Data generation method, text generation method and electronic equipment

Abstract

The invention discloses a data generation method, a text generation method and electronic equipment, and the data generation method comprises the steps: responding to a target trigger event, and obtaining target watermark information; generating target data based on the target watermark information and the input content of the target input data, so that the target watermark information is directly or indirectly embedded into the target data; wherein the target watermark information is used for identifying the source of the target data, and the watermark information embedded in the target data corresponding to different trigger events is different.

Description

Technical Field

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

[0002] With the continuous advancement of computer technology, many data generation tools are now available, which can generate data according to user needs. However, although the process of data generation has become increasingly convenient, how to effectively track and determine the specific source of the data generated by the data generation tools has become a problem that needs to be solved urgently. Summary of the invention

[0003] This application provides the following technical solutions:

[0004] The first aspect of the present application provides a data generation method, comprising:

[0005] In response to a target triggering event, obtaining target watermark information;

[0006] generating target data based on the target watermark information and the target input content, so that the target watermark information is directly or indirectly embedded in the target data;

[0007] The target watermark information is used to identify the source of the target data, and the watermark information embedded in the target data corresponding to different triggering events is different.

[0008] The target triggering event is generated from at least one of the following:

[0009] In response to monitoring first input content input into the first application, generating the target trigger event;

[0010] generating the target triggering event in response to obtaining second input content, wherein the second input content is used to call a target generation model of the electronic device;

[0011] In response to obtaining third input content, generating the target trigger event, wherein the third input content can trigger the electronic device to perform an operation of adding watermark information;

[0012] In response to monitoring the operation of sharing the target file, the target trigger event is generated.

[0013] Wherein, obtaining the target watermark information includes at least one of the following:

[0014] Obtaining source information of the target triggering event, and obtaining target watermark information matching the source information based on the source information;

[0015] Obtaining user information and / or model information associated with the target triggering event, and generating the target watermark information based on the user information and / or model information;

[0016] A user intention represented by the target input content is obtained, and target watermark information matching the user intention is obtained based on the user intention.

[0017] The generating of the target watermark information based on the user information and / or the model information includes at least one of the following:

[0018] Processing the model information of the target generation model into first character string data and / or a first image element set as the target watermark information;

[0019] Processing user information associated with the target triggering event into second string data and / or a second image element set as the target watermark information;

[0020] Processing the user information and time information associated with the target triggering event into a second string data and / or a second image element set as the target watermark information;

[0021] Processing model information of a target generation model into first string data and / or a first image element set, processing user information associated with a target triggering event into second string data and / or a second image element set, and processing the first string data and / or the first image element set, and the second string data and / or the second image element set into target watermark information;

[0022] The model information of the target generation model is processed into a first string data and / or a first image element set, the user information and time information associated with the target trigger event is processed into a second string data and / or a second image element set, and the first string data and / or the first image element set, as well as the second string data and / or the second image element set, are processed into target watermark information.

[0023] The generating of target data based on the target watermark information and the target input content includes at least one of the following:

[0024] Identify the initial user intention represented by the target input content, use the target watermark information or its associated data to update the initial user intention to obtain the target user intention, and generate target data matching the target user intention;

[0025] Inputting the target watermark information and the target input content into a target generation model matching the target trigger event, and in the process of the target generation model performing a generation operation corresponding to the target input content, embedding at least part of the target watermark information or its associated data into the generated target data;

[0026] A target position is determined in the initial generated data generated based on the target input content, and at least part of the target watermark information or its associated data is inserted into the corresponding target position to generate the target data.

[0027] Wherein, so that the target watermark information is directly or indirectly embedded into the target data, including at least one of the following:

[0028] Obtaining attribute information of the target data, and embedding the target watermark information or its associated data into the target data according to a target embedding strategy based on the attribute information, wherein the amount of the target watermark information or its associated data embedded into the target data under different embedding strategies is different;

[0029] Obtaining usage information of the target data, and embedding a target portion of the target watermark information or its associated data into the target data based on the usage information, wherein the target portion is a portion matching the usage information;

[0030] The authorization information of the target user is obtained, and based on the authorization information, it is determined to embed the target watermark information or the associated data of the target watermark information into the target data.

[0031] Also includes at least one of the following:

[0032] After generating the target data, performing quality inspection on the target data;

[0033] After the target data is generated, the watermark information embedded in the target data is verified.

[0034] Also includes:

[0035] In response to obtaining a traceability request for the target data, extracting target watermark information in the target data;

[0036] The extracted target watermark information is decrypted, and a tracing result of the target data is obtained based on the decryption result.

[0037] Another aspect of the present application provides a text generation method, comprising:

[0038] In response to obtaining a target text generation requirement, obtaining target watermark information, wherein the target watermark information includes first character string data and / or second character string data;

[0039] generating target text content based on the target watermark information and the target input content, so that the first character string data and / or the second character string data are directly or indirectly embedded into the target text content;

[0040] The target watermark information is used to identify the source of the target text, and the watermark information embedded in the target text content corresponding to different text generation requirements is different.

[0041] A third aspect of the present application provides an electronic device, including:

[0042] The memory is used to store computer programs;

[0043] The processor is used to execute the computer program so that the electronic device can implement any one of the data generation methods described above or the text generation method described above.

[0044] A fourth aspect of the present application provides a data generating device, including:

[0045] An acquisition module, used for obtaining target watermark information in response to a target trigger event;

[0046] A generating module, configured to generate target data based on the target watermark information and target input content, so that the target watermark information is directly or indirectly embedded in the target data;

[0047] The target watermark information is used to identify the source of the target data, and the watermark information embedded in the target data corresponding to different triggering events is different.

[0048] A fifth aspect of the present application provides a text generation device, comprising:

[0049] An obtaining module, configured to obtain target watermark information in response to a target text generation requirement, wherein the target watermark information includes first character string data and / or second character string data;

[0050] A generating module, configured to generate target text content based on the target watermark information and target input content, so that the first character string data and / or the second character string data are directly or indirectly embedded into the target text content;

[0051] The target watermark information is used to identify the source of the target text, and the watermark information embedded in the target text content corresponding to different text generation requirements is different. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] The above and other features, advantages and aspects of the embodiments of the present disclosure will become more apparent with reference to the following detailed description in conjunction with the accompanying drawings. Throughout the accompanying drawings, the same or similar reference numerals represent the same or similar elements. It should be understood that the drawings are schematic and the originals and elements are not necessarily drawn to scale.

[0053] Figure 1A flowchart of a data generation method provided in Example 1 of the present application;

[0054] Figure 2 A schematic diagram of a scenario implementation of a data generation method provided in this application;

[0055] Figure 3 A schematic diagram of another implementation scenario of a data generation method provided in this application;

[0056] Figure 4 A schematic diagram of another implementation scenario of a data generation method provided in this application;

[0057] Figure 5 A schematic diagram of another implementation scenario of a data generation method provided in this application;

[0058] Figure 6 A schematic diagram of another implementation scenario of a data generation method provided in this application;

[0059] Figure 7 A flowchart of a data generation method provided in Example 7 of the present application;

[0060] Figure 8 A flowchart of a data generation method provided in Example 8 of the present application;

[0061] Fig. 9 A flowchart of a data generation method provided in Example 9 of the present application;

[0062] Fig.10 A flowchart of a text generation method provided in Example 10 of the present application;

[0063] Fig.11 A schematic diagram of the structure of a data generation device provided in this application. DETAILED DESCRIPTION

[0064] The following describes the embodiments of the present application in conjunction with the drawings in the embodiments of the present application. The terms used in the implementation method section of the present application are only used to explain the specific embodiments of the present application, and are not intended to limit the present application.

[0065] The embodiments of the present application are described below in conjunction with the accompanying drawings. Those skilled in the art will appreciate that, with the development of technology and the emergence of new scenarios, the technical solutions provided in the embodiments of the present application are also applicable to similar technical problems.

[0066] The terms "first", "second", etc. in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects, and need not be used to describe a specific order or sequential order. It should be understood that the terms used in this way can be interchangeable under appropriate circumstances, which is only to describe the distinction mode adopted by the objects of the same attributes when describing in the embodiments of the present application. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, so that the process, method, system, product or equipment comprising a series of units need not be limited to those units, but may include other units that are not clearly listed or inherent to these processes, methods, products or equipment.

