Data processing method and system, electronic device and storage medium

Abstract

The application discloses a data processing method and system, electronic equipment and storage medium, and relates to the technical field of large models, artificial intelligence and data security. The method comprises the following steps: verifying first reply information displayed in a generative interaction interface to obtain a first verification result; in the case that the first verification result is that the content of the first reply information is unsafe, and the first reply information is information generated by a generative model based on first inquiry information, determining the first inquiry information based on data stored in a block chain; and adjusting original training data of the generative model based on the first inquiry information to obtain target training data, wherein the target training data is used to adjust model parameters of the generative model. The application solves the technical problem of low safety of content generation by using a large model.

Description

Technical Field

[0001] This application relates to the fields of large model technology, artificial intelligence, and data security. Specifically, it relates to a data processing method, system, electronic device, and storage medium. Background Technology

[0002] The application of large models for content generation is becoming increasingly widespread, and data security issues during the content generation process have also received widespread attention. Because content generation using large models relies on both user-input queries and the large model itself, when data security issues arise, it is difficult to accurately pinpoint the root cause of the problem, resulting in relatively low security for content generation using large models.

[0003] There is currently no effective solution to the above problems. Summary of the Invention

[0004] This application provides a data processing method, system, electronic device, and storage medium to at least address the technical problem of low security when generating content using large models.

[0005] According to one aspect of the embodiments of this application, a data processing method is provided, comprising: verifying first response information displayed in a generative interactive interface to obtain a first verification result, wherein the first verification result is used to characterize whether the content of the first response information is secure, and the first response information is used to characterize information responding to a first query information input in the generative interactive interface; if the first verification result indicates that the content of the first response information is insecure, and the first response information is information generated by a generative model based on the first query information, determining the first query information based on data stored in a blockchain, wherein the blockchain is used to store data corresponding to different query information and data corresponding to corresponding response information, and the corresponding response information is information generated by a generative model based on different query information; and adjusting the original training data of the generative model based on the first query information to obtain target training data, wherein the target training data is used to adjust the model parameters of the generative model.

[0006] According to another aspect of the embodiments of this application, a data processing method is also provided, including: acquiring first query information displayed in a generative interactive interface; generating first response information based on the first query information using a generative model; displaying the first response information in the generative interactive interface; and storing the data corresponding to the first query information and the data corresponding to the first response information in a blockchain, wherein the blockchain is used to store data corresponding to different query information and data corresponding to the corresponding response information, and the corresponding response information is information generated based on different query information using a generative model.

[0007] According to another aspect of the embodiments of this application, a data processing method is also provided, comprising: displaying first response information in a generative interactive interface; displaying a first verification result on first query information in response to a verification instruction applied to the generative interactive interface, wherein the first verification result is used to characterize whether the content of the first response information is secure, and the first response information is used to characterize information responding to the first query information input in the generative interactive interface; and displaying target training data of a generative model on the first query information in response to an adjustment instruction applied to the generative interactive interface, wherein the target training data is data obtained by adjusting the original training data of the generative model based on the first query information, the first query information is information determined based on data stored in a blockchain when the first verification result indicates that the content of the first response information is insecure, and the first response information is information generated by the generative model based on the first query information, the blockchain is used to store data corresponding to different query information and data corresponding to the corresponding response information, and the corresponding response information is information generated by the generative model based on different query information.

[0008] According to another aspect of the embodiments of this application, a data processing method is also provided, comprising: responding to an input instruction applied to a generative interactive interface, displaying first query information on the generative interactive interface; responding to a generation instruction applied to the first query information, displaying first response information on the first query information, wherein the first response information is information generated based on the first query information using a generative model, and the data corresponding to the first query information and the data corresponding to the first response information are stored in a blockchain, the blockchain being used to store data corresponding to different query information and data corresponding to corresponding response information, and the corresponding response information being information generated based on different query information using a generative model.

[0009] According to another aspect of the embodiments of this application, a data processing method is also provided, comprising: obtaining first response information displayed in a generative interactive interface by calling a first interface, wherein the first interface includes a first parameter, and the parameter value of the first parameter is the first response information; verifying the first response information to obtain a first verification result, wherein the first verification result is used to characterize whether the content of the first response information is secure, and the first response information is used to characterize the information responding to the first query information input in the generative interactive interface; in the case that the first verification result indicates that the content of the first response information is insecure, and the first response information is information generated by a generative model based on the first query information, determining the first query information based on data stored in a blockchain, wherein the blockchain is used to store data corresponding to different query information and data corresponding to corresponding response information, and the corresponding response information is information generated by a generative model based on different query information; adjusting the original training data of the generative model based on the first query information to obtain target training data, wherein the target training data is used to adjust the model parameters of the generative model; and outputting the target training data by calling a second interface, wherein the second interface includes a second parameter, and the parameter value of the second parameter is the target training data.

[0010] According to another aspect of the embodiments of this application, a data processing method is also provided, comprising: inputting multimodal information in a dialogue interface, wherein the type of multimodal information includes at least one of the following: text information containing character information, video frame information containing frame image information, and audio information; generating response information corresponding to the multimodal information based on the multimodal information using a generative model, wherein the type of response information includes at least one of the following: text information, image information, video information, and voice information; storing the data corresponding to the multimodal information and the data corresponding to the response information in a blockchain, wherein the blockchain is used to store data corresponding to different query information and data corresponding to the corresponding response information, and the corresponding response information is information generated by the generative model based on different query information.

[0011] According to another aspect of the embodiments of this application, a data processing system is also provided, including: a front-end client for displaying a generative interactive interface and capturing first query information input in the generative interactive interface; a back-end server connected to the front-end client for generating first response information based on the first query information using a generative model; and a blockchain connected to the back-end server for storing data corresponding to the first query information and data corresponding to the first response information.

[0012] According to another aspect of the embodiments of this application, an electronic device is also provided, including: a memory storing an executable program; and a processor for running the program, wherein the program executes any of the methods described above when it runs.

[0013] According to another aspect of the embodiments of this application, a computer-readable storage medium is also provided, the computer-readable storage medium including a stored executable program, wherein, when the executable program is running, the device where the computer-readable storage medium is located executes any of the methods described above.

[0014] In this embodiment, a first verification result is obtained by verifying the first response information displayed in the generative interactive interface. The first verification result is used to characterize whether the content of the first response information is safe, and the first response information is used to characterize the information that responds to the first query information input in the generative interactive interface. If the first verification result indicates that the content of the first response information is unsafe, and the first response information is generated by the generative model based on the first query information, the first query information is determined based on the data stored in the blockchain. The blockchain is used to store data corresponding to different query information and data corresponding to the corresponding response information, and the corresponding response information is generated by the generative model based on different query information. Based on the first query information, the original training data of the generative model is adjusted to obtain target training data, whereby the target training data is used to adjust the model parameters of the generative model. It is noteworthy that by using blockchain to store the data corresponding to different query information and the corresponding response information, when it is determined that the content of the first response information is insecure, and the first response information is generated by the generative model based on the first query information, the reason for the insecurity of the content of the first response information can be accurately located based on the first query information, and the parameters of the generative model can be adjusted. This improves the security of content generation using large models and solves the technical problem of low security when using large models for content generation.

[0015] It is worth noting that the general description above and the detailed description that follow are merely for illustrative purposes and do not constitute a limitation on this application. Attached Figure Description

[0016] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments of this application and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings:

[0017] Figure 1 This is a schematic diagram of an optional data processing method according to an embodiment of this application;

[0018] Figure 2 This is a flowchart of the data processing method according to Embodiment 1 of this application;

[0019] Figure 3 This is a schematic diagram of a data processing method according to an embodiment of this application;

[0020] Figure 4 This is a flowchart of the data processing method according to Embodiment 2 of this application;

[0021] Figure 5 This is a schematic diagram of information interaction according to Embodiment 2 of this application;

[0022] Figure 6 This is a flowchart of the data processing method according to Embodiment 3 of this application;

[0023] Figure 7 This is a flowchart of the data processing method according to Embodiment 4 of this application;

[0024] Figure 8 This is a flowchart of the data processing method according to Embodiment 5 of this application;

[0025] Figure 9 This is a flowchart of the data processing method according to Embodiment 6 of this application;

[0026] Figure 10 This is a schematic diagram of the data processing system according to Embodiment 7 of this application;

[0027] Figure 11 This is a schematic diagram of a data processing apparatus according to Embodiment 8 of this application;

[0028] Figure 12 This is a schematic diagram of a data processing apparatus according to Embodiment 9 of this application;

[0029] Figure 13 This is a schematic diagram of a data processing apparatus according to Embodiment 10 of this application;

[0030] Figure 14 This is a schematic diagram of a data processing apparatus according to Embodiment 11 of this application;

[0031] Figure 15 This is a schematic diagram of a data processing apparatus according to Embodiment 12 of this application;

[0032] Figure 16 This is a schematic diagram of a data processing apparatus according to Embodiment 13 of this application;

[0033] Figure 17 This is a structural block diagram of a computer terminal according to an embodiment of this application. Detailed Implementation

[0034] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present application, and not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative effort should fall within the scope of protection of the present application.

