail, dress, two-dimensional, movie light effect, ultra-high definition}. The user selects "big eyes". In response to the user's selection operation, big eyes are added to the first input data, and the first prompt word data is obtained as "the picture shows a little girl, age 8, with a pair of big eyes".

[0096] According to the method of the above embodiment, the first input data is optimized by using the target prompt word data through various processing methods, so that the first input data is more accurate and rich, and the accuracy of the generated target file is further improved to make it more in line with user needs.

[0097] In some embodiments, the first prompt word data is adjusted based on the target guidance data to obtain the second prompt word data, including: obtaining the type and / or source of the target guidance data, and updating the first prompt word data according to the corresponding update method based on the type and / or source to obtain the second prompt word data.

[0098] The types of target guidance data include first input data, user feedback information and user subsequent input information. According to the types of target guidance data, it can be determined whether the source of the target guidance data is from the original input data of the user, or from the user feedback data or interaction data.

[0099] For each type and / or source of target guidance data, a corresponding updating method of the first prompt word data may be set, for example, the data may be changed through UI interaction or through a small model, or updated through a driver.

[0100] Exemplarily, the type of target guidance data can be identified and its source can be determined. For example: after obtaining the image generated by the first input data, the user inputs feedback information as the target guidance data: "I hope the sky in the picture is brighter." At this time, it can be identified that the target guidance data comes from the user's feedback. If the first prompt word data is {seaside, sunset, beach, waves,}, the target guidance data is to hope that "the sky is brighter." According to the type of the target guidance data being text and the source being user feedback, it can be determined that the update method is to add a new prompt word, that is, {bright, sky} is added to the first prompt word data, and the second prompt word data is obtained as: {seaside, sunset, beach, waves, bright, sky}.

[0101] In some embodiments, the first prompt word data is adjusted based on the target guidance data, including at least one of the following: adjusting the first prompt word data based on the first input data; adjusting the first prompt word data based on the second input data input to the target application; adjusting the first prompt word data based on feedback data of the second response result obtained in response to the first prompt word data.

[0102] The first prompt word data is adjusted based on the first input data. For example, the first input data can be directly used as the target guidance data, or the dimension-related keywords in the first input data can be extracted as the target guidance data. For example, the first input data is: "Generate a little girl, 8 years old, wearing a dress, reading." The first input data is dimensionally extracted, and keywords describing gender, age, clothing, and action are extracted respectively, and the target guidance data is obtained as: "Girl, 8 years old, dress, reading." The first prompt word data includes: {Girl, 8 years old, dress, reading, 18 years old, jeans, white shirt, rest}. At this time, the target guidance data can be used to adjust the first prompt word data, and the prompt words in the first prompt word data that do not match the age, clothing, and action descriptions in the first input data are deleted, and the second prompt word data is obtained as: {Girl, 8 years old, dress, reading}.

[0103] The first prompt word data is adjusted based on second input data input into the target application.

[0104] The second input data may include a description of the generation requirement or prompt word data that needs to be updated, a selected guidance recommendation option, or optimization suggestions or design suggestions for interface display components input by the user.

[0105] For example, see Figure 7 and Figure 8 . Figure 7 FIG. 1 is a schematic diagram of a window interface of a target application of a data generation method according to an embodiment of the present disclosure. Figure 7 As shown, the first input data is: "Draw a little girl, age 8 years old". The target prompt word data matched in the target database is displayed in the form of guiding recommended options. The target prompt word data is: {big eyes, high ponytail, dress, two-dimensional, movie lighting effects, ultra-high definition}. The user selected "big eyes", and in response to the user's selection operation, big eyes are added to the first input data, and the first prompt word data is obtained as "The picture shows a little girl, age 8, with a pair of big eyes". The target file generated by calling the target processing model based on the first prompt word data is the second response result, and the second response result is Picture 2. At this time, the user enters the second input data in the user input area: "wearing a straw hat, with a smiling expression". Click the send button in the lower left corner of the user input area to send the second input data. Figure 8 FIG. 1 is a schematic diagram of a window interface of a target application of a data generation method according to an embodiment of the present disclosure. Figure 8 As shown, after the second input data is sent out, the keywords describing the specific dimension included in the second input data can be added to the first prompt word, and the second prompt word data is obtained as: "The picture shows a little girl, 8 years old, with big eyes, wearing a straw hat, and smiling." The target processing model generates a target file based on the second prompt word data, namely, Picture 3.