[0067] In the field of natural language processing (NLP) and text generation, with the rapid development of technology, target generation models have become core tools in scenarios such as text generation, automatic writing, and dialogue systems. These models can generate high-quality, coherent texts, greatly improving the efficiency and diversity of text creation. However, texts generated by target generation models face severe challenges in copyright protection, model traceability, and user and time tracking.

[0068] At present, although the target generation model has achieved remarkable results in text creation, it is impossible to accurately identify the specific target generation model that generates the text or track the specific generation time of the text and related user information based on the text it generates. This situation has brought great trouble to copyright protection, because once the generated text is abused or infringed, it is difficult to determine the identity of the model, the generation time of the text and related user information, making it difficult to hold accountable.

[0069] Although existing watermarking technology provides a solution to this problem to a certain extent, it has obvious limitations.

[0070] Static watermarking is one of the common methods, which implements copyright protection by embedding fixed identifiers (such as model names or copyright statements) in text after it is generated. Although static watermarks are relatively simple to embed, they lack dynamic changes and cannot track specific user and time information, so their traceability capabilities are limited. In addition, static watermarks are easy to detect and remove, which reduces their effectiveness in practical applications.

[0071] Another common watermarking method is visual watermarking and metadata watermarking. Visual watermarking usually places the watermark information in a prominent position of the text, while metadata watermarking uses the metadata of the text to carry the watermark information. Although these two methods have the advantages of direct visibility or detection, they have obvious defects in concealment. Visual watermarks affect the appearance and reading experience of the text, while metadata can be easily tampered with or removed, thus losing the protective effect of the watermark. Therefore, the present application provides a data generation method to enhance the copyright protection and traceability capabilities of model-generated data.

[0072] The present application is further described in detail below in conjunction with the accompanying drawings and specific implementation methods.

[0073] Reference Figure 1 , is a flow chart of a data generation method provided in Example 1 of the present application. The method can be applied to electronic devices. The present application does not limit the product type of the electronic device. For example, Figure 1 As shown, the method may include but is not limited to the following steps:

[0074] Step S101: In response to a target triggering event, obtain target watermark information.

[0075] In this embodiment, in response to different target trigger events, the target watermark information obtained may be different. For example, in a multifunctional content creation application, when a user selects a text generation function, a target trigger event is generated, and string data (an implementation of the target watermark information) may be obtained.

[0076] In the above-mentioned multifunctional content creation application, when the user selects the image generation function, a target trigger event is generated, and string data (an implementation method of the target watermark information) can be obtained; or, string data and image elements (an implementation method of the target watermark information) can be obtained.

[0077] In the above-mentioned multifunctional content creation application, when the user selects the audio generation function, a target trigger event is generated and audio data can be obtained ((an implementation method of the target watermark information)).

[0078] The character string data obtained in the image generation function and the text generation function may be different.

[0079] Step S102: Generate target data based on the target watermark information and the target input content, so that the target watermark information is directly or indirectly embedded in the target data.

[0080] The target input content can be used to guide the generation process of the target data. The target input content can be at least one of text input content, image input content, and voice input content. Accordingly, the target data can include but is not limited to: at least one of text data, image elements, video data, and audio data.

[0081] For example, in a text processing scenario (text-to-text scenario), the target input content can be at least one of text input content and image input content, which is used to guide the generation process of text data; in an image processing scenario (text-to-image scenario or image-to-image scenario), the target input content can be at least one of image input content and text input content, which is used to guide the generation process of image elements; in an audio processing scenario, the target input content can be voice input content, which is used to guide the generation process of audio data.

[0082] The target watermark information is directly embedded into the target data, that is, the target data contains the target watermark information, but the target watermark information exists in the target data in a non-directly readable or non-directly visible form. For example, the target watermark information may include: Lenovo, and the target data may include the following text: "In today's digital age, the Internet has penetrated into all aspects of our lives. Whether it is work, study, entertainment, or social interaction, network connections play an indispensable role. ... In addition, through the Internet, we can want to get to know more people, expand our social circles, and communicate, learn, and progress with like-minded friends. ... These problems have prompted us to think about how to better use the Internet to maintain and develop healthy interpersonal relationships. In short, while the Internet helps us expand interpersonal relationships, it also brings challenges. We need to pay attention to and maintain interpersonal relationships in real life while taking advantage of its convenience." Among them, "lian", "xiang", "suo", and "you" are included in the target data, but are distributed in different positions in the target data. For example, in the target watermark information, "lian" is located in "network contact", in the target watermark information, "xiang" is located in "want to get to know", in the target watermark information, "suo" is located in "thinking", and in the target watermark information, "you" is located in "help".

[0083] The target watermark information is indirectly embedded into the target data, that is, the target data does not directly contain the original form of the target watermark information, but the target data may contain deformed data of the target watermark information, or the target data and the target watermark information have an implicit association relationship.

[0084] Among them, the target watermark information can be used to identify the source of the target data, so that when the target data needs to be traced, the generation information of the target data can be traced back by parsing the target watermark information, such as which generation model the target data was generated by, the generation time, the operating user information, etc.

[0085] In response to different target triggering events, the target watermark information obtained may be different, so that the target watermark information can be flexibly embedded, and the watermark information embedded in the target data corresponding to different triggering events may be different.

[0086] In this embodiment, by responding to a target trigger event, target watermark information is obtained, and target data is generated based on the target watermark information and target input content, so that the target watermark information is directly or indirectly embedded in the target data. This ensures the concealment of the target watermark information and reduces the risk of the target watermark information being maliciously detected or removed.

[0087] Moreover, by responding to different target trigger events to obtain different target watermark information, it can flexibly adapt to a variety of application scenarios and needs, so that the embedding of target watermark information is no longer limited to a fixed mode or content, but can be adjusted and optimized according to actual needs.

[0088] As another optional embodiment of the present application, a data generation method is provided in Embodiment 2 of the present application. This embodiment is mainly an implementation method for generating a target trigger event in Embodiment 1. The target trigger event may be, but is not limited to, generated from at least one of the following:

[0089] Step S11: In response to monitoring the first input content input into the first application, generating the target trigger event.

[0090] The first application may have an input box or an input button (eg, a microphone control), and the input box may be used to receive the first input content.

[0091] Correspondingly, the first application or the first application calls the intent recognition model or other sub-model in the target generation model to parse the first input content (an implementation method of the target input content), identify the intent from the first input content, and then call the target generation model or the processing sub-model in the target generation model to generate target data.

[0092] For example, the first application may include but is not limited to any of the following:

[0093] Image processing applications (such as Lenovo's Creator Zone): mainly used to generate images or process images. Image processing applications have a main editing interface for displaying the image currently being edited or created. Image processing applications may also include auxiliary tool panels, which may include text input boxes. Users can enter text descriptions or select images for editing in the input boxes, and image processing applications can generate or modify images based on these inputs. For example, Figure 2As shown, if the first input content is monitored in the text input box in the raw image application, such as "Please add some sunset effects to this picture", a target trigger event can be generated. The text input box is different from the functional area (such as the prompt input box) specifically used to call the AI ​​generation model.

[0094] Generated text application: mainly used to generate text content, such as articles, reports, etc. Users can enter the first input content such as keywords, topics or sentences in its text input box, and the generated text application can generate corresponding text content based on the first input content. Similarly, the text input box is different from the functional area specifically used to call the AI ​​generation model.

[0095] Video generation application: used to generate or edit video content. Users can enter video descriptions, select video clips, or add special effects in its input box. Video generation applications can generate or modify video content. This input box is different from the functional area specifically used to call AI generation models.

[0096] Intelligent agent applications: such as Lenovo Xiaotian or AI NOW, which can integrate multiple functions, including but not limited to text generation, image processing, speech recognition, etc. Users can generate or process data by interacting with intelligent agent applications.

[0097] Other applications that can call the above applications: such as conference applications, social media applications (such as WeChat), photo album applications, etc. Although these applications may not have the ability to generate or process data themselves, they can achieve the effect of model processing services by calling the APIs or plug-ins of the above-mentioned raw image applications, intelligent agent applications, etc. For example, in conference applications, users can generate images of meeting minutes by calling the API of raw image applications; in social media applications, users can generate interesting text or image content by calling the plug-ins of intelligent agent applications.