[0035] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0036] The technical solution provided in this application is mainly implemented using large-scale model technology. Here, "large-scale model" refers to a deep learning model with a massive number of parameters, typically containing hundreds of millions, tens of billions, hundreds of billions, trillions, or even tens of trillions of parameters. Large-scale models can also be called foundational models. They are pre-trained using large-scale unlabeled corpora to produce pre-trained models with hundreds of millions of parameters. Such models can adapt to a wide range of downstream tasks and have good generalization ability. Examples include Large Language Models (LLMs) and multimodal pre-training models.

[0037] It should be noted that, in practical applications, large models can be fine-tuned using a small number of samples to adapt them to different tasks. For example, large models can be widely used in Natural Language Processing (NLP), computer vision, and speech processing. Specifically, they can be applied to computer vision tasks such as Visual Question Answering (VQA), Image Captioning (IC), and Image Generation, as well as NLP tasks such as text-based sentiment classification, text summarization, and machine translation. Therefore, the main application scenarios for large models include, but are not limited to, digital assistants, intelligent robots, search, online education, office software, e-commerce, and intelligent design. In this embodiment, the example of generating corresponding response information based on user queries in a human-computer interaction scenario is used for explanation.

[0038] First, some nouns or terms that appear in the description of the embodiments of this application shall be interpreted as follows:

[0039] Generative interactive interface: This can be an interactive interface that generates natural language text through machine learning algorithms. Optionally, the generative interactive interface can generate relevant text responses based on user input, and can be used in chatbots, intelligent assistants, question answering systems and other related scenarios.

[0040] Generative models can be a type of machine learning model whose goal is to generate new sample data by learning the statistical characteristics of sample data. Generative models can use existing training data and learn the distribution characteristics of the data to generate new data samples.

[0041] Blockchain: It can be a distributed ledger technology that uses cryptography and distributed consensus algorithms to link data together in the form of blocks, forming an immutable chain. Each block contains the hash value of the previous block, ensuring the security and integrity of the data.

[0042] Example 1

[0043] According to an embodiment of this application, a data processing method is provided. It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. Furthermore, although a logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in a different order than that shown here.

[0044] Considering the large number of model parameters in large models and the limited computing resources of mobile terminals, the data processing method provided in this application embodiment can be applied to, for example, Figure 1 The application scenarios shown are not limited to these. In, for example... Figure 1 In the application scenario shown, the large model is deployed on server 10. Server 10 can connect to one or more client devices 20 via a local area network (LAN), wide area network (WAN), internet connection, or other types of data network. These client devices 20 may include, but are not limited to, smartphones, tablets, laptops, PDAs, personal computers, smart home devices, and in-vehicle devices. Client devices 20 can interact with users through a graphical user interface to access the large model, thereby implementing the method provided in this embodiment. Figure 1 This is a schematic diagram of an optional data processing method according to an embodiment of this application, such as... Figure 1 As shown, the user can input the first query information on the graphical user interface on the client device 20 and click the send button, thereby obtaining the first reply information displayed on the graphical user interface. The graphical user interface can be a generative interactive interface. The server 10 can verify the first reply information displayed in the generative interactive interface and obtain a first verification result. If the first verification result is that the content of the first reply information is insecure and the first reply information is generated by the generative model based on the first query information, the server 10 determines the first query information based on the data stored in the blockchain. Based on the first query information, the server adjusts the original training data of the generative model to obtain the target training data.

[0045] In this embodiment, the system consisting of a client device and a server can perform the following steps: The client device inputs a first query message; the server generates a first response message based on the first query message, obtains the first response message, displays the first response message on a generative interactive interface, verifies the first response message displayed on the generative interactive interface, and obtains a first verification result; if the first verification result indicates that the content of the first response message is insecure and the first response message is generated by the generative model based on the first query message, the first query message is determined based on the data stored in the blockchain; based on the first query message, the original training data of the generative model is adjusted to obtain the target training data. It should be noted that this embodiment can be performed on the client device if the client device's operating resources can meet the deployment and operation conditions of a large model.

[0046] Under the aforementioned operating environment, this application provides the following: Figure 2 The data processing method shown. Figure 2 This is a flowchart of the data processing method according to Embodiment 1 of this application. For example... Figure 2 As shown, the method may include the following steps:

[0047] Step S102: Verify the first response information displayed in the generative interactive interface to obtain a first verification result. The first verification result is used to characterize whether the content of the first response information is safe. The first response information is used to characterize the information that responds to the first query information input in the generative interactive interface.

[0048] The aforementioned first response information can be a system response to relevant content entered by the user on the generative interactive interface. The number of first response messages can be one or more. Optionally, the first response information can include text, images, and videos. Images can include three-dimensional graphics (also known as 3D images) and other images, and videos can include 3D videos and other videos. This application does not impose specific limitations on the content of the first response information.

[0049] The first query information mentioned above can be content entered by the user on the generative interactive interface. Optionally, the first query information can include one or more types of content such as text, voice, and images. In this application, there are no specific limitations on the content of the first query information.

[0050] In one optional embodiment, a user can input their query on the generative interactive interface by entering text, voice, or voice-to-text, i.e., inputting a first query. After receiving the first query from the user, the generative interface can respond based on its content, i.e., display a first response on the generative interactive interface. The basic components of the generative interactive interface include an input box, an output box, and a send button. The user can enter a question or instruction in the input box and then click the send button. The system will generate a corresponding answer or operation result based on the user's input and display it in the output box. Optionally, after receiving the first response, it can be uploaded to the cloud and verified by a cloud server to determine whether the content of the first response complies with legal regulations, thus obtaining a first verification result. Optionally, after receiving the first response, it can also be reviewed by auditors to determine whether the content of the first response complies with legal regulations, thus obtaining a first verification result.

[0051] Step S104: If the first verification result indicates that the content of the first response information is insecure, and the first response information is generated by the generation model based on the first query information, the first query information is determined based on the data stored in the blockchain. The blockchain is used to store data corresponding to different query information and data corresponding to the corresponding response information. The corresponding response information is generated by the generation model based on different query information.

[0052] The generative models described above can be used to generate corresponding responses based on user-input queries. The types of input and response information can include images, audio, and text. By using generative models, new samples similar to the training data can be generated, which helps in tasks such as expanding datasets and performing data augmentation. Generative models have wide applications in many fields, such as image generation, music generation, and natural language processing.

[0053] The aforementioned characteristics of blockchain include decentralization, transparency, immutability, and anonymity. Decentralization means that there is no centralized control institution, and all nodes have the right to participate in verifying and updating data. Transparency means that all data records can be publicly viewed, and anyone can verify the authenticity of the data. Immutability means that once data is recorded on the blockchain, it cannot be tampered with or deleted. Anonymity means that users participating in transactions can remain anonymous, and only they know the specific details of the transaction.

[0054] In an optional embodiment, assuming the first verification result indicates that the content of the first response information is insecure, that is, the content of the first response information is invalid, and the first response information is generated by the generative model based on the first query information, then the first query information corresponding to the first response information can be determined by retrieving the data stored in the blockchain. Then, based on the first query information, the specific reason why the content of the first response information is insecure can be accurately determined, thereby adjusting the original training data of the generative model to avoid generating insecure first response information in the future.

[0055] Optionally, since the encoder (Bidirectional Encoder Representations from Transformers, or BERT for short) of the bidirectional encoder-transformer is a natural language processing model that adopts the Transformer architecture, it can learn language representations from massive amounts of text data through large-scale unsupervised training. Its characteristic is that it is a bidirectional encoder, which can simultaneously consider contextual information and better understand the meaning of words. It performs well in various natural language processing tasks (including text classification, named entity recognition, question answering systems, and other natural language processing tasks). Therefore, BERT can also be used as a generation mode and applied to the technical solutions mentioned in this application.

[0056] Step S106: Based on the first query information, adjust the original training data of the generator model to obtain the target training data, wherein the target training data is used to adjust the model parameters of the generator model.

[0057] The original training data mentioned above can refer to the training data used to train the generative model. Different types of training data can be used for different application scenarios, as shown in the following examples:

[0058] Text generation models can use a large amount of text data as raw training data, such as novels, news articles, blog posts, etc. This text data can be crawled from the Internet or obtained from existing datasets.

[0059] Image generation models can use a large amount of image data as raw training data, such as photos, paintings, cartoons, etc. This image data can be crawled from the Internet or obtained from image databases or existing datasets.

[0060] Music generation models can use a large amount of music data as raw training data, such as music files, sheet music, etc., which can be obtained from music databases, music websites, or existing datasets.

[0061] Video generation models can use a large amount of video data as raw training data, such as movies, TV series, video clips, etc. This video data can be obtained from video databases, video websites, or existing datasets.