[0106] The first prompt word data is adjusted based on feedback data of a second response result obtained in response to the first prompt word data.

[0107] The feedback data for the second response result may include the user's evaluation information for the second response result, or optimization suggestions for the second response result.

[0108] For example, see Fig. 9 and Fig.10 . Fig. 9 A schematic diagram of a window interface of a target application of a data generation method according to an embodiment of the present disclosure.

[0109] like Fig. 9As shown, the first input data is: "Draw a little girl, age 8 years old". The target prompt word data matched in the target database is displayed in the form of guiding recommended options. The target prompt word data is: {big eyes, high ponytail, dress, two-dimensional, movie lighting effects, ultra-high definition}. The user selected "big eyes", and in response to the user's selection operation, big eyes are added to the first input data, and the first prompt word data is obtained as "The picture shows a little girl, age 8, with a pair of big eyes". The target file generated by calling the target processing model based on the first prompt word data is the second response result, and the second response result is Picture 2. At this time, if the character in Picture 2 is not clear, the user can enter feedback information in the user input area: "The character in the picture is not prominent". Click the send button in the lower left corner of the user input area to send feedback information.

[0110] Fig.10 FIG. 1 is a schematic diagram of a window interface of a target application of a data generation method according to an embodiment of the present disclosure. Fig.10 As shown, after the feedback information is sent out, the prompt word describing the clarity of the person in the picture: the person is prominent, can be added to the first prompt word data according to the feedback information, and the second prompt word data is obtained as: "The picture shows a little girl, 8 years old, with big eyes, and the person is prominent." The target processing model generates a target file according to the second prompt word data, that is, Picture 3.

[0111] According to the method provided in the above embodiment, the first prompt word data is adjusted based on the first input data, ensuring that the first prompt word data is consistent with the original intention of the user. The first prompt word data is adjusted by the second input data input to the target application, so that the generated target file adapts to the change of user needs. The first prompt word data is adjusted according to the feedback data of the second response result obtained for the first prompt word data, ensuring that the generated target file meets the dynamic needs of the user.

[0112] In some embodiments, using target prompt word data to process first input data into first prompt word data also includes: obtaining scoring data of each prompt word data under the first dimension to determine fourth prompt word data under the first dimension; integrating the fourth prompt word data with the first input data to obtain fifth prompt word data; obtaining scoring data of each prompt word data under the second dimension to determine sixth prompt word data under the second dimension; integrating the sixth prompt word data with the first input data to obtain first prompt word data.

[0113] The target prompt word data includes prompt word data under the first dimension and prompt word data under the second dimension. The first dimension and the second dimension are used to characterize different feature dimensions of the target file to be generated. The prompt word data under the first dimension is used to describe the specific features of the target file under the first feature dimension. The prompt word data under the second dimension is used to describe the specific features of the target file under the second feature dimension. Each dimension can correspond to multiple prompt word data.

[0114] For example, a category-dimension mapping table is stored in the target database, as shown in Table 2.

[0115] Table 2

[0116]

[0117] For the image generation task of the character scene category, the first dimension and the second dimension corresponding to the character scene category can be queried in the target database. The first dimension may include image content features such as character posture or action, background and scene elements, and the second dimension may include image parameter features such as lighting and shadow, clarity, artistic style, emotional atmosphere, hue, vision and composition. The prompt word data under the first dimension may include prompt words for describing image content features such as character posture or action, background and scene elements. The prompt word data under the second dimension may include prompt words for describing image parameter features such as lighting and shadow, clarity, artistic style, emotional atmosphere, hue, vision and composition.