[0098] Step S12: generating the target triggering event in response to obtaining the second input content, wherein the second input content is used to call the target generation model of the electronic device;

[0099] The first application may have a functional area specifically for calling the AI ​​generation model. The functional area may include one or more prompt input boxes for receiving second input content, and the second input content may be used to directly call the target generation model of the electronic device. For example, in the raw image application, in addition to the auxiliary tool panel, a prompt input box may also be included, such as Figure 3 As shown in the figure, in the prompt input box of the raw image application, you can enter "generate a landscape painting with mountains and lakes". Accordingly, the raw image application can directly call the target generation model to generate a landscape painting such as Figure 4 Landscape painting with mountains and lake shown.

[0100] Of course, in addition to entering the second input content in the prompt input box, the second input content can also come from the task location (i.e., the location when performing a specific task in the first application). For example, at the task location, there are several preset options, such as "Mountain Scenery", "Beach Sunset", "City Night Scenery", etc. If the "Mountain Scenery" option is clicked, the click behavior can be identified as the second input content, and the second input content can be used to call the AI ​​generation model to generate a mountain scenery image.

[0101] Correspondingly, after receiving the second input content (an implementation of the target input content), the first application can directly call the target generation model to generate target data.

[0102] Step S13: in response to obtaining the third input content, generating the target trigger event, wherein the third input content can trigger the electronic device to perform an operation of adding watermark information.

[0103] In this embodiment, the third input content can be used to trigger the addition of watermark information when the specified content is sent out, so that the user can actively embed the watermark information according to their own needs. For example, after the raw image application has generated a picture (an implementation method of the specified content), the user can click the watermark option of the raw image application, and the click behavior can be identified as the third input content.

[0104] Accordingly, generating target data based on target watermark information and target input content may include: embedding the target watermark information in designated content to obtain the target data.

[0105] The third input content can also be used to trigger the addition of watermark information when generating content, so that users can actively embed watermark information according to their own needs. For example, the user can enter "generate a landscape painting with mountains and lakes and add a watermark" in the prompt input box in the raw image application, and the content entered in the prompt input box is the third input content.

[0106] Accordingly, generating the target data based on the target watermark information and the target input content may include: in the process of performing a generation operation corresponding to the third input content, embedding the target watermark information into the generated target data.

[0107] Step S14: In response to monitoring the operation of sharing the target file, a target trigger event is generated.

[0108] In this embodiment, the smart watermark embedding function can be configured in the electronic device, and when the electronic device monitors the operation of sharing the target file (which can be a file determined based on the naming of the file, content analysis, or the user's previous mark), a target trigger event can be generated. Accordingly, in response to the target trigger event, the smart watermark embedding function can be triggered to obtain the target watermark information, embed the target watermark information into the target file (an implementation method of the target input content), and realize that the electronic device automatically embeds the watermark information into the target file to be analyzed, which can not only improve the convenience of the user, but also effectively identify the source of the target file.

[0109] For example, when a user completes a market analysis report and sends it to a partner via email, when the user clicks the "Send" button in the email client, the electronic device can monitor the sharing operation and identify the market analysis report as a target file, and can generate a target trigger event. Accordingly, in response to the target trigger event, the target watermark information can be obtained and embedded in the market analysis report.

[0110] As another optional embodiment of the present application, a data generation method is provided in Embodiment 3 of the present application. This embodiment is mainly an implementation method for obtaining target watermark information in Embodiment 1. The target watermark information can be obtained by but is not limited to at least one of the following:

[0111] Step S21: Obtain source information of a target triggering event, and obtain target watermark information matching the source information based on the source information.

[0112] The source information of the target trigger event can be used to indicate the type of the target trigger event. For example, the source information can distinguish whether the target trigger event is triggered by a text-to-text operation (text-to-text), a text-to-image operation (text-to-image), an image-to-image operation (image-to-image), or a text-to-video operation (text-to-video).

[0113] In this embodiment, a watermark library containing multiple watermark information may be pre-established, and each type of watermark information may be associated with a specific type of triggering event.

[0114] After obtaining the source information of the target trigger event, the watermark information matching the source information can be searched in the watermark library. Once the matching watermark information is found, it can be determined as the target watermark information and prepared for use in subsequent data generation or processing.

[0115] Step S22: Obtain user information and / or model information associated with the target triggering event, and generate the target watermark information based on the user information and / or model information.

[0116] The user information may be used to identify the user, and the model information may be used to identify the target generation model used to generate data.

[0117] For example, in a raw image application, a user can generate a new image with a specific style by uploading his own image and entering some descriptive text. When a user uploads a picture and enters descriptive text, a target trigger event, namely a raw image event, can be generated. Accordingly, in response to the raw image event, the user name "Alice" and the user ID "220345" are obtained, as well as the name "ArtGAN" and the version number "v2.0" of the image generation model of the raw image application. Based on the user name "Alice" and the user ID "220345", as well as the name "ArtGAN" and the version number "v2.0" of the image generation model of the raw image application, the target watermark information is generated, and the target watermark information may include "User: Alice, ID: 12345, Image Generation Model: ArtGAN v2.0".

[0118] Step S23: obtaining the user intention represented by the target input content, and obtaining target watermark information matching the user intention based on the user intention.

[0119] In this embodiment, the user intention represented by the target input content may include but is not limited to: the specific scenario of data generation (such as generating images, generating text, generating videos, etc.), the user's needs (for example, whether the format of the generated image is 2D or 3D, the word limit of the generated text, the frame rate and clarity requirements of the video, etc.), and the final effect that the user expects to achieve (such as document summary, document expansion, image beautification, video editing, etc.).

[0120] In this embodiment, a watermark library containing multiple watermark information may be pre-established. Each watermark information may be associated with a specific user intention (eg, generation scenario, user demand, effect expected by the user, etc.).

[0121] After obtaining the user's intention, the watermark information that matches the user's intention can be searched in the watermark library. Once the matching watermark information is found, it can be reused as the target watermark information to improve efficiency and reduce resource consumption.

[0122] If no matching watermark information is found in the watermark library, new watermark information can be generated in real time as the target watermark information according to the user's intention.

[0123] For example, in an intelligent writing assistant, a user inputs a text "Artificial intelligence is changing our lives" (i.e., an implementation of the target input content). By analyzing the text input by the user, it is determined that the user intends to generate an extended document about "the impact of artificial intelligence on life" based on this text. Based on the user's intention, a watermark information related to "document extension" is selected from the watermark library as the target watermark information.

[0124] As another optional embodiment of the present application, a data generation method is provided in Embodiment 4 of the present application. This embodiment is mainly an implementation method for generating target watermark information based on user information and / or model information in Embodiment 3. Generating target watermark information based on user information and / or model information may include but is not limited to at least one of the following:

[0125] Step S31: Process the model information of the target generation model into a first string data and / or a first image element set as target watermark information.

[0126] The target generation model can be used to generate target data. The model information can be used to identify the target generation model. For example, the model information can include but is not limited to: model name, model version number, model manufacturer, etc.

[0127] For example, when the target generation model generates text data, the model information may be processed into first character string data.

[0128] Alternatively, when the target generation model generates image data, the model information can be processed into a first set of image elements, such as generating a small, unique icon representing the target generation model, which can be any shape or pattern.

[0129] Alternatively, when the target generation model generates image data, the model information may be processed into the first string data and the first image element set.

[0130] Step S32: Process the user information associated with the target triggering event into a second string data and / or a second image element set as target watermark information.

[0131] User information can be used to identify the user who triggers the target trigger event (such as a picture generation event or a text generation event). User information can include user name, ID, social name, IP address, geographic location, etc.

[0132] For example, when the target generation model generates text data, the user information associated with the target triggering event may be processed into second character string data.

[0133] Alternatively, when the target generation model generates image data, the user information associated with the target trigger event can be processed into a second set of image elements. For example, a small, unique icon is generated to represent the user information associated with the target trigger event. The icon can be any shape or pattern.

[0134] Alternatively, when the target generation model generates image data, the user information associated with the target triggering event may be processed into the second string data and the second image element set.

[0135] Step S33: Process the user information and time information associated with the target triggering event into a second string data and / or a second image element set as target watermark information.