[0062] 3D model generation: A large amount of 3D model data can be used as raw training data, such as architectural models, human models, object models, etc. This 3D model data can be obtained from 3D model databases, 3D model websites, or existing datasets.

[0063] The target training data mentioned above can be the new training data obtained after adjusting the original data of the generative model. Optionally, the target training data mentioned above can include various types of data such as text data, speech data, and video data.

[0064] In one optional embodiment, since the verification of the first response information is usually not performed simultaneously with its generation, if the first verification result indicates that the content of the first response information is insecure, it is necessary to further determine whether the first response information was generated using a generative model. If the first response information was generated using a generative model—that is, generated based on the first query information—then the original training data of the generative model can be adjusted according to the first query information to obtain target training data. In other words, the training data of the generative model is adjusted, and the parameters of the generative model are updated using the target training data, thereby improving the security of the first response information generated when the generative model generates the first response information based on the first query information. If the first response information was not generated using a generative model—for example, if it was information directly read from a database or searched from the internet—other methods can be used for processing, such as removing the first response information from the database or deleting it from the internet.

[0065] Optionally, based on the data platform providing the first response information, it can be confirmed whether the first response information was generated using a generative model based on the first query information. For example, if the data platform uses a generative model for content generation, then it can be determined that the first response information was generated using a generative model based on the first query information. Optionally, a watermark can be added to the first response information. This watermark records the object that generated the first response information. Therefore, by extracting the watermark from the first response information, it can be determined whether the first response information was generated using a generative model based on the first query information. For example, if the watermark in the first response information records that the object that generated the first response information is a generative model, then it can be determined that the first response information was generated using a generative model based on the first query information.

[0066] In this embodiment, a first verification result is obtained by verifying the first response information displayed in the generative interactive interface. The first verification result is used to characterize whether the content of the first response information is safe, and the first response information is used to characterize the information that responds to the first query information input in the generative interactive interface. If the first verification result indicates that the content of the first response information is unsafe, and the first response information is generated by the generative model based on the first query information, the first query information is determined based on the data stored in the blockchain. The blockchain is used to store data corresponding to different query information and data corresponding to the corresponding response information, and the corresponding response information is generated by the generative model based on different query information. Based on the first query information, the original training data of the generative model is adjusted to obtain target training data, whereby the target training data is used to adjust the model parameters of the generative model. It is noteworthy that by using blockchain to store the data corresponding to different query information and the corresponding response information, when it is determined that the content of the first response information is insecure, and the first response information is generated by the generative model based on the first query information, the reason for the insecurity of the content of the first response information can be accurately located based on the first query information, and the parameters of the generative model can be adjusted. This improves the security of content generation using large models and solves the technical problem of low security when using large models for content generation.

[0067] In the above embodiments of this application, determining the first query information based on the data stored in the blockchain includes one of the following: reading the data corresponding to the first query information stored in the blockchain based on the data corresponding to the first response information, and determining the data corresponding to the first query information as the first query information; determining the data corresponding to the first response information, reading the data corresponding to the first query information stored in the blockchain based on the data corresponding to the first response information, and reading the first query information from a preset storage space based on the data corresponding to the first query information, wherein the preset storage space is used to store data corresponding to different query information and different query information.

[0068] The data corresponding to the first reply information mentioned above can be the first reply information itself, or it can be the hash value of the first reply information.

[0069] In one optional embodiment, the first query information and the first response information corresponding to the first query information can be compressed, and the compressed first query information and the corresponding first response information can be stored in the blockchain, so that when it is necessary to determine the first query information, the desired first query information can be quickly determined from the blockchain.

[0070] In another optional embodiment, the hash value of the first response information can be determined by hash calculation, and the calculated hash value of the first response information can be matched with the hash values ​​of different first query information stored in the blockchain, so as to determine the hash value of the first query information corresponding to the first response information stored in the blockchain. Furthermore, the first query information can be read from a preset storage space outside the blockchain based on the hash value of the first query information.

[0071] In the above embodiments of this application, determining the data corresponding to the first reply information includes: generating a hash value of the first reply information to obtain the data corresponding to the first reply information.

[0072] In one optional embodiment, the hash value corresponding to the first reply information can be obtained by performing a hash calculation on the first reply information, and the hash value corresponding to the first reply information can be determined as the data corresponding to the first reply information.

[0073] In the above embodiments of this application, adjusting the original training data of the generative model based on the first query information to obtain target training data includes: verifying the first query information to obtain a second verification result, wherein the second verification result is used to characterize whether the first query information causes the content of the first response information to be insecure; if the second verification result indicates that the first query information causes the content of the first response information to be insecure, using the first query information and the first response information as negative samples and adding the negative samples to the original training data to obtain target training data; if the second verification result indicates that the first query information does not cause the content of the first response information to be insecure, cleaning the original training data to obtain target training data.

[0074] In one optional embodiment, if the content of the first response information is insecure, it can generally be assumed that the generation model itself has a problem, resulting in the insecurity of the generated first response information, or that the user-inputted first query information is insecure, thus making the generated first response information insecure. Therefore, when adjusting the original training data of the generation model based on the first query information, the security of the first query information can be determined by verifying the first query information. Optionally, during the verification process, the first query information can be uploaded to the cloud for verification, thereby obtaining a second verification result; alternatively, after determining the first query information, those skilled in the art can identify the content in the first query information to verify it, thereby obtaining a second verification result.

[0075] Optionally, if the second verification result indicates that the content of the first response is insecure due to the insecurity of the first query information, the first query information and the first response information can be identified as negative samples and added to the original training data to obtain the target training data. Optionally, by adding negative samples, the generation model can be prevented from generating insecure first response information based on this type of first query information during subsequent data generation.

[0076] Optionally, if the second verification result indicates that the first query information did not cause the content of the first response information to be insecure, that is, the first query information is secure, then it can be assumed that there is insecure data in the original training data of the generative model. Therefore, the original training data can be cleaned to remove insecure training data and obtain the target training data. Optionally, by cleaning the original data, the training data of the generative model can be made to be secure data, thereby avoiding the generation of insecure first response information due to the generative model itself.

[0077] In the above embodiments of this application, the method further includes: detecting the first response information to obtain a detection result corresponding to the first response information, wherein the detection result is used to characterize whether the first response information contains first watermark content; if the detection result indicates that the first response information contains first watermark content, extracting the first watermark content added to the first response information; and determining, based on the first watermark content, whether the first response information is information generated by a generation model based on the first query information.

[0078] The aforementioned first watermark content can be used to indicate the source and attribution information of the first response information. Optionally, when the first response information is audio, video, image, or text data, the first watermark content can be added to the first response information to determine whether the first response information was generated by the generative model.

[0079] Optionally, the aforementioned first watermark content can be a visible watermark or a hidden watermark. A visible watermark can be directly embedded in the first response message, and the user and those skilled in the art can determine the watermark content by observation. A hidden watermark is information secretly embedded in digital images, audio, video, or text. It does not cause visible changes to the original file. It is usually generated by a specific algorithm and embedded in the original file; only those with the corresponding decoding algorithm can extract the information from the hidden watermark.

[0080] In one optional embodiment, after obtaining the first response information, the first response information can be detected to determine whether there is a watermark on the first response information, that is, whether the first watermark content is present on the first response information. Optionally, if there is a watermark in the first response information, the first response information can be extracted to extract the watermark content on the first response information, that is, the first watermark content. Further, the watermark content can be used to determine whether the first response information was generated using a generative model.

[0081] In the above embodiments of this application, the method further includes: obtaining second query information displayed in a generative interactive interface; generating second response information based on the second query information using a generative model; displaying the second response information in the generative interactive interface; and storing the data corresponding to the second query information and the data corresponding to the second response information in a blockchain.

[0082] The aforementioned second response information can be a system response to relevant content entered by the user on the generative interactive interface. The number of second response messages can be one or more. Optionally, the second response information can include text, images, and videos. Images can include three-dimensional graphics (also known as 3D images) and other images, and videos can include 3D videos and other videos. This application does not impose specific limitations on the content of the second response information.

[0083] The second query information mentioned above can be content entered by the user on the generative interactive interface. Optionally, the second query information can include one or more of the following: text, voice, and images. In this application, no specific restrictions are placed on the content of the second query information.

[0084] In one optional embodiment, after a user inputs a second query on the generative interactive interface, a corresponding second response can be generated through a generative model, or a second response matching the second query can be retrieved from the Internet. Furthermore, the second query and the second response can be stored, that is, stored on the blockchain, so that the stored data cannot be modified, thereby ensuring the integrity of the data preservation. This allows the second query corresponding to the second response to be retrieved in a timely manner when the security of the second response is compromised, and the cause of the security problem can be determined based on the second query, whether it is due to the insecurity of the second query itself or due to the generative model.