[0118] The scoring data of each prompt word data in the target database can be determined according to the generation effect of the target processing model on the specific features described by each prompt word data, or according to the relevance of each prompt word data to the first input data, or according to the feedback information of the user on the target file generated according to the first input data containing the prompt word data in the user's historical generation record. For example, according to the following steps, the picture generated by the prompt word data and the picture-text pair composed of the prompt word data are scored, and the score is used as the scoring data of the prompt word data:

[0119] Generate historical images based on user history generation operations and perform statistics on the categories of historical images;

[0120] According to the prompt word data corresponding to the historical pictures, a mapping relationship between the category of the picture and the dimension of the prompt word data is generated;

[0121] Based on the generation effect of each historical picture, multiple high-quality input data are selected from the historical input data corresponding to each historical picture;

[0122] Combine high-quality input data with corresponding historical images to form image-text pairs, and score each image-text pair;

[0123] From the image-text pairs with scores higher than a preset score threshold, filter out the image-text pairs with poor image generation effects;

[0124] According to the mapping relationship between the image category and the prompt word data dimension, the image category in the image-text pairs with poor image generation effect and the corresponding prompt word data dimension in the historical input data are screened out according to the image-text pairs with poor image generation effect, and a category-dimension blacklist is generated;

[0125] According to the mapping relationship between the image category and the prompt word data dimension, as well as the prompt word data in the historical input data corresponding to the image-text pairs with poor image generation effect, the prompt word data with poor generation effect in each dimension are screened out to generate a category-dimension-vocabulary blacklist; finally, the target prompt word data is filtered according to the category-dimension vocabulary blacklist.

[0126] According to the scoring data of each prompt word data under the first dimension, the fourth prompt word data under the first dimension is determined. For example, the prompt word data under the first dimension can be sorted from large to small according to the scores of each prompt word data, and the first N prompt word data can be screened out as the fourth prompt word data. A scoring threshold can also be set, and the prompt word data with a score greater than the scoring threshold can be used as the fourth prompt word data. The recommendation attribute of each prompt word data can also be determined according to the score of each prompt word data, for example, the recommendation attribute of the prompt word data with a score greater than the first scoring threshold is determined as strongly recommended, the recommendation attribute of the prompt word data with a score less than or equal to the first scoring threshold and greater than the second scoring threshold is determined as optional recommendation, and the recommendation attribute of the prompt word data with a score less than or equal to the first scoring threshold and greater than the second scoring threshold is determined as optional recommendation, as the fourth prompt word data.

[0127] The fourth prompt word data is integrated with the first input data to obtain the fifth prompt word data. For example, the fourth prompt word can be added to the corresponding position of the first input data to obtain the fifth prompt word data, or the fifth prompt word data can be generated according to the fourth prompt word and the first input data.

[0128] Similarly, the scoring data of each prompt word data under the second dimension is obtained to determine the sixth prompt word data under the second dimension. The sixth prompt word data is integrated with the fifth prompt word data to obtain the first prompt word data.

[0129] For example, see Fig.11 . Fig.11 FIG. 4 is a flow chart of determining fifth prompt word data according to a data generation method of an embodiment of the present disclosure. Fig.11 As shown, the determination of the first prompt word data includes the following operations:

[0130] Operation S1101: Add each prompt word data under the first dimension to the first input data in turn to obtain extended input data. For example, the first input data is: "Draw a little girl, age 8". The target prompt word data is: {big eyes, high ponytail, dress, two-dimensional, movie light effect, ultra-high definition}. First, add "big eyes" to the first input data to obtain the extended input data "Draw a little girl, age 8, big eyes";

[0131] Operation S1102: After each addition is completed, determine whether the scoring data of the extended input data meets the preset conditions. If so, execute operations S1103 to S1105. If not, execute operation S1101. For example: input the extended input data into the scoring system, obtain the score of the extended input data, and determine whether the score is greater than the preset scoring threshold. The score can be determined based on the relevance of the added prompt word data and the first input data. For example: if the first input data is "Draw a little girl, reading", then the extended input data: "Draw a little girl, reading, on the playground" will score higher than the extended input data: "Draw a little girl, reading, in the classroom";

[0132] Operation S1103: adding the prompt word data to the recommended prompt word set. For example, if the score of the extended input data "draw a little girl, age 8, with big eyes" is greater than a preset threshold, then "big eyes" is added to the recommended prompt word set;

[0133] Operation S1104: Determine whether the traversal of each prompt word data under the first dimension is complete. If the traversal is complete, execute operations S1105 to S1106. If the traversal is not complete, execute S1101. For example: if the traversal is not complete, add "high ponytail" to the first input data, and obtain the extended input data as "draw a little girl, age 8, high ponytail";

[0134] Operation S1105: Determine fourth prompt word data based on the recommended prompt word set. For example, if the recommended prompt word set is: {big eyes, high ponytail, dress, two-dimensional}, the recommended prompt word set is used as the fourth prompt word data.