[0136] The time information can be used to identify the execution time of a generation task (eg, a generated image event or a generated text event).

[0137] For example, when the target generation model generates text data, the user information and time information associated with the target triggering event may be processed into second character string data.

[0138] Alternatively, when the target generation model generates image data, the user information and time information associated with the target trigger event can be processed into a second set of image elements. For example, a small, unique first icon is generated to represent the user information associated with the target trigger event. A second icon is generated to represent the time information. The second icon can be a non-obtrusive clock, sun, or other image elements that can represent time information.

[0139] Alternatively, when the target generation model generates image data, the user information and time information associated with the target triggering event may be processed into the second string data and the second image element set.

[0140] Step S34, processing the model information of the target generation model into a first string data and / or a first image element set, processing the user information associated with the target trigger event into a second string data and / or a second image element set, and processing the first string data and / or the first image element set, and the second string data and / or the second image element set into target watermark information.

[0141] In this embodiment, the target watermark information may be used to identify model information of the target generation model and user information associated with the target triggering event.

[0142] For example, in a raw image application, user "Charlie" generates an image using a target generative model named "ArtGAN". The information "ArtGAN, Charlie" can be converted into a string data and embedded into the image generated by the target generative model. Alternatively, the information "ArtGAN, Charlie" can be converted into a string data, and the information "ArtGAN, Charlie" can be processed into a set of image elements, and the string data and the set of image elements can be embedded into the image generated by the target generative model.

[0143] Step S35: Process the model information of the target generation model into a first string data and / or a first image element set, process the user information and time information associated with the target trigger event into a second string data and / or a second image element set, and process the first string data and / or the first image element set, and the second string data and / or the second image element set into target watermark information.

[0144] In this embodiment, when generating watermark information, a certain order may be followed, that is, the first string data and / or the first image element set are placed first, and the second string data and / or the second image element set are placed later, thereby forming a watermark list.

[0145] In this embodiment, if the watermark list is too long, it may interfere with the semantics and readability of the text data, or cause obvious visual impact in the image data. Therefore, in this embodiment, the length of the watermark list can be controlled to ensure that it contains the necessary information without causing adverse effects on the target data.

[0146] The target watermark information can be used to identify the model information of the target generation model, the user information and time information associated with the target triggering event.

[0147] For example, in an intelligent writing assistant, a user inputs a text "Artificial intelligence is changing our lives" (i.e., an implementation of the target input content). The intelligent writing assistant can generate an extended document about "the impact of artificial intelligence on life" based on the target generation model, and can process the model information of the target generation model into a first image element set (e.g., a model logo), and process the user information and the time information when the user inputs the text into a second string data. Each time the user inputs new text, the second string data will be updated.

[0148] The first image element set is equivalent to a fixed image element set, and the second character string data is equivalent to a dynamic character string data embedded in the target data.

[0149] In step S35, not only the model information of the target generation model is processed into the first string data and / or the first image element set, but also the user information and time information associated with the target trigger event are processed into the second string data and / or the second image element set. These two sets of data together constitute the target watermark information. Since the target watermark information contains the model information of the target generation model, the user information and time information associated with the target trigger event, the model, user and time of the specific generated text or image can be easily tracked according to the target watermark information, which is helpful for copyright protection, content auditing and error tracking.

[0150] Importantly, the second string data and / or the second image element set are dynamic and will change according to the user and time. Therefore, the target watermark information generated each time is unique, which makes the target watermark information difficult to be predicted or copied, thereby greatly improving the security of the target watermark information.

[0151] In this embodiment, after obtaining the source information of the target trigger event, the type of the target trigger event (such as text-generated text, text-generated image, image-generated image, text-generated video, etc.) can be determined according to the source information. If the target trigger event is a text-generated scene (such as text-generated text), the model information, user information, etc. can be processed into the first string data. If the target trigger event is a picture-generated scene, the model information, user information, etc. can be processed into the first image element set.

[0152] After obtaining the user intention represented by the target input content, the specific scenario of data generation (such as generating images, generating text, etc.) and the user's needs and expected effects can be determined according to the user's intention. If the user's intention is to generate text (such as document summary, document extension, etc.), the model information, user information, etc. can be processed into the first string data. If the user's intention is to generate images (such as image beautification, frame images in video clips, etc.), the model information, user information, etc. can be processed into the first image element set.

[0153] As another optional embodiment of the present application, a data generation method is provided in Embodiment 5 of the present application. This embodiment is mainly an implementation of step S102 in Embodiment 1. Step S102 may include but is not limited to at least one of the following:

[0154] Step S1021, identifying the initial user intention represented by the target input content, using the target watermark information or its associated data to update the initial user intention to obtain the target user intention, and generating target data that matches the target user intention.

[0155] In this embodiment, the target input content may be understood and analyzed to determine the initial user intent, for example, what type of data (such as text, image, video, etc.) the user wants to generate and the specific content or format requirements.

[0156] Then, the target watermark information or its associated data can be used to update the initial user intent to form the target user intent.

[0157] The associated data of the target watermark information may include but is not limited to at least one of the following: data that can replace the target watermark information without affecting the data integrity and functionality. It may contain information similar to or related to the target watermark information, but in a different format or expression.

[0158] Data after the target watermark information is encrypted by an encryption algorithm. For example, the model information of the target generation model in the target watermark information is hashed to generate a hash value to ensure the uniqueness and non-tamperability of the model information, and the user information and time information in the target watermark information are encrypted by a symmetric or asymmetric encryption algorithm to further enhance the security and uniqueness of the target watermark information.

[0159] For example, a user enters a piece of text as the target input content: "Please generate a promotional poster for Lenovo Group's latest product for me." At this point, the identifiable initial user intent is to generate a promotional poster with the theme of Lenovo Group's latest product. Then, the target watermark information (such as the model name "AI Poster Generator V1.0", user ID "223045", etc.) can be used to update the initial user intent to form the target user intent: "Please use AI Poster Generator V1.0 to generate a promotional poster for Lenovo Group's latest product for me, and reflect the information of user ID 220345 in the poster." In the process of generating the poster, a small icon or text containing the model name and user ID can be embedded in the corner or bottom of the poster as the target watermark information.

[0160] Step S1022: input the target watermark information and the target input content into a target generation model that matches the target trigger event, and in the process of the target generation model performing a generation operation corresponding to the target input content, embed at least part of the target watermark information or its associated data into the generated target data.

[0161] In the process of the target generation model performing the generation operation corresponding to the target input content, embedding at least part of the target watermark information or its associated data into the generated target data may include but is not limited to:

[0162] During the process of the target generation model executing the generation operation corresponding to the target input content, the model information of the target generation model in the target watermark information or its associated data is directly embedded into the generated target data, and the user information and time information in the target watermark information or its associated data are indirectly embedded into the generated target data.

[0163] During the process of the target generation model executing the generation operation corresponding to the target input content, the model information of the target generation model in the target watermark information or its associated data is directly embedded into the generated target data, which may include but is not limited to: updating the candidate words of the target generation model based on the model information of the target generation model in the target watermark information or its associated data, and based on the candidate words of the updated target generation model, during the process of the target generation model executing the generation operation corresponding to the target input content, embedding the model information into the generated target data.

[0164] Updating the candidate words of the target generation model based on the model information of the target generation model in the target watermark information or its associated data may include but is not limited to: adding the model information of the target generation model in the target watermark information or its associated data, synonyms, phrases or specific vocabulary corresponding to the model information to the candidate words of the target generation model.

[0165] The candidate words based on the updated target generation model can ensure that the semantic coherence of the generated target data is not destroyed when embedding the model information.

[0166] In the process of the target generation model performing the generation operation corresponding to the target input content, indirectly embedding the target watermark information or the user information and time information in its associated data into the generated target data may include:

[0167] In the process of the target generation model executing the generation operation corresponding to the target input content, the target data is generated by influencing the target generation model to predict the selection range of the next word based on the target watermark information or the user information and time information in its associated data.

[0168] For example, a mapping table may be constructed to record the mapping relationship between user information, time information and codes. The codes are used to represent the probability distribution of each type of word after the candidate words of the target generation model are classified.