[0085] In the above embodiments of this application, displaying second response information in a generative interactive interface and storing the data corresponding to the second query information and the data corresponding to the second response information in the blockchain includes: displaying abbreviated information corresponding to the second response information in the generative interactive interface, wherein the amount of information in the abbreviated information is less than the amount of information in the second response information; outputting prompt information in the generative interactive interface, wherein the prompt information is used to prompt whether to store the data corresponding to the second query information and the data corresponding to the second response information in the blockchain; and responding to a confirmation command acting on the generative interactive interface, displaying the second response information in the generative interactive interface and storing the data corresponding to the second query information and the data corresponding to the second response information in the blockchain, wherein the confirmation command is used to indicate confirmation of storing the data corresponding to the second query information and the data corresponding to the second response information in the blockchain.

[0086] The abbreviated information corresponding to the second response information mentioned above can be the second response information with lower clarity.

[0087] The information content of the aforementioned abbreviation can be used to represent the attribute information of the second response information. Specifically, the following examples illustrate this: if the second response information is text information, the information content of the abbreviation can be the number of characters; if the second response information is image information, the information content of the abbreviation can be the image resolution; if the second response information is video, the information content of the abbreviation can be the video resolution or the video duration; if the second response information is audio, the information content of the abbreviation can be the audio duration or the audio clarity.

[0088] The aforementioned prompts can be displayed in various forms such as dialog boxes, pop-ups, voice messages, and virtual buttons. Optionally, the aforementioned prompts can be used to ask the user whether to store the data corresponding to the second query and the data corresponding to the second reply in the blockchain.

[0089] The aforementioned confirmation command can be expressed through voice, text, or pressing a virtual button. Optionally, this application does not impose specific limitations on the form of the confirmation command. Optionally, a prompt message can be displayed on the generative interactive interface. If the prompt message is displayed in the form of a dialog box or pop-up window, the user can enter text such as "confirm" to indicate confirmation, thereby issuing a confirmation command. If the prompt message is expressed through voice, for example, the generative interactive interface issues a voice prompt "Confirm on-chain?", that is, whether the user confirms storing the second query information and the second reply information on the blockchain, the user can reply "confirm" through voice to issue a confirmation command. If the prompt message is expressed through a virtual button, for example, a virtual button is displayed on the generative interactive interface with the text "Click to confirm on-chain", then the user issues a confirmation command by clicking the virtual button.

[0090] In one optional embodiment, when a user inputs a second query on the generative interactive interface, a thumbnail of the corresponding second response can be displayed on the interface. Optionally, since the displayed thumbnail of the second response has low clarity, a prompt can be displayed on the interface to remind the user that to view the clearer second response, the data corresponding to the second query and the data corresponding to the second response must be stored on the blockchain. Optionally, if the user confirms storing the data corresponding to the second query and the data corresponding to the second response on the blockchain, a clearer second response can be displayed on the interface, and the data corresponding to the second query and the data corresponding to the second response can be stored on the blockchain, thereby ensuring that the relevant data generated each time the user uses the generative model to generate data can be stored.

[0091] In the above embodiments of this application, if no confirmation instruction is received within a preset time period, the display of the second response information in the generative interactive interface is prohibited, and the data corresponding to the second query information and the data corresponding to the second response information are prohibited from being stored in the blockchain.

[0092] The aforementioned preset time period can be set by those skilled in the art according to their needs. In this application, there is no specific limitation on the length of the preset time.

[0093] In one optional embodiment, if no confirmation instruction is received from the user within a preset time period after the prompt information is displayed on the generative interactive interface, it indicates that the user does not want the data corresponding to the second query information and the data corresponding to the second reply information to be stored in the blockchain. Here, the generative interactive interface only displays the first reply information with a smaller amount of information, and the user cannot view the second reply information with a larger amount of information.

[0094] In the above embodiments of this application, the method further includes: obtaining the second watermark content corresponding to the second reply information; adding the second watermark content to the second reply information by means of a dark watermark to obtain the target reply information; and displaying the target reply information in a generative interactive interface.

[0095] The aforementioned hidden watermarking method is a technique that embeds hidden information into digital media to prove the copyright ownership of the digital media, prevent piracy, and protect the integrity of the content. Optionally, hidden watermarks can only be detected by specific decoding methods. The following are some common hidden watermarking methods:

[0096] Bit modulation: This method involves embedding hidden information into specific bits of digital media. This can be achieved by adding tiny noise to specific bits or changing the brightness of pixels.

[0097] Transform domain watermarking: By embedding hidden information into the transform domain of digital media, such as Fourier transform or wavelet transform, the watermark information is embedded by changing the values ​​of the transform coefficients to ensure minimal impact on the original media.

[0098] Spread spectrum watermarking: This method embeds hidden information into high-frequency regions by using spread spectrum technology in digital media. It hides watermark information by adding high-frequency signals to the media.

[0099] Frequency domain watermarking: This involves embedding hidden information into the frequency domain of digital media, such as the spectrum or frequency domain. In other words, watermark information is embedded by changing the spectrum distribution or the position of frequency components.

[0100] The aforementioned second watermark content can be used to indicate the source of the second response information. The second watermark content may include whether the second response information is generated by a generative model or obtained from the Internet.

[0101] The target response information mentioned above can be obtained by adding a second watermark to the second response information.

[0102] In one optional embodiment, after obtaining the second response information, a second watermark can be added to the second response information using a hidden watermark method to obtain the target response information, which is then displayed on the generative interactive interface. Optionally, adding the watermark using a hidden watermark method ensures that it does not cause visible changes to the original data, while allowing those skilled in the art to easily obtain the source of the second response content through relevant decryption methods when needed. In other words, this improves the user experience and also allows for the identification of the cause of security issues when they arise in the second response content.

[0103] In the above embodiments of this application, the method further includes: when the first response information is information generated by a generative model based on the first query information, controlling the attribute of the generative interactive interface to be a first attribute, wherein the attribute includes at least one of the following: the interface color of the generative interactive interface, the interface size of the generative interactive interface, the font of the content displayed in the generative interactive interface, the content color of the displayed content, and the content size of the displayed content; when the first response information is not information generated by a generative model based on the first query information, controlling the attribute of the generative interactive interface to be a second attribute; responding to a confirmation instruction, controlling the attribute of the generative interactive interface to be a third attribute; and after adding the second watermark content to the second response information, controlling the attribute of the generative interactive interface to be a fourth attribute.

[0104] The first attribute mentioned above can be a generative interaction mode.

[0105] The second attribute mentioned above can be a retrieval-based interactive mode.

[0106] The aforementioned third attribute can be a generative interaction mode for blockchain.

[0107] The fourth attribute mentioned above can be a watermark-generating interactive mode.

[0108] In one optional embodiment, the generative interactive interface can be divided into attributes. Optionally, if the first response information is generated using a generative model based on the first query information, the attribute of the generative interactive interface can be controlled to be a generative interactive mode. Specifically, in the case of generative interactive mode, the interface color, interface size, font of the content displayed in the generative interactive interface, content color of the displayed content, and content size of the displayed content can be adjusted.

[0109] In another alternative embodiment, if the first response information is generated using a generative model that is not based on the first query information, that is, if the first response information is determined from an internet database based on the first query information, then the attribute of the generative interactive interface can be controlled to be a retrieval-based interactive mode.

[0110] In another optional embodiment, if a confirmation instruction is received indicating confirmation to store the first query information and the first response information in the blockchain, the attribute of the generative interactive interface can be controlled to be a blockchain generative interactive mode. In this mode, all the first query information input into the generative interactive interface, as well as the first response information corresponding to the first query information, can be stored in the blockchain.

[0111] In another optional embodiment, if the second watermark content has already been added to the second reply information, the attribute of the generative interactive interface can be controlled to be watermark generative interactive mode. That is, watermarks can be added to all reply information generated in this mode.

[0112] In the above embodiments of this application, the method further includes: in response to a confirmation instruction, or after adding a second watermark to the second response information, archiving the second query information to obtain an archived record; summarizing the archived record and the second response information, or summarizing the archived record and the target response information to obtain copyright information; and storing the copyright information.

[0113] In an optional embodiment, if a confirmation instruction indicating confirmation to store the second query information and the second response information in the blockchain has been received, or if the second watermark content has been added to the second response information, the corresponding second query information can be stored, and an archived record can be determined according to the storage order. The archived record can be an archive of multiple query information inputs by the user in the generative interactive interface, which can be a screenshot or content similar to chat history, recording the time of information sending.

[0114] Optionally, the archived records and each second query message contained therein, along with the corresponding second response message, can be aggregated to obtain copyright information, which can then be stored in the blockchain. This allows for the timely identification of the source of the second response message and the corresponding second query message when a security issue is discovered, thus quickly determining the cause of the security problem.