[0135] Operation S1106: adding the fourth prompt word data to the first input data to obtain fifth prompt word data.

[0136] According to operations S1101 to S1106, and so on, each prompt word data under the second dimension can be sequentially added to the fifth prompt word data to obtain extended prompt word data, obtain the scoring data corresponding to the extended prompt word data as the scoring data of each prompt word data, and perform subsequent steps to determine the sixth prompt word data under the second dimension. The sixth prompt word data is integrated with the fifth prompt word data to obtain the first prompt word data.

[0137] According to the method provided by the above embodiment, prompt words can be effectively screened and optimized, ensuring that each prompt word is verified by scoring and is relevant to the original input data of the user.

[0138] In some embodiments, a data generation method further includes: in response to obtaining a configuration operation on a target file, updating configuration parameters of the target file and generating model parameters of a target processing model.

[0139] The configuration operation of the target file may include inputting configuration parameters for the target file, or uploading a reference target file, or directly adjusting relevant configuration parameters on the target file. The configuration parameters of the target file are updated according to the configuration operation of the target file, for example: according to the input configuration parameters, the target processing model is called to generate an updated target file that conforms to the configuration parameters. Or the configuration parameters of the reference target file are parsed, and according to the parsed configuration parameters, the target processing model is called to generate an updated target file that conforms to the configuration parameters. Or in response to the adjustment operation on the target file, such as the shearing and scaling operation on the generated image, the adjusted target file is returned as the result.

[0140] In response to obtaining the configuration operation on the target file, the obtained configuration operation can be used as the historical behavior information of the target user, and the model parameters of the target processing model can be updated according to the historical behavior information to generate a target file that better meets the user's needs.

[0141] By updating the configuration parameters of the file and the model parameters of the processing model in response to the user's configuration operation on the target file, dynamic adjustment of the target file generation effect is achieved, ensuring that the target file can adapt to changes in user needs and improving user experience.

[0142] Fig.12 A schematic block diagram of an electronic device that can be used to implement the method of the embodiment of the present disclosure is schematically shown.

[0143] like Fig.12 As shown, the electronic device 1200 according to an embodiment of the present disclosure includes a processor 1201, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 1202 or a program loaded from a storage part 1208 to a random access memory (RAM) 1203. The processor 1201 may include, for example, a general-purpose microprocessor (e.g., a CPU), an instruction set processor and / or a related chipset and / or a dedicated microprocessor (e.g., an application-specific integrated circuit (ASIC)), etc. The processor 1201 may also include an onboard memory for caching purposes. The processor 1201 may include a single processing unit or multiple processing units for performing different actions of the method flow according to an embodiment of the present disclosure.

[0144] In RAM 1203, various programs and data required for the operation of electronic device 1200 are stored. Processor 1201, ROM 1202 and RAM 1203 are connected to each other through bus 1204. Processor 1201 performs various operations of the method flow according to the embodiment of the present disclosure by executing the program in ROM 1202 and / or RAM 1203. It should be noted that the program can also be stored in one or more memories other than ROM 1202 and RAM 1203. Processor 1201 can also implement the method provided by the embodiment of the present disclosure by executing the program stored in the one or more memories: in response to obtaining the first input data, obtaining the target prompt word data matching the first input data from the target database, so as to process the first input data into the first prompt word data by using the target prompt word data; adjusting the first prompt word data based on the target guide data to obtain the second prompt word data; generating the target file based on the second prompt word data.

[0145] In other embodiments, the electronic device further comprises a first body and a second body that are rotatably connected, and at least one of the first body and the second body has a display screen. That is, the electronic device can be a laptop computer with a single display screen, or a dual-screen laptop computer with two display screens.

[0146] According to an embodiment of the present disclosure, the electronic device 1200 may further include an input / output (I / O) interface 1205, which is also connected to the bus 1204. The electronic device 1200 may further include one or more of the following components connected to the I / O interface 1205: an input portion 1206 including a keyboard, a mouse, etc.; an output portion 1207 including a cathode ray tube (CRT), a liquid crystal display (LCD), etc., and a speaker, etc.; a storage portion 1208 including a hard disk, etc.; and a communication portion 1209 including a network interface card such as a LAN card, a modem, etc. The communication portion 1209 performs communication processing via a network such as the Internet. A drive 1210 is also connected to the I / O interface 1205 as needed. A removable medium 1211, such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, etc., is installed on the drive 1210 as needed, so that a computer program read therefrom is installed into the storage portion 1208 as needed.