[0169] For example, the user information includes a user ID of 220345. For example, the 2 in 220345 corresponds to the code 123, and the 0 in 220345 corresponds to another code 234. The target generation model first performs the generation operation corresponding to the target input content according to the code 123. 123 means that the probability of the first category of words is the highest, the probability of the second category of words is in the middle, and the probability of the third category of words is the lowest. When generating the first 100 words, if 10 words have been generated, when generating 11 words, the target generation model predicts the probability of each candidate word in the 30,000 candidate words, and then selects 6,000 words belonging to the first category from the 30,000 candidate words, and extracts one output as the 11th word based on the probability of the 6,000 words; when generating 12 words, the target generation model predicts the probability of each candidate word in the 30,000 candidate words. The target generation model predicts the probability of each candidate word in the 30,000 candidate words, and then selects 6,000 words belonging to the first category from the 30,000 candidate words, and extracts an output as the 12th word according to the probability of the 6,000 words; and so on, when 14 words are generated, the target generation model predicts the probability of each candidate word in the 30,000 candidate words, and then selects 6,000 words belonging to the first category from the 30,000 candidate words, and extracts an output as the 14th word according to the probability of the 6,000 words, thereby achieving 4 consecutive times of extracting an output from the first category of words.

[0170] Next, when the 15th word is generated, the target generation model predicts the probability of each candidate word among the 30,000 candidate words, and then selects 5,000 words belonging to the second category from the 30,000 candidate words, and extracts an output as the 15th word based on the probability of 5,000 words. Similarly, when the 17th word is generated, the target generation model predicts the probability of each candidate word among the 30,000 candidate words, and then selects 5,000 words belonging to the second category from the 30,000 candidate words, and extracts an output as the 17th word based on the probability of 5,000 words, thereby achieving three consecutive times of extracting an output from the words in the second category.

[0171] Next, when the 18th word is generated, the target generation model predicts the probability of each candidate word among the 30,000 candidate words, and then selects 8,000 words belonging to the third category from the 30,000 candidate words, and extracts an output as the 18th word based on the probability of the 8,000 words. Similarly, when the 19th word is generated, the target generation model predicts the probability of each candidate word among the 30,000 candidate words, and then selects 8,000 words belonging to the third category from the 30,000 candidate words, and extracts an output as the 19th word based on the probability of the 8,000 words, thereby achieving two consecutive extractions of an output from the words of the third category.

[0172] In the process of the target generation model performing the generation operation corresponding to the target input content, the target generation model can distinguish and insert the model information, user information and time information. The target generation model can embed the model information into the generated target data corresponding to the target input content at the initial stage of the process of performing the generation operation corresponding to the target input content. That is, the generation of the model information and the target data is performed simultaneously, rather than embedding the model information after the target data is generated.

[0173] The target generation module may embed the user information and the time information at a subsequent stage of the process of performing the generation operation corresponding to the target input content.

[0174] For example, a user inputs a text as the target input content: "Please generate a promotional poster for Lenovo Group's latest product for me." A target trigger event, i.e., a generated image event, is generated. At this time, the target watermark information can be obtained: model name "AI Poster Generator V1.0", user ID "220345", time "10.2". The target watermark information and the target input content are input into the target generation model that matches the generated image event. When the target generation model performs the generation operation corresponding to the target input content, at least part of the target watermark information or its associated data is embedded into the generated poster.

[0175] Step S1023: determine a target position in the initial generated data generated based on the target input content, and insert at least part of the target watermark information or its associated data into the corresponding target position to generate the target data.

[0176] The target location may be determined by, but is not limited to, at least one of the following methods:

[0177] Step S41: randomly selecting a position in the initially generated data as a target position based on a random number generator.

[0178] The random selection of the position based on the random number generator can ensure the randomness and unpredictability of the target watermark information insertion position, thereby increasing the concealment and security of the target watermark information.

[0179] Step S42: Analyze the content of the target watermark information or its associated data, and compare it with the initially generated data to find the parts with the same content or high similarity, and replace the parts with the same content or high similarity with the target watermark information to obtain the target data.

[0180] Through step S42, the consistency of the target data can be maintained and the watermark can be effectively hidden. It should be noted that when replacing, the replaced content will not change the semantics of the initially generated data or the main content of the image.

[0181] Step S43: When a specific word or symbol appears in the initially generated data, its position is determined as the insertion point of the target watermark information or its associated data.

[0182] In step S1023, by determining the target position, it can be ensured that the insertion of the target watermark information should not cause confusion in the text semantics or obvious changes in the image content, so as to ensure that the target watermark information can realize its traceability function without affecting the normal use of the data. For example, the target watermark information may include: Lenovo all. The initial generated data generated based on the target input content (i.e., "generate a text about the impact between the Internet and life") may include the following text: "In today's digital age, the Internet has penetrated into all aspects of our lives, whether it is work, study, entertainment, or social interaction, the Internet plays an indispensable role. ... In addition, through the Internet, we can meet more people, expand our social circles, and communicate, learn, and progress with like-minded friends. However, the Internet also brings some negative effects. First of all, ..., sometimes we may be hurt because of trusting people on the Internet, leading to a breakdown in trust relationships."

[0183] Determine the target position in the above-mentioned initial generated data, insert all associations into the corresponding target position, and the generated target data may include: "In today's digital age, the Internet has penetrated into all aspects of our lives, whether it is work, study, entertainment, or social interaction, network connections play an indispensable role. ... In addition, through the Internet, we can get to know more people, expand our social circles, and communicate, learn, and make progress with like-minded friends. However, the Internet also brings some negative effects. First of all, ..., sometimes we may be hurt because of trusting people on the Internet, resulting in a breakdown of trust. ... These problems have prompted us to think about how to better use the Internet to maintain and develop healthy interpersonal relationships. In short, the Internet helps us expand our interpersonal relationships, but it also brings challenges. We need to pay attention to and maintain interpersonal relationships in real life while taking advantage of its convenience."

[0184] In the target data, association does not lead to confusion of text semantics and semantic coherence, and the target watermark information can realize the traceability function.

[0185] In this embodiment, it is possible to flexibly determine how much of the target watermark information or its associated data to insert into the target location based on factors such as the type of initially generated data, the size of the data, and the application scenario.

[0186] For example, different types of initial generated data have different requirements for embedding watermark information:

[0187] Text data: In text data, according to the content and format of the text, you can choose to insert the target watermark information at the beginning, end, between paragraphs or in a specific position of the text. In addition, you also need to consider factors such as the language characteristics and encoding method of the text to ensure that the insertion of the target watermark information does not destroy the grammar and semantics of the text.

[0188] Image data: In image data, the target watermark information is usually embedded in the form of image elements. According to the image resolution, color depth and other characteristics, you can choose to insert the target watermark information in a specific area of ​​the image (such as corners, edges or background), or use the image's texture, color and other characteristics to insert the target watermark information.

[0189] Audio / video data: In audio or video data, the target watermark information is usually inserted in the form of a slight change in the audio signal, a pixel change in the video frame, or a specific encoding mode. The watermark insertion of this type of data needs to consider factors such as the audio or video compression format, bit rate, frame rate, etc. to ensure that the embedding of the watermark information does not significantly affect the quality of the audio or video.

[0190] When the amount of initially generated data is large, the target watermark information can be distributed and inserted in multiple locations to increase the concealment and robustness of the watermark.

[0191] When the amount of initially generated data is small, the insertion position and the amount of target watermark information need to be selected more carefully. Too much watermark information may cause data damage or distortion, so it is necessary to weigh and compromise according to the actual situation.

[0192] The application scenario is also an important factor in determining the amount of insertion. For example, in the copyright protection scenario, the watermark information needs to be sufficiently concealed and robust to deal with possible tampering and attacks. Therefore, it is necessary to select an appropriate insertion position and insertion amount to ensure that the watermark information can effectively play a role when needed. Alternatively, in the data tracking scenario, the watermark information needs to be able to accurately identify the source and flow of the data. Therefore, it is necessary to select an appropriate insertion method and insertion amount based on the characteristics of the data and the application scenario to ensure that the watermark information can effectively convey the required information.