[0115] In the above embodiments of this application, generating second response information based on second query information using a generative model includes at least one of the following: sending second query information to a cloud server and receiving second response information returned by the cloud server, wherein the generative model is deployed in the cloud server; obtaining the generative model by calling a preset interface and generating second response information based on the second query information using the generative model; sending second query information to the cloud server, receiving a first semantic analysis result corresponding to the second query information returned by the cloud server, determining the target data corresponding to the first semantic analysis result using the generative model, sending the target data to the cloud server, receiving a second semantic analysis result corresponding to the target data returned by the cloud server, and generating second response information based on the first semantic analysis result and the second semantic analysis result using the generative model.

[0116] The aforementioned preset interface can be a network interface.

[0117] The first semantic analysis result mentioned above can be the result obtained by parsing the second query information through a cloud server.

[0118] The target data mentioned above can be the semantic content generated by the generative model based on the first semantic analysis result.

[0119] The second speech analysis result mentioned above can be used to represent the result obtained by parsing the target data through a cloud server.

[0120] In one alternative embodiment, the second query information can be sent to a cloud server, a generative model deployed on the cloud server can generate a second response information corresponding to the second query information, and the second response information can be displayed on a generative interactive interface.

[0121] In another optional embodiment, the second query information can be sent to a cloud server, and the cloud server can parse the second query information to obtain the first semantic analysis result, that is, determine the content corresponding to the second query information. Optionally, the cloud server can parse the second query information through a pre-deployed semantic analysis module, or through a semantic analysis module provided by a third party, or by calling an interface to use a semantic analysis module to parse the second query information. Further, a generative model can generate corresponding target data based on the first semantic analysis result, and send the target data to the cloud server to obtain the second semantic analysis result corresponding to the target data. That is, the second semantic analysis result related to the target data on the Internet can be obtained from the cloud server. Finally, the generative model uses the first and second semantic analysis results to generate the second response information.

[0122] Figure 3 This is a schematic diagram of a data processing method according to an embodiment of this application, such as... Figure 3 As shown, the scenario of generating images or videos using a generative model is illustrated. Users can send their input queries to the question understanding module by calling the corresponding interface of the generative interactive interface on the client side. The question understanding module then calls the semantic analysis module on the cloud server to perform semantic understanding on the received queries. Further, a large model generates corresponding response content, and the result understanding module calls the semantic analysis module on the cloud server to understand the response content. Finally, the large model generates the final image or video, which is then displayed in low resolution on the generative interactive interface by calling the interface. For example... Figure 3The system presents results 1, 2, and 3, resulting in unclear images or videos for the user. Furthermore, after the user selects their desired response from the different options, they can be prompted to choose whether to upload the query and response to the blockchain. If the user selects result 1 and chooses to upload it to the blockchain, result 1 can be displayed in high resolution on the generative interactive interface, and hash values ​​for both the query and result 1 can be generated. These two hash values ​​are stored in the blockchain, and the hash values ​​and compressed text are stored off-chain. Optionally, if the user does not choose to upload to the blockchain, the high-resolution image or video will not be improved.

[0123] Optionally, assuming that the result 1 in the second resolution form has a content security issue discovered externally, it can be extracted through dark watermarking to verify whether the content is the result generated by the large model. That is, the hash value of the corresponding query information is searched in the blockchain, and the text stored off-chain is further extracted. If it is determined that it is not a result induced by the user, it means that the corresponding large model training data is contaminated and the original training data needs to be cleaned. If it is determined that it is a result induced by the user, that is, the query information entered by the user has a security issue, then the text content needs to be used as negative text training data, and the original training data is updated on the cloud service.

[0124] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties. Furthermore, the collection, use and processing of the relevant data must comply with the relevant laws, regulations and standards of the relevant countries and regions, and corresponding operation portals are provided for users to choose to authorize or refuse.

[0125] It should be noted that, for the sake of simplicity, the foregoing method embodiments are all described as a series of actions. However, those skilled in the art should understand that this application is not limited to the described order of actions, as some steps may be performed in other orders or simultaneously according to this application. Furthermore, those skilled in the art should also understand that the embodiments described in the specification are preferred embodiments, and the actions and modules involved are not necessarily essential to this application.

[0126] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods according to the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, they can also be implemented by hardware. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk), and includes several instructions to cause a terminal device (which may be a mobile phone, computer, server, or network device, etc.) to execute the methods of the various embodiments of this application.

[0127] Example 2

[0128] According to an embodiment of this application, a data processing method is also provided. Figure 4 This is a flowchart of the data processing method according to Embodiment 2 of this application, as follows: Figure 4 As shown, the method includes the following steps:

[0129] Step S402: Obtain the first query information displayed in the generative interactive interface;

[0130] Step S404: Generate the first response information based on the first query information using the generative model;

[0131] Step S406: Display the first response information in the generative interactive interface;

[0132] Step S408: Store the data corresponding to the first query information and the data corresponding to the first response information in the blockchain. The blockchain is used to store the data corresponding to different query information and the data corresponding to the corresponding response information. The corresponding response information is generated by the generative model based on different query information.

[0133] In one optional embodiment, the user can input the first query information into the generative interactive interface. After obtaining the first query information input by the user, the generative model can be used to generate the corresponding first response information based on the first query information, and the first response information can be displayed on the generative interactive interface. Furthermore, the data corresponding to different query information and the data corresponding to the response information of different query information can be stored on the blockchain.

[0134] Figure 5 This is a schematic diagram of information interaction according to Embodiment 2 of this application, such as... Figure 5 As shown, users can enter the first query information in the input box of the generative interactive interface and send it via the motor button. After receiving the first query information, the user can generate the corresponding first response information through the generative model and display the first response information on the generative interactive interface.

[0135] In the above embodiments of this application, displaying a first response message in a generative interactive interface and storing the data corresponding to the first query message and the data corresponding to the first response message in the blockchain includes: displaying abbreviated information corresponding to the first response message in the generative interactive interface, wherein the amount of information in the abbreviated information is less than the amount of information in the first response message; outputting a prompt message in the generative interactive interface, wherein the prompt message is used to prompt whether to store the data corresponding to the first query message and the data corresponding to the first response message in the blockchain; and responding to a confirmation command acting on the generative interactive interface, displaying the first response message in the generative interactive interface and storing the data corresponding to the first query message and the data corresponding to the first response message in the blockchain, wherein the confirmation command is used to indicate confirmation of storing the data corresponding to the first query message and the data corresponding to the first response message in the blockchain.

[0136] In the above embodiments of this application, if no confirmation instruction is received within a preset time period, the first response information is prohibited from being displayed in the generative interactive interface, and the data corresponding to the first inquiry information and the data corresponding to the first response information are prohibited from being stored in the blockchain.

[0137] In the above embodiments of this application, the method further includes: obtaining the first watermark content corresponding to the first reply information; adding the first watermark content to the first reply information by means of a dark watermark to obtain the target reply information; and displaying the target reply information in a generative interactive interface.

[0138] In the above embodiments of this application, generating a first response based on a first query by a generative model includes at least one of the following: sending the first query to a cloud server and receiving the first response returned by the cloud server, wherein the generative model is deployed in the cloud server; obtaining the generative model by calling a preset interface and generating the first response based on the first query by the generative model; sending the first query to the cloud server, receiving the first semantic analysis result corresponding to the first query returned by the cloud server, determining the target data corresponding to the first semantic analysis result using the generative model, sending the target data to the cloud server, receiving the first semantic analysis result corresponding to the target data returned by the cloud server, and generating the first response based on the first semantic analysis result and the first semantic analysis result using the generative model.

[0139] Example 3

[0140] According to an embodiment of this application, a data processing method is also provided. Figure 6 This is a flowchart of the data processing method according to Embodiment 3 of this application, as follows: Figure 6 As shown, the method includes the following steps:

[0141] Step S602: The first response information displayed in the generative interactive interface;

[0142] Step S604: In response to the verification command applied to the generative interactive interface, display the first verification result on the first query information, wherein the first verification result is used to characterize whether the content of the first response information is safe, and the first response information is used to characterize the information that responds to the first query information input in the generative interactive interface;

[0143] In one alternative embodiment, after the first response information is displayed on the generative interactive interface, a verification instruction can be displayed simultaneously, thereby instructing the display of a first verification result on the first query information, wherein the first verification result can be used to indicate whether the first response information is secure.

[0144] Step S606: In response to the adjustment command applied to the generative interactive interface, display the target training data of the generative model on the first query information. The target training data is the data obtained by adjusting the original training data of the generative model based on the first query information. The first query information is the information determined based on the data stored in the blockchain when the first verification result is that the content of the first response information is insecure and the first response information is the information generated by the generative model based on the first query information. The blockchain is used to store the data corresponding to different query information and the data corresponding to the corresponding response information. The corresponding response information is the information generated by the generative model based on different query information.