[0147] The present disclosure also provides a computer-readable storage medium, which may be included in the device / apparatus / system described in the above embodiments; or may exist independently without being assembled into the device / apparatus / system. The above computer-readable storage medium carries one or more programs, and when the above one or more programs are executed, the method according to the embodiment of the present disclosure is implemented.

[0148] According to an embodiment of the present disclosure, a computer-readable storage medium may be a non-volatile computer-readable storage medium, for example, may include but is not limited to: 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), a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination thereof. In the present disclosure, a computer-readable storage medium may be any tangible medium containing or storing a program, which may be used by or in combination with an instruction execution system, an apparatus, or a device. For example, according to an embodiment of the present disclosure, a computer-readable storage medium may include the ROM 1202 and / or RAM 1203 described above and / or one or more memories other than ROM 1202 and RAM 1203.

[0149] The embodiment of the present disclosure also includes a computer program product, which includes a computer program, and the computer program contains program code for executing the method shown in the flowchart. When the computer program product is run in a computer system, the program code is used to enable the computer system to implement the method provided by the embodiment of the present disclosure.

[0150] The above functions defined in the system / device of the embodiment of the present disclosure are performed when the computer program is executed by the processor 1201. According to the embodiment of the present disclosure, the system, device, module, unit, etc. described above can be implemented by a computer program module.

[0151] In one embodiment, the computer program may rely on tangible storage media such as optical storage devices, magnetic storage devices, etc. In another embodiment, the computer program may also be transmitted and distributed in the form of signals on a network medium, and downloaded and installed through the communication part 1209, and / or installed from the removable medium 1211. The program code contained in the computer program may be transmitted using any appropriate network medium, including but not limited to: wireless, wired, etc., or any suitable combination of the above.

[0152] In such an embodiment, the computer program can be downloaded and installed from the network through the communication part 1209, and / or installed from the removable medium 1211. When the computer program is executed by the processor 1201, the above functions defined in the system of the embodiment of the present disclosure are performed. According to the embodiment of the present disclosure, the system, device, means, module, unit, etc. described above can be implemented by a computer program module.

[0153] It should be noted that the collection, storage, use, processing, transmission, provision, disclosure and application of user personal information involved in the technical solution of this disclosure are in compliance with the provisions of relevant laws and regulations, necessary confidentiality measures have been taken, and they do not violate public order and good morals. In the technical solution of this disclosure, the user's authorization or consent is obtained before obtaining or collecting user personal information.

[0154] According to an embodiment of the present disclosure, the program code for executing the computer program provided by the embodiment of the present disclosure can be written in any combination of one or more programming languages. Specifically, these computing programs can be implemented using high-level process and / or object-oriented programming languages, and / or assembly / machine languages. Programming languages ​​include, but are not limited to, Java, C++, python, "C" language or similar programming languages. The program code can be executed entirely on the user computing device, partially on the user device, partially on the remote computing device, or entirely on the remote computing device or server. In the case of a remote computing device, the remote computing device can be connected to the user computing device through any type of network, including a local area network (LAN) or a wide area network (WAN), or can be connected to an external computing device (for example, using an Internet service provider to connect through the Internet).

[0155] The flow charts and block diagrams in the accompanying drawings illustrate the possible architecture, functions and operations of the systems, methods and computer program products according to various embodiments of the present disclosure. In this regard, each box in the flow chart or block diagram can represent a module, a program segment, or a part of a code, and the above-mentioned module, program segment, or a part of a 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 box can also occur in a different order from the order marked in the accompanying drawings. For example, two 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 box in the block diagram or flow chart, and the combination of the boxes in the block diagram 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.

[0156] It will be appreciated by those skilled in the art that the features described in the various embodiments and / or claims of the present disclosure may be combined and / or coupled in a variety of ways, even if such combinations and / or couplings are not explicitly described in the present disclosure. In particular, the features described in the various embodiments and / or claims of the present disclosure may be combined and / or coupled in a variety of ways without departing from the spirit and teachings of the present disclosure. All of these combinations and / or couplings fall within the scope of the present disclosure.