[0193] As another optional embodiment of the present application, a data generation method is provided in Embodiment 6 of the present application. This embodiment is mainly an implementation method of enabling target watermark information to be directly or indirectly embedded into target data in Embodiment 1 or 5. The method of enabling target watermark information to be directly or indirectly embedded into target data may include but is not limited to at least one of the following:

[0194] Step S51, obtaining attribute information of the target data, and embedding the target watermark information or its associated data into the target data with a target embedding strategy based on the attribute information, wherein the amount of target watermark information or its associated data embedded into the target data under different embedding strategies is different.

[0195] The attribute information of the target data may characterize the type (eg, text, image, audio, video, etc.) and size (eg, the length of text data or the number of image elements of image data) of the target data.

[0196] Based on the attribute information of the target data, a suitable target embedding strategy can be selected to embed the target watermark information. For example, in long text data (the length of the text data meets the set threshold, the text data contains multiple paragraphs, sentences, etc.), select a suitable embedding position to embed the target watermark information, such as the beginning of a paragraph, the end of a paragraph, between sentences, or near specific keywords. Selecting a suitable embedding position can ensure the concealment and robustness of the watermark. For example, one or several watermark characters can be inserted at the end of each paragraph or in a specific sentence, so that even if the text is partially copied or modified, the target watermark information may still be retained.

[0197] Alternatively, in order to enhance concealment, the target watermark information can be dispersed to multiple locations in the long text data. This can be achieved by gradually inserting the target watermark information into different paragraphs or sentences of the long text data. For example, the target watermark information can be divided into multiple parts, each of which is embedded in a different paragraph of the long text data. In this way, even if a paragraph is deleted or modified, the watermark information parts in other paragraphs can still exist.

[0198] Step S52: Obtain usage information of the target data, and embed the target watermark information or the target part in its associated data into the target data based on the usage information, where the target part is the part that matches the usage information.

[0199] Usage information can be used to characterize the intended use or usage scenario of the target data. For example, usage information can characterize whether the target data is used for personal learning (self-use scenario) or for sharing with others (sharing scenario), whether it is used for making summary reports or for content expansion, whether it is presented in formal meetings or shared on social media, etc.

[0200] For different types of target data (such as narrative text, conversational text, etc.), different embedding strategies can be used to adapt to their specific structure and content. For example, if the target data is used to make a summary report, the type of the target data is narrative text, and the target watermark information is embedded in the target data in the form of academic words or phrases to increase the concealment of the watermark information. In addition, if the target data is long text data, the target watermark information can be dispersed to multiple locations in the long text data.

[0201] Alternatively, if the target data is used for sharing in conversations on social media, the target watermark information can be embedded in the target data in the form of inserting conversations, comments, or emoticons, etc., to ensure the concealment of the watermark information.

[0202] Step S53: Obtain the authorization information of the target user, and determine to embed the target watermark information or the associated data of the target watermark information into the target data based on the authorization information.

[0203] In this embodiment, the target user can be the user who triggers the target trigger time or inputs the target input content.

[0204] The authorization information can indicate whether to allow the explicit display of the watermark information.

[0205] If the authorization information of the target user indicates that it allows the explicit display of the watermark information, the target watermark information can be clearly displayed in the generated target data. For example, if the target watermark information is "Lenovo All", as Figure 5 shown, the four characters "Lenovo All" can be displayed in bold and underlined to ensure that users can easily identify the target watermark information.

[0206] If the authorization information of the target user indicates that it does not allow the explicit display of the target watermark information, the invisible watermark scheme can be default selected, that is, the target watermark information is not prominently displayed in the generated target data. For example, if the target watermark information is "Lenovo All", as Figure 6 shown, the four characters "Lenovo All" can be embedded in the same format as other text data, so that ordinary users cannot directly identify the target watermark information.

[0207] As another optional embodiment of the present application, referring to Figure 7 , it is a schematic flowchart of a data generation method provided in Embodiment 7 of the present application. As Figure 7 shown, the method may include but is not limited to the following steps:

[0208] Step S201: In response to a target trigger event, obtain target watermark information.

[0209] Among them, the target watermark information is used to identify the source of the target data, and the watermark information embedded in the target data corresponding to different trigger events is different.

[0210] Step S202: Generate target data based on the target watermark information and the target input content, so that the target watermark information is directly or indirectly embedded in the target data.

[0211] The detailed process of steps S201-S202 can refer to the relevant introduction of steps S101-S102 in Example 1, which will not be repeated here.

[0212] Step S203: After the target data is generated, quality inspection is performed on the target data.

[0213] If the result of the quality detection indicates that the quality of the target data does not meet the set conditions (for example, the semantic coherence and readability of the text data do not meet the coherence and readability set conditions, and the display layout and hierarchical relationship of the image data do not meet the display layout and hierarchical relationship set conditions), the target generation model can be adjusted and the target data can be regenerated based on the adjusted target generation model.

[0214] For example, after the text data is generated, the text data can be tested for semantic coherence and readability to ensure that the text data after embedding the target watermark information still maintains the original semantic integrity and logical order, and there is no semantic break or ambiguity caused by watermark embedding. At the same time, the readability of the text data also needs to be guaranteed, that is, the text data after embedding the target watermark information can maintain a clear font, appropriate font size and spacing, and an easy-to-read format.

[0215] In this embodiment, it is also possible to detect whether the text data after the target watermark information is embedded still conforms to the user's intention represented by the target input content. For example, if the user input text (i.e., the target input content) expresses a specific viewpoint or emotion, it is possible to detect whether the text data after the target watermark information is embedded changes the expression of these viewpoints or emotions.

[0216] After the image data is generated, the display layout and hierarchical relationship of the image data can be detected to ensure that the image data after embedding the target watermark information still maintains the original image structure and layout, and the embedding of the target watermark information does not destroy the integrity or hierarchy of the image data. At the same time, the position and size of the target watermark information can also be reasonably controlled to ensure that it does not affect the visual effect of the image data and can play a role in identification.

[0217] In this embodiment, it is also possible to detect whether the image data after embedding the target watermark information maintains the original image quality and clarity, so as to ensure that the embedding of the target watermark information will not cause the image to be blurred, distorted or have color deviation.

[0218] As another optional embodiment of the present application, refer to Figure 8 , is a flow chart of a data generation method provided in Example 8 of the present application, such as Figure 8 As shown, the method may include but is not limited to the following steps:

[0219] Step S301: In response to a target triggering event, obtain target watermark information.

[0220] The target watermark information is used to identify the source of the target data, and the watermark information embedded in the target data corresponding to different triggering events is different.

[0221] Step S302: Generate target data based on the target watermark information and the target input content, so that the target watermark information is directly or indirectly embedded in the target data.

[0222] The detailed process of steps S301-S302 can refer to the relevant introduction of steps S101-S102 in Example 1, which will not be repeated here.

[0223] Step S303: After generating the target data, verify the watermark information embedded in the target data.

[0224] In this embodiment, a watermark list may be constructed according to the target watermark information, and the watermark list may include all watermark information that should appear in the target data and their expected order.

[0225] The watermark information embedded in the target data is matched with the watermark list. If the matching degree reaches a preset threshold, it can be determined that the target data contains valid watermark information.

[0226] In this embodiment, a reasonable fault tolerance threshold may be set according to the importance of the target watermark information and the characteristics of the data. The fault tolerance threshold may represent the maximum proportion of watermark information mismatch allowed.

[0227] In the process of matching the watermark information embedded in the target data with the watermark list, if it is found that the mismatch degree of the watermark information is within the tolerance threshold, the watermark is still determined to be valid.

[0228] The fault tolerance threshold can also be adjusted dynamically according to the needs of actual application scenarios. For example, in scenarios where data transmission security is required to be high, the fault tolerance threshold can be lowered to increase the strictness of watermark verification; in scenarios where data transmission efficiency is required to be high, the fault tolerance threshold can be appropriately increased to improve verification efficiency at the expense of a certain degree of security.

[0229] In this embodiment, taking into account possible damage or changes (such as character loss, replacement, insertion, etc.) during text transmission, a fault tolerance threshold can be set to allow the validity of the watermark to be verified even when a certain proportion of the watermark information is missing or changed.

[0230] In this embodiment, the watermark information is used as a data identifier, and its integrity is directly related to the source confirmation and authenticity assurance of the data. By verifying the watermark information in the target data, it can be ensured that the target data has not been tampered with or damaged during transmission or processing, thereby maintaining the integrity and credibility of the data.