[0145] In one optional embodiment, after an adjustment instruction is displayed on the generative interface, the original training data of the generative model can be adjusted to obtain the target training data. Optionally, if the first verification result indicates that the content of the first response information is insecure, and the first response information is generated by the generative model based on the first query information, the first query information can be determined from the data corresponding to different query information and the corresponding response information stored in the blockchain. Based on the first query information, it can be determined whether the insecurity of the first response information is due to the insecurity of the first query information or the insecurity of the original training data of the generative model. If it is due to the insecurity of the first query information, the first query information can be used as a negative sample to obtain the target training data. If it is due to the insecurity of the original training data, the original training data can be cleaned to obtain the target training data.

[0146] Example 4

[0147] According to an embodiment of this application, a data processing method is also provided. Figure 7 This is a flowchart of the data processing method according to Embodiment 4 of this application, as follows: Figure 7As shown, the method includes the following steps:

[0148] Step S702: In response to the input command applied to the generative interactive interface, display the first query information on the generative interactive interface;

[0149] Step S704: In response to the generation instruction applied to the first query information, display the first response information on the first query information. The first response information is generated based on the first query information using a generation model. The data corresponding to the first query information and the data corresponding to the first response information are stored in the blockchain. The blockchain is used to store the data corresponding to different query information and the data corresponding to the corresponding response information. The corresponding response information is generated based on different query information using a generation model.

[0150] In one optional embodiment, after an input command is displayed on the generative interactive interface, a first query message can be displayed on the generative interactive interface. Furthermore, when a generation command is displayed on the generative interactive interface, a first response message can be displayed on the first query message, and the first query message and the first response message are stored in the blockchain.

[0151] Example 5

[0152] According to an embodiment of this application, a data processing method is also provided. Figure 8 This is a flowchart of the data processing method according to Embodiment 5 of this application, as follows: Figure 8 As shown, the method includes the following steps:

[0153] Step S802: Obtain the first response information displayed in the generative interactive interface by calling the first interface, wherein the first interface includes a first parameter, and the parameter value of the first parameter is the first response information;

[0154] The first interface mentioned above can be a data connection cable interface or a virtual interface. Optionally, the content of the first interface is not specifically limited in this invention.

[0155] Step S804: Verify the first response information to obtain a first verification result, wherein the first verification result is used to characterize whether the content of the first response information is safe, and the first response information is used to characterize the information that responds to the first query information input in the generative interactive interface;

[0156] Step S806: If the first verification result indicates that the content of the first response information is insecure, and the first response information is information generated by the generation model based on the first query information, the first query information is determined based on the data stored in the blockchain. The blockchain is used to store data corresponding to different query information and data corresponding to the corresponding response information. The corresponding response information is information generated by the generation model based on different query information.

[0157] Step S808: Based on the first query information, adjust the original training data of the generator model to obtain the target training data, wherein the target training data is used to adjust the model parameters of the generator model;

[0158] Step S810: Output the target training data by calling the second interface, wherein the second interface includes a second parameter, and the parameter value of the second parameter is the target training data.

[0159] The second interface mentioned above can be a data connection cable interface or a virtual interface. Optionally, the content of the second interface is not specifically limited in this invention.

[0160] In one optional embodiment, the first response information can be obtained through a first interface and verified to determine whether the first response information is secure. Optionally, if the first verification result indicates that the content of the first response information is insecure, and the first response information is generated by a generative model based on the first query information, the corresponding first query information can be determined by retrieving data stored in the blockchain. Based on the first query information, it can be determined whether the insecurity of the first response information is due to the insecurity of the first query information or the insecurity of the original training data of the generative model. If it is due to the insecurity of the first query information, the first query information can be used as a negative sample to obtain the target training data. If it is due to the insecurity of the original training data, the original training data can be cleaned to obtain the target training data. Furthermore, the target training data can be output through a second interface.

[0161] Example 6

[0162] According to an embodiment of this application, a data processing method is also provided. Figure 9 This is a flowchart of the data processing method according to Embodiment 6 of this application, as follows: Figure 9 As shown, the method includes the following steps:

[0163] Step S902: Input multimodal information in the dialog interface, wherein the type of multimodal information includes at least one of the following: text information containing character information, video frame information containing frame image information, and audio information;

[0164] The aforementioned dialogue interface can be a generative interactive interface or other dialogue interfaces. Optionally, this application does not specifically limit the type of dialogue interface.

[0165] Step S904: Generate response information corresponding to the multimodal information based on the multimodal information using a generative model, wherein the type of response information includes at least one of the following: text information, image information, video information, and voice information;

[0166] Step S906: Store the data corresponding to the multimodal information and the data corresponding to the response information in the blockchain. The blockchain is used to store the data corresponding to different query information and the data corresponding to the corresponding response information. The corresponding response information is generated by the generative model based on different query information.

[0167] In one alternative embodiment, the user can input multimodal information on the dialogue interface, and the generative model can generate corresponding response information based on the multimodal information input by the user. Furthermore, the multimodal information input by the user and the corresponding response information generated by the generative model can be stored in a blockchain.

[0168] Example 7

[0169] According to an embodiment of this application, a data processing system is also provided. Figure 10 This is a schematic diagram of the data processing system according to Embodiment 7 of this application, as shown below. Figure 10 As shown, the system includes:

[0170] The front-end client 1002 is used to display the generative interactive interface and capture the first query information entered in the generative interactive interface;

[0171] Backend server 1004 connects to the frontend client and is used to generate the first response information based on the first query information using the generative model.

[0172] Blockchain 1006 connects to the backend server and is used to store the data corresponding to the first query information and the data corresponding to the first response information.

[0173] Example 8

[0174] According to an embodiment of this application, an apparatus for implementing the above-described data processing method is also provided. Figure 11 This is a schematic diagram of a data processing apparatus according to Embodiment 8 of this application, as shown below. Figure 11 As shown, the device includes: a verification module 1102, a determination module 1104, and an adjustment module 1106.

[0175] The aforementioned verification module 1102 is used to verify the first response information displayed in the generative interactive interface to obtain a first verification result. The first verification result is used to characterize whether the content of the first response information is safe, and the first response information is used to characterize the information that responds to the first query information input in the generative interactive interface. The determination module 1104 is used to determine the first query information based on data stored in the blockchain when the first verification result indicates that the content of the first response information is unsafe and the first response information is information generated by the generative model based on the first query information. The blockchain is used to store data corresponding to different query information and data corresponding to the corresponding response information. The corresponding response information is information generated by the generative model based on different query information. The adjustment module 1106 is used to adjust the original training data of the generative model based on the first query information to obtain target training data. The target training data is used to adjust the model parameters of the generative model.

[0176] It should be noted that the verification module 1102, determination module 1104, and adjustment module 1106 mentioned above correspond to steps S202 to S206 in Embodiment 1. The modules and corresponding steps implement the same instances and application scenarios, but are not limited to the content disclosed in Embodiment 1. It should be noted that the above modules or units can be hardware or software components stored in memory and processed by one or more processors. The above modules can also be part of the device and run in the server 10 provided in Embodiment 1.

[0177] In the above embodiments of this application, the determining module 1104, based on the data stored in the blockchain, includes one of the following: a first determining unit, configured to read the data corresponding to the first query information stored in the blockchain based on the data corresponding to the first reply information, and determine the data corresponding to the first query information as the first query information; a second determining unit, configured to determine the data corresponding to the first reply information, read the data corresponding to the first query information stored in the blockchain based on the data corresponding to the first reply information, and read the first query information from a preset storage space based on the data corresponding to the first query information, wherein the preset storage space is used to store data corresponding to different query information and different query information.

[0178] In the above embodiments of this application, the second determining unit includes: a generating subunit, used to generate a hash value of the first reply information and obtain the data corresponding to the first reply information.

[0179] In the above embodiments of this application, the adjustment module 1106 includes: a verification unit, used to verify the first query information and obtain a second verification result, wherein the second verification result is used to characterize whether the first query information causes the content of the first response information to be insecure; an addition unit, used to add the first query information and the first response information as negative samples and add the negative samples to the original training data to obtain target training data when the second verification result is that the first query information causes the content of the first response information to be insecure; and a cleaning unit, used to clean the original training data to obtain target training data when the second verification result is that the first query information does not cause the content of the first response information to be insecure.

[0180] In the above embodiments of this application, the device further includes: a detection module, used to detect the first response information and obtain a detection result corresponding to the first response information, wherein the detection result is used to characterize whether the first response information contains first watermark content; an extraction module, used to extract the first watermark content added to the first response information when the detection result indicates that the first response information contains first watermark content; and a second determination module, used to determine, based on the first watermark content, whether the first response information is information generated by a generation model based on the first query information.

[0181] In the above embodiments of this application, the device further includes: a first acquisition module, used to acquire second query information displayed in a generative interactive interface; a generation module, used to generate second response information based on the second query information using a generative model; and a storage module, used to display the second response information in the generative interactive interface and store the data corresponding to the second query information and the data corresponding to the second response information in a blockchain.