[0157] The embodiments of the present disclosure are described above. However, these embodiments are only for illustrative purposes and are not intended to limit the scope of the present disclosure. Although the embodiments are described above, this does not mean that the measures in the various embodiments cannot be used in combination. The scope of the present disclosure is defined by the appended claims and their equivalents. Without departing from the scope of the present disclosure, those skilled in the art may make a variety of substitutions and modifications, which should all fall within the scope of the present disclosure.

Claims

1. A data generation method, comprising: In response to obtaining first input data, acquiring target prompt word data matching the first input data from a target database, so as to process the first input data into first prompt word data by using the target prompt word data; Adjusting the first prompt word data based on the target guidance data to obtain second prompt word data; The target processing model is called to generate a target file based on the second prompt word data.

2. The method according to claim 1, wherein: Acquiring target prompt word data matching the first input data from a target database includes at least one of the following: In response to obtaining the first input data, identifying a first user intention represented by the first input data, and acquiring matching target prompt word data from a first database based on the first user intention; In response to obtaining the first input data, obtaining user portrait information of a target user, and obtaining matching target prompt word data from a second database based on the user portrait information; In response to obtaining the first input data, outputting a first response result for the first input data, and acquiring matching target prompt word data from a third database based on the first response result; In response to obtaining the first input data, the intention type represented by the first input data is identified, and target prompt word data matching the first input data is acquired from a fourth database based on the intention type.

3. The method according to claim 2, wherein: Acquiring matching target prompt word data from a fourth database corresponding to the intention type based on the first input data includes: extracting target feature data from the first input data based on the intent type; Retrieving a matching prompt word data set from a fourth database based on the target feature data, and determining evaluation information corresponding to the prompt word data set; The target prompt word data is determined from the prompt word data set based on the evaluation information.

4. The method according to claim 2, wherein: Acquiring matching target prompt word data from a second database based on the user portrait information includes: Obtaining historical operation data of a target user, and processing the historical operation data into structured data; The structured data is used to construct a knowledge graph to obtain the target prompt word data from a corresponding node of the knowledge graph based on the first input data.

5. The method according to claim 1 or 2, wherein: Processing the first input data into first prompt word data using the target prompt word data includes at least one of the following: Performing prompt engineering processing on the first input data to obtain third prompt word data, and embedding the target prompt word data into a corresponding position of the third prompt word data to obtain the first prompt word data; Performing generation processing on the target prompt word data and the first input data to obtain the first prompt word data; Outputting target prompt word data for the first input data in a window interface of a target application, and in response to obtaining a selection operation on the target prompt word data, adding the prompt word data determined by the selection operation to the first input data to obtain the first prompt word data.

6. The method according to claim 1, wherein: The first prompt word data is adjusted based on the target guidance data to obtain the second prompt word data, including: The type and / or source of the target guidance data is obtained, and the first prompt word data is updated according to a corresponding updating method based on the type and / or source to obtain the second prompt word data.

7. The method according to claim 1, wherein: Adjusting the first prompt word data based on the target guidance data includes at least one of the following: Adjusting the first prompt word data based on the first input data; adjusting the first prompt word data based on second input data input into the target application; The first prompt word data is adjusted based on feedback data of a second response result obtained in response to the first prompt word data.

8. The method according to claim 1, wherein: Processing the first input data into first prompt word data using the target prompt word data includes: Obtaining score data of each prompt word data under the first dimension to determine fourth prompt word data under the first dimension; Integrate the fourth prompt word data with the first input data to obtain fifth prompt word data; Obtaining score data of each prompt word data under the second dimension to determine sixth prompt word data under the second dimension; The sixth prompt word data and the fifth prompt word data are integrated to obtain the first prompt word data.

9. The method according to claim 1, further comprising: In response to obtaining a configuration operation on the target file, configuration parameters of the target file and model parameters of a processing model for generating the target file are updated.

10. An electronic device comprising at least one processor and at least one processing model capable of running on the processor, wherein the processing model can be called by a target application to perform at least one of the following: In response to obtaining first input data, acquiring target prompt word data matching the first input data from a target database, so as to process the first input data into first prompt word data by using the target prompt word data; Adjusting the first prompt word data based on the target guidance data to obtain second prompt word data; A target file is generated based on the second prompt word data.

Images