[0231] In this embodiment, the following steps may also be performed after steps S301-S302:

[0232] After the target data is generated, the target data is quality checked.

[0233] After the target data is generated, the detailed process of performing quality inspection on the target data can be found in the related introduction of the above step S203, which will not be repeated here.

[0234] As another optional embodiment of the present application, refer to Fig. 9 , is a flow chart of a data generation method provided in Example 9 of the present application, such as Fig. 9 As shown, the method may include but is not limited to the following steps:

[0235] Step S401: In response to a target triggering event, obtain target watermark information.

[0236] The target watermark information is used to identify the source of the target data, and the watermark information embedded in the target data corresponding to different triggering events is different.

[0237] Step S402: Generate target data based on the target watermark information and the target input content, so that the target watermark information is directly or indirectly embedded in the target data.

[0238] The detailed process of steps S401-S402 can refer to the relevant introduction of steps S101-S102 in Example 1, which will not be repeated here.

[0239] Step S403: In response to obtaining a traceability request for target data, extract target watermark information from the target data.

[0240] If the target watermark information is embedded in the target data after being encrypted, the target watermark information extracted from the target data is the encrypted target watermark information.

[0241] In this embodiment, extracting target watermark information from target data may include but is not limited to:

[0242] Step S4031, extract model information from target data.

[0243] Step S4032: construct a watermark list according to the model information of the target generation model in the target watermark information. The watermark list may include all the model information that should appear in the target data and their expected order.

[0244] Step S4033: Match the extracted model information with the watermark list. If the matching degree reaches a preset threshold, it can be determined that the target data contains valid model information, and the user information and time information are extracted from the target data.

[0245] Corresponding to an implementation method of generating target data by influencing the target generation model to predict the next word selection range based on the target watermark information or the user information and time information in its associated data during the target generation model's execution of the generation operation of the corresponding target input content, when extracting user information and time information from the target data, statistics can be performed on various types of words in the target data to obtain statistical results, and the encoding that matches the statistical results can be searched in the mapping table to obtain the user information and / or time information corresponding to the encoding in the mapping table.

[0246] Step S404: decrypt the extracted target watermark information, and obtain the tracing result of the target data based on the decryption result.

[0247] In this embodiment, the extracted model information, user information and time information may be decrypted to obtain the model information, user information and time information.

[0248] The traceability result of the target data can clearly indicate the model name and / or version number that generated the target data, as well as the user name and user ID that generated the target data, and display the generation time of the target data, which may include a date and timestamp.

[0249] In this embodiment, the tracing result may be presented to the user or administrator in a visual form or a text form so that they can understand the source of the target data.

[0250] Of course, each traceability result can also be recorded in the system log for subsequent audit or analysis.

[0251] In this embodiment, by decrypting the extracted watermark information, the original watermark information content can be restored, thereby accurately revealing the source of the target data, so that the target data can be effectively tracked and located during the circulation process, thereby improving the traceability and manageability of the target data.

[0252] Moreover, during the data generation and tracing process, encryption and decryption technologies can be used to ensure the privacy and security of target data, thus avoiding the leakage and abuse of target data.

[0253] In addition, by extracting user information and time information, dual traceability of time and events can be achieved, that is, not only knowing who generated the data, but also knowing when the data was generated.

[0254] As another optional embodiment of the present application, refer to Fig.10 , is a flow chart of a text generation method provided in Example 10 of the present application, such as Fig.10 As shown, the method may include but is not limited to the following steps:

[0255] Step S501: in response to obtaining a target text generation requirement, obtaining target watermark information, the target watermark information including first character string data and / or second character string data.

[0256] With reference to the introduction in the foregoing embodiments, the model information of the target generation model may be processed into first string data, and at least one of the user information and the time information associated with the target triggering event may be processed into second string data.

[0257] Step S502: Generate target text content based on the target watermark information and the target input content, so that the first character string data and / or the second character string data are directly or indirectly embedded in the target text content.

[0258] The detailed process of step S502 can refer to the related introduction of generating target data based on target watermark information and target input content in the aforementioned embodiments so that the target watermark information is directly or indirectly embedded into the target data, which will not be repeated here.

[0259] For example, the target input content can be used to call the text generation model of the electronic device. Accordingly, the target watermark information and the target input content can be input into the text generation model, and in the process of the text generation model performing the generation operation corresponding to the target input content, at least part of the target watermark information or its associated data is embedded into the generated target text content.

[0260] The target watermark information is used to identify the source of the target text, and the watermark information embedded in the target text content corresponding to different text generation requirements is different.

[0261] In this embodiment, embedding is controlled by target input content so that the first character string data and / or the second character string data are directly or indirectly embedded into the target text content, ensuring that the target watermark information is difficult to detect or remove without affecting the natural fluency of the target text content.

[0262] By directly or indirectly embedding the first character string data and / or the second character string data into the target text content, not only can the text generation model be identified, but also the user and time of the specific generated text can be tracked, and it has a strong traceability capability.

[0263] Furthermore, the second string data is dynamic and will change according to the user and time. Therefore, the target watermark information generated each time is unique, which makes the target watermark information difficult to be predicted or copied, thereby greatly improving the security of the target watermark information.

[0264] Next, the data generating device provided by the present application is introduced. The data generating device introduced below and the data generating method introduced above can refer to each other.

[0265] Reference Fig.11 , the data generating device includes: an obtaining module 100 and a generating module 200.

[0266] An acquisition module 100, for acquiring target watermark information in response to a target triggering event;

[0267] A generating module 200, configured to generate target data based on the target watermark information and target input content, so that the target watermark information is directly or indirectly embedded in the target data;

[0268] The target watermark information is used to identify the source of the target data, and the watermark information embedded in the target data corresponding to different triggering events is different.

[0269] The target trigger event can be generated by at least one of the following:

[0270] In response to monitoring first input content input into the first application, generating the target trigger event;

[0271] generating the target triggering event in response to obtaining second input content, wherein the second input content is used to call a target generation model of the electronic device;

[0272] In response to obtaining third input content, generating the target trigger event, wherein the third input content can trigger the electronic device to perform an operation of adding watermark information;

[0273] In response to monitoring the operation of sharing the target file, the target trigger event is generated.

[0274] The obtaining module 100 obtains the target watermark information, which may specifically include at least one of the following:

[0275] Obtaining source information of the target triggering event, and obtaining target watermark information matching the source information based on the source information;

[0276] Obtaining user information and / or model information associated with the target triggering event, and generating the target watermark information based on the user information and / or model information;

[0277] A user intention represented by the target input content is obtained, and target watermark information matching the user intention is obtained based on the user intention.

[0278] The obtaining module 100 generates the target watermark information based on the user information and / or the model information, which may specifically include at least one of the following:

[0279] Processing the model information of the target generation model into first character string data and / or a first image element set as the target watermark information;

[0280] Processing user information associated with the target triggering event into second string data and / or a second image element set as the target watermark information;

[0281] Processing the user information and time information associated with the target triggering event into a second string data and / or a second image element set as the target watermark information;

[0282] Processing model information of a target generation model into first string data and / or a first image element set, processing user information associated with a target triggering event into second string data and / or a second image element set, and processing the first string data and / or the first image element set, and the second string data and / or the second image element set into target watermark information;

[0283] The model information of the target generation model is processed into a first string data and / or a first image element set, the user information and time information associated with the target trigger event is processed into a second string data and / or a second image element set, and the first string data and / or the first image element set, as well as the second string data and / or the second image element set, are processed into target watermark information.

[0284] The generating module 200 generates target data based on the target watermark information and the target input content, which may specifically include at least one of the following:

[0285] Identify the initial user intention represented by the target input content, use the target watermark information or its associated data to update the initial user intention to obtain the target user intention, and generate target data matching the target user intention;

[0286] Inputting the target watermark information and the target input content into a target generation model matching the target trigger event, and in the process of the target generation model performing a generation operation corresponding to the target input content, embedding at least part of the target watermark information or its associated data into the generated target data;

[0287] A target position is determined in the initial generated data generated based on the target input content, and at least part of the target watermark information or its associated data is inserted into the corresponding target position to generate the target data.