[0182] In the above embodiments of this application, the storage module includes: a first display unit, configured to display abbreviated information corresponding to the second response information in a generative interactive interface, wherein the amount of information in the abbreviated information is less than the amount of information in the second response information; an output unit, configured to output prompt information in the generative interactive interface, wherein the prompt information is used to prompt whether to store the data corresponding to the second query information and the data corresponding to the second response information in the blockchain; and a second display unit, configured to respond to a confirmation command acting on the generative interactive interface, display the second response information in the generative interactive interface, and store the data corresponding to the second query information and the data corresponding to the second response information in the blockchain, wherein the confirmation command is used to indicate confirmation of storing the data corresponding to the second query information and the data corresponding to the second response information in the blockchain.

[0183] In the above embodiments of this application, the device further includes: a second acquisition module, used to acquire the second watermark content corresponding to the second reply information; an addition module, used to add the second watermark content to the second reply information by means of a dark watermark, to obtain the target reply information; and a display module, used to display the target reply information in a generative interactive interface.

[0184] In the above embodiments of this application, the device further includes: a first control module, configured to control the attribute of the generative interactive interface to a first attribute when the first response information is information generated by a generative model based on the first query information, wherein the attribute includes at least one of the following: the interface color of the generative interactive interface, the interface size of the generative interactive interface, the font of the content displayed in the generative interactive interface, the content color of the displayed content, and the content size of the displayed content; a second control module, configured to control the attribute of the generative interactive interface to a second attribute when the first response information is not information generated by a generative model based on the first query information; a third control module, configured to control the attribute of the generative interactive interface to a third attribute in response to a confirmation command; and a fourth control module, configured to control the attribute of the generative interactive interface to a fourth attribute after adding second watermark content to the second response information.

[0185] In the above embodiments of this application, the device further includes: an archiving module, used to archive the second query information in response to a confirmation command or after adding second watermark content to the second reply information, to obtain an archived record; a summarizing module, used to summarize the archived record and the second reply information, or to summarize the archived record and the target reply information, to obtain copyright information; and a notarization module, used to notarize the copyright information.

[0186] In the above embodiments of this application, the generation module includes: a first sending unit, configured to send a second query message to a cloud server and receive a second response message returned by the cloud server, wherein the generation model is deployed in the cloud server; a calling unit, configured to obtain the generation model by calling a preset interface and generate the second response message based on the second query message using the generation model; the first sending unit is configured to send the second query message to the cloud server, receive a first semantic analysis result corresponding to the second query message returned by the cloud server, determine the target data corresponding to the first semantic analysis result using the generation model, send the target data to the cloud server, receive a second semantic analysis result corresponding to the target data returned by the cloud server, and generate the second response message based on the first and second semantic analysis results using the generation model.

[0187] It should be noted that the preferred implementation schemes involved in the above embodiments of this application are the same as the schemes, application scenarios and implementation processes provided in Embodiment 1, but are not limited to the schemes provided in Embodiment 1.

[0188] Example 9

[0189] According to an embodiment of this application, an apparatus for implementing the above-described data processing method is also provided. Figure 12 This is a schematic diagram of a data processing apparatus according to Embodiment 9 of this application, as shown below. Figure 12 As shown, the device includes: an acquisition module 1202, a generation module 1204, a display module 1206, and a storage module 1208.

[0190] Specifically, the acquisition module 1202 is used to acquire the first query information displayed in the generative interactive interface; the generation module 1204 is used to generate the first response information based on the first query information using a generative model; the display module 1206 is used to display the first response information in the generative interactive interface; and the storage module 1208 is used to store the data corresponding to the first query information and the data corresponding to the first response information in the blockchain, wherein the blockchain is used to store the data corresponding to different query information and the data corresponding to the corresponding response information, and the corresponding response information is the information generated by the generative model based on different query information.

[0191] In the above embodiments of this application, the storage module 1208 includes: a first display unit, configured to display abbreviated information corresponding to the first reply information in a generative interactive interface, wherein the amount of information in the abbreviated information is less than the amount of information in the first reply information; an output unit, configured to output prompt information in the generative interactive interface, wherein the prompt information is used to prompt whether to store the data corresponding to the first query information and the data corresponding to the first reply information in the blockchain; and a second display unit, configured to respond to a confirmation command acting on the generative interactive interface, display the first reply information in the generative interactive interface, and store the data corresponding to the first query information and the data corresponding to the first reply information in the blockchain, wherein the confirmation command is used to indicate confirmation of storing the data corresponding to the first query information and the data corresponding to the first reply information in the blockchain.

[0192] In the above embodiments of this application, the device further includes: a second acquisition module, used to acquire the first watermark content corresponding to the first reply information; an addition module, used to add the first watermark content to the first reply information by means of a dark watermark, to obtain the target reply information; and a second display module, used to display the target reply information in a generative interactive interface.

[0193] In the above embodiments of this application, the generation module 1204 includes at least one of the following: a first sending unit, configured to send first query information to a cloud server and receive first response information returned by the cloud server, wherein the generation model is deployed in the cloud server; an acquisition unit, configured to acquire the generation model by calling a preset interface and generate first response information based on the first query information using the generation model; and a second sending unit, configured to send the first query information to the cloud server, receive the first semantic analysis result corresponding to the first query information returned by the cloud server, determine the target data corresponding to the first semantic analysis result using the generation model, send the target data to the cloud server, receive the first semantic analysis result corresponding to the target data returned by the cloud server, and generate the first response information based on the first semantic analysis result and the first semantic analysis result using the generation model.

[0194] It should be noted that the acquisition module 1202, generation module 1204, display module 1206, and storage module 1208 mentioned above correspond to steps S402 to S408 in Embodiment 2. The modules and corresponding steps implement the same instances and application scenarios, but are not limited to the content disclosed in Embodiment 1. It should be noted that the above modules or units can be hardware or software components stored in memory and processed by one or more processors. The above modules can also run as part of the device in the server 10 provided in Embodiment 1.

[0195] Example 10

[0196] According to an embodiment of this application, an apparatus for implementing the above-described data processing method is also provided. Figure 13 This is a schematic diagram of a data processing apparatus according to Embodiment 10 of this application, as shown below. Figure 13 As shown, the device includes: a first display module 1302, a second display module 1304, and a third display module 1306.

[0197] The first display module 1302 is used to display the first response information in the generative interactive interface; the second display module 1304 is used to respond to the verification command applied to the generative interactive interface and display the first verification result on the first query information, wherein the first verification result is used to characterize whether the content of the first response information is safe, and the first response information is used to characterize the information that responds to the first query information input in the generative interactive interface; the third display module 1306 is used to respond to the adjustment command applied to the generative interactive interface and display the target training data of the generative model on the first query information, wherein the target training data is the data obtained by adjusting the original training data of the generative model based on the first query information, the first query information is the information determined based on the data stored in the blockchain when the first verification result is that the content of the first response information is unsafe, and the first response information is the information generated by the generative model based on the first query information, and the blockchain is used to store the data corresponding to different query information and the data corresponding to the corresponding response information, and the corresponding response information is the information generated by the generative model based on different query information.

[0198] It should be noted that the first display module 1302, the second display module 1304, and the third display module 1306 mentioned above correspond to steps S602 to S606 in Embodiment 3. The modules and corresponding steps implement the same instances and application scenarios, but are not limited to the content disclosed in Embodiment 1. It should be noted that the above modules or units can be hardware or software components stored in memory and processed by one or more processors. The above modules can also run as part of the device in the server 10 provided in Embodiment 1.

[0199] Example 11

[0200] According to an embodiment of this application, an apparatus for implementing the above-described data processing method is also provided. Figure 14 This is a schematic diagram of a data processing apparatus according to Embodiment 11 of this application, as shown below. Figure 14 As shown, the device includes: a first display module 1402 and a second display module 1404.

[0201] The first display module 1402 is used to respond to input commands applied to the generative interactive interface and display first query information on the generative interactive interface; the second display module 1404 is used to respond to generation commands applied to the first query information and display first response information on the first query information. The first response information is generated based on the first query information using a generative model. The data corresponding to the first query information and the data corresponding to the first response information are stored in a blockchain. The blockchain is used to store the data corresponding to different query information and the data corresponding to the corresponding response information. The corresponding response information is generated based on different query information using a generative model.

[0202] It should be noted that the first display module 1402 and the second display module 1404 mentioned above correspond to steps S702 to S704 in Embodiment 4. The modules and corresponding steps implement the same instances and application scenarios, but are not limited to the content disclosed in Embodiment 1. It should be noted that the above modules or units can be hardware components or software components stored in memory and processed by one or more processors. The above modules can also be part of the device and run in the server 10 provided in Embodiment 1.

[0203] Example 12

[0204] According to an embodiment of this application, an apparatus for implementing the above-described data processing method is also provided. Figure 15 This is a schematic diagram of a data processing apparatus according to Embodiment 12 of this application, as shown below. Figure 15 As shown, the device includes: a calling module 1502, a verification module 1504, a determination module 1506, an adjustment module 1508, and an output module 1510.