[0288] In this embodiment, the generation module 200 enables the target watermark information to be directly or indirectly embedded into the target data, and specifically may include at least one of the following:

[0289] Obtaining attribute information of the target data, and embedding the target watermark information or its associated data into the target data according to a target embedding strategy based on the attribute information, wherein the amount of the target watermark information or its associated data embedded into the target data under different embedding strategies is different;

[0290] Obtaining usage information of the target data, and embedding a target portion of the target watermark information or its associated data into the target data based on the usage information, wherein the target portion is a portion matching the usage information;

[0291] The authorization information of the target user is obtained, and based on the authorization information, it is determined to embed the target watermark information or the associated data of the target watermark information into the target data.

[0292] The data generating device may further include at least one of the following:

[0293] A detection module, used for performing quality detection on the target data after the target data is generated;

[0294] The verification module is used to verify the watermark information embedded in the target data after the target data is generated.

[0295] The data generating device may further include:

[0296] An extraction module, configured to extract target watermark information from the target data in response to obtaining a traceability request for the target data;

[0297] The tracing module is used to decrypt the extracted target watermark information and obtain the tracing result of the target data based on the decryption result.

[0298] In another embodiment of the present application, a text generation device is provided. The text generation device and the text generation method introduced above can refer to each other.

[0299] The text generating means may include:

[0300] An obtaining module, configured to obtain target watermark information in response to a target text generation requirement, wherein the target watermark information includes first character string data and / or second character string data;

[0301] A generating module, configured to generate target text content based on the target watermark information and target input content, so that the first character string data and / or the second character string data are directly or indirectly embedded into the target text content;

[0302] The target watermark information is used to identify the source of the target text, and the watermark information embedded in the target text content corresponding to different text generation requirements is different.

[0303] In another embodiment of the present application, an electronic device is provided.

[0304] Electronic equipment may include:

[0305] The memory is used to store computer programs;

[0306] The processor is used to execute the computer program so that the electronic device can implement the data generation method introduced in any one of Examples 1-9 or the text generation method introduced in Example 10.

[0307] A computer-readable storage medium is also provided in an embodiment of the present application. The storage medium carries one or more computer programs. When the one or more computer programs are executed by an electronic device, the electronic device can implement the data generation method described in any one of Embodiments 1-9 or the text generation method described in Embodiment 10.

[0308] It should also be noted that the device embodiments described above are merely schematic, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed over multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. In addition, in the drawings of the device embodiments provided by the present application, the connection relationship between the modules indicates that there is a communication connection between them, which may be specifically implemented as one or more communication buses or signal lines.

[0309] Through the description of the above implementation mode, the technicians in the field can clearly understand that the present application can be implemented by means of software plus necessary general hardware, and of course, it can also be implemented by special hardware including special integrated circuits, special CPUs, special memories, special components, etc. In general, all functions completed by computer programs can be easily implemented by corresponding hardware, and the specific hardware structure used to implement the same function can also be various, such as analog circuits, digital circuits or special circuits. However, for the present application, software program implementation is a better implementation mode in more cases. Based on such an understanding, the technical solution of the present application is essentially or the part that contributes to the prior art can be embodied in the form of a software product, which is stored in a readable storage medium, such as a computer floppy disk, a U disk, a mobile hard disk, a ROM, a RAM, a disk or an optical disk, etc., including a number of instructions to enable a computer device (which can be a personal computer, a training device, or a network device, etc.) to execute the methods described in each embodiment of the present application.

[0310] In the above embodiments, all or part of the embodiments may be implemented by software, hardware, firmware or any combination thereof. When implemented by software, all or part of the embodiments may be implemented in the form of a computer program product.

[0311] The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, the process or function described in the embodiment of the present application is generated in whole or in part. The computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices. The computer instructions may be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions may be transmitted from a website site, a computer, a training device, or a data center by wired (e.g., coaxial cable, optical fiber, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) mode to another website site, computer, training device, or data center. The computer-readable storage medium may be any available medium that a computer can store or a data storage device such as a training device, a data center, etc. that includes one or more available media integrations. The available medium may be a magnetic medium, (e.g., a floppy disk, a hard disk, a tape), an optical medium (e.g., a DVD), or a semiconductor medium (e.g., a solid-state drive (SSD)), etc.

Claims

1. A data generation method, comprising: In response to a target triggering event, obtaining target watermark information; generating target data based on the target watermark information and the target input content, so that the target watermark information is directly or indirectly embedded in the target data; The target watermark information is used to identify the source of the target data, and the watermark information embedded in the target data corresponding to different triggering events is different.

2. The method according to claim 1, wherein: The target trigger event is generated by at least one of the following: In response to monitoring first input content input into the first application, generating the target trigger event; generating the target triggering event in response to obtaining second input content, wherein the second input content is used to call a target generation model of the electronic device; In response to obtaining third input content, generating the target trigger event, wherein the third input content can trigger the electronic device to perform an operation of adding watermark information; In response to monitoring the operation of sharing the target file, the target trigger event is generated.

3. The method according to claim 1, wherein: Obtaining target watermark information includes at least one of the following: Obtaining source information of the target triggering event, and obtaining target watermark information matching the source information based on the source information; Obtaining user information and / or model information associated with the target triggering event, and generating the target watermark information based on the user information and / or model information; A user intention represented by the target input content is obtained, and target watermark information matching the user intention is obtained based on the user intention.

4. The method according to claim 3, wherein: Generating the target watermark information based on the user information and / or the model information includes at least one of the following: Processing the model information of the target generation model into first character string data and / or a first image element set as the target watermark information; Processing user information associated with the target triggering event into second string data and / or a second image element set as the target watermark information; Processing the user information and time information associated with the target triggering event into a second string data and / or a second image element set as the target watermark information; Processing model information of a target generation model into first string data and / or a first image element set, processing user information associated with a target triggering event into second string data and / or a second image element set, and processing the first string data and / or the first image element set, and the second string data and / or the second image element set into target watermark information; The model information of the target generation model is processed into a first string data and / or a first image element set, the user information and time information associated with the target trigger event is processed into a second string data and / or a second image element set, and the first string data and / or the first image element set, as well as the second string data and / or the second image element set, are processed into target watermark information.

5. The method according to claim 1, wherein: Generating target data based on the target watermark information and the target input content includes at least one of the following: Identify the initial user intention represented by the target input content, use the target watermark information or its associated data to update the initial user intention to obtain the target user intention, and generate target data matching the target user intention; Inputting the target watermark information and the target input content into a target generation model matching the target trigger event, and in the process of the target generation model performing a generation operation corresponding to the target input content, embedding at least part of the target watermark information or its associated data into the generated target data; A target position is determined in the initial generated data generated based on the target input content, and at least part of the target watermark information or its associated data is inserted into the corresponding target position to generate the target data.

6. The method according to claim 1 or 5, wherein: So that the target watermark information is directly or indirectly embedded into the target data, including at least one of the following: Obtaining attribute information of the target data, and embedding the target watermark information or its associated data into the target data according to a target embedding strategy based on the attribute information, wherein the amount of the target watermark information or its associated data embedded into the target data under different embedding strategies is different; Obtaining usage information of the target data, and embedding a target portion of the target watermark information or its associated data into the target data based on the usage information, wherein the target portion is a portion matching the usage information; The authorization information of the target user is obtained, and based on the authorization information, it is determined to embed the target watermark information or the associated data of the target watermark information into the target data.

7. The method according to claim 1, further comprising at least one of the following: After generating the target data, performing quality inspection on the target data; After the target data is generated, the watermark information embedded in the target data is verified.

8. The method according to claim 1, further comprising: In response to obtaining a traceability request for the target data, extracting target watermark information in the target data; The extracted target watermark information is decrypted, and a tracing result of the target data is obtained based on the decryption result.

9. A text generation method, comprising: In response to obtaining a target text generation requirement, obtaining target watermark information, wherein the target watermark information includes first character string data and / or second character string data; generating target text content based on the target watermark information and the target input content, so that the first character string data and / or the second character string data are directly or indirectly embedded into the target text content; The target watermark information is used to identify the source of the target text, and the watermark information embedded in the target text content corresponding to different text generation requirements is different.

10. An electronic device comprising: The memory is used to store computer programs; The processor is used to execute the computer program so that the electronic device can implement the data generation method as described in any one of claims 1 to 8 or the text generation method as described in claim 9.

Images