[0205] The aforementioned calling module 1502 is used to obtain the first response information displayed in the generative interactive interface by calling the first interface, wherein the first interface includes a first parameter, and the parameter value of the first parameter is the first response information; the verification module 1504 is used to verify the first response information and obtain a first verification result, wherein the first verification result is used to characterize whether the content of the first response information is safe, and the first response information is used to characterize the information that responds to the first query information input in the generative interactive interface; the determining module 1506 is used to determine whether the first response information is unsafe based on the first query information in the first verification result. In the case of generated information, based on the data stored in the blockchain, the first query information is determined, wherein the blockchain is used to store data corresponding to different query information and data corresponding to the corresponding response information, and the corresponding response information is information generated by the generation model based on different query information; the adjustment module 1508 is used to adjust the original training data of the generation model based on the first query information to obtain target training data, wherein the target training data is used to adjust the model parameters of the generation model; the output module 1510 is used to output the target training data by calling the second interface, wherein the second interface includes a second parameter, and the parameter value of the second parameter is the target training data.

[0206] It should be noted that the aforementioned calling module 1502, verification module 1504, determination module 1506, adjustment module 1508, and output module 1510 correspond to steps S802 to S810 in Embodiment 5. The instances and application scenarios implemented by the modules and corresponding steps are the same, but are not limited to the content disclosed in Embodiment 1. It should be noted that the aforementioned modules or units can be hardware or software components stored in memory and processed by one or more processors. The aforementioned modules can also run as part of the device in the server 10 provided in Embodiment 1.

[0207] Example 13

[0208] According to an embodiment of this application, an apparatus for implementing the above-described data processing method is also provided. Figure 16 This is a schematic diagram of a data processing apparatus according to Embodiment 13 of this application, as shown below. Figure 16 As shown, the device includes: an input module 1602, a generation module 1604, and a storage module 1606.

[0209] The input module 1602 is used to input multimodal information into the dialogue interface. The type of multimodal information includes at least one of the following: text information containing character information, video frame information containing frame image information, and audio information. The generation module 1604 is used to generate response information corresponding to the multimodal information based on the multimodal information using a generation model. The type of response information includes at least one of the following: text information, image information, video information, and voice information. The storage module 1606 is used to store the data corresponding to the multimodal information and the data corresponding to the response information in the blockchain. The blockchain is used to store the data corresponding to different query information and the data corresponding to the corresponding response information. The corresponding response information is information generated by the generation model based on different query information.

[0210] It should be noted that the input module 1602, generation module 1604, and storage module 1606 mentioned above correspond to steps S902 to S906 in Embodiment 6. The modules and corresponding steps implement the same instances and application scenarios, but are not limited to the content disclosed in Embodiment 1. It should be noted that the above modules or units can be hardware or software components stored in memory and processed by one or more processors. The above modules can also run as part of the device in the server 10 provided in Embodiment 1.

[0211] Example 14

[0212] Embodiments of this application may provide a computer terminal, which may be any computer terminal device in a group of computer terminals. Optionally, in this embodiment, the aforementioned computer terminal may also be replaced with a mobile terminal or other terminal device.

[0213] Optionally, in this embodiment, the computer terminal may be located in at least one of a plurality of network devices in a computer network.

[0214] In this embodiment, the computer terminal described above can execute the program code for the following steps in the data processing method: verifying the first response information displayed in the generative interactive interface to obtain a first verification result, wherein the first verification result is used to characterize whether the content of the first response information is secure, and the first response information is used to characterize the information responding to the first query information input in the generative interactive interface; if the first verification result indicates that the content of the first response information is insecure, and the first response information is information generated by the generative model based on the first query information, determining the first query information based on data stored in the blockchain, wherein the blockchain is used to store data corresponding to different query information and data corresponding to the corresponding response information, and the corresponding response information is information generated by the generative model based on different query information; adjusting the original training data of the generative model based on the first query information to obtain target training data, wherein the target training data is used to adjust the model parameters of the generative model.

[0215] Optionally, Figure 17 This is a structural block diagram of a computer terminal according to an embodiment of this application. Figure 17 As shown, the computer terminal A may include: one or more (only one is shown in the figure) processors 1702, memory 1704, memory controller, and peripheral interfaces, wherein the peripheral interfaces are connected to a radio frequency module, an audio module, and a display.

[0216] The memory can be used to store software programs and modules, such as the program instructions / modules corresponding to the data processing method and apparatus in this application embodiment. The processor executes various functional applications and data processing by running the software programs and modules stored in the memory, thereby implementing the aforementioned data processing method. The memory may include high-speed random access memory, and may also include non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some instances, the memory may further include memory remotely located relative to the processor, and these remote memories can be connected to terminal A via a network. Examples of such networks include, but are not limited to, the Internet, corporate intranets, local area networks, mobile communication networks, and combinations thereof.

[0217] The processor can invoke information and applications stored in the memory via a transmission device to perform the following steps: verifying the first response information displayed in the generative interactive interface to obtain a first verification result, wherein the first verification result is used to characterize whether the content of the first response information is secure, and the first response information is used to characterize the information responding to the first query information input in the generative interactive interface; if the first verification result indicates that the content of the first response information is insecure, and the first response information is information generated by the generative model based on the first query information, determining the first query information based on data stored in the blockchain, wherein the blockchain is used to store data corresponding to different query information and data corresponding to the corresponding response information, and the corresponding response information is information generated by the generative model based on different query information; adjusting the original training data of the generative model based on the first query information to obtain target training data, wherein the target training data is used to adjust the model parameters of the generative model.

[0218] Optionally, the processor may also execute program code for the following steps: reading data corresponding to the first query information stored in the blockchain based on the data corresponding to the first response information, and determining the data corresponding to the first query information as the first query information; determining the data corresponding to the first response information, reading the data corresponding to the first query information stored in the blockchain based on the data corresponding to the first response information, and reading the first query information from a preset storage space based on the data corresponding to the first query information, wherein the preset storage space is used to store data corresponding to different query information and different query information.

[0219] Optionally, the processor may also execute program code that performs the following steps: determining the data corresponding to the first reply information, including: generating a hash value for the first reply information to obtain the data corresponding to the first reply information.

[0220] Optionally, the processor may also execute program code with the following steps: verifying the first query information to obtain a second verification result, wherein the second verification result is used to characterize whether the first query information causes the content of the first response information to be insecure; if the second verification result indicates that the first query information causes the content of the first response information to be insecure, using the first query information and the first response information as negative samples and adding the negative samples to the original training data to obtain target training data; if the second verification result indicates that the first query information does not cause the content of the first response information to be insecure, cleaning the original training data to obtain target training data.

[0221] Optionally, the processor may also execute program code for the following steps: detecting the first response information to obtain a detection result corresponding to the first response information, wherein the detection result is used to characterize whether the first watermark content has been added to the first response information; if the detection result indicates that the first watermark content has been added to the first response information, extracting the first watermark content added to the first response information; and based on the first watermark content, determining whether the first response information is information generated by a generative model based on the first query information.

[0222] Optionally, the processor may also execute program code that performs the following steps: obtaining the second query information displayed in the generative interactive interface; generating the second response information based on the second query information using a generative model; displaying the second response information in the generative interactive interface; and storing the data corresponding to the second query information and the data corresponding to the second response information in the blockchain.

[0223] Optionally, the processor may also execute program code that performs the following steps: displaying abbreviated information corresponding to the second response information in a generative interactive interface, wherein the amount of information in the abbreviated information is less than the amount of information in the second response information; outputting prompt information in the generative interactive interface, wherein the prompt information is used to indicate whether to store the data corresponding to the second query information and the data corresponding to the second response information in the blockchain; and responding to a confirmation command acting on the generative interactive interface, displaying the second response information in the generative interactive interface and storing the data corresponding to the second query information and the data corresponding to the second response information in the blockchain, wherein the confirmation command is used to indicate confirmation of storing the data corresponding to the second query information and the data corresponding to the second response information in the blockchain.

[0224] Optionally, the processor may also execute program code that performs the following steps: if no confirmation instruction is received within a preset time period, the second response information is prohibited from being displayed in the generative interactive interface, and the data corresponding to the second query information and the data corresponding to the second response information are prohibited from being stored in the blockchain.

[0225] Optionally, the processor may also execute program code that performs the following steps: obtains the second watermark content corresponding to the second reply information; adds the second watermark content to the second reply information using a hidden watermark method to obtain the target reply information; and displays the target reply information in a generative interactive interface.

[0226] Optionally, the processor may also execute program code with the following steps: when the first response information is information generated by a generative model based on the first query information, controlling the attribute of the generative interactive interface to be a first attribute, wherein the attribute includes at least one of the following: the interface color of the generative interactive interface, the interface size of the generative interactive interface, the font of the content displayed in the generative interactive interface, the content color of the displayed content, and the content size of the displayed content; when the first response information is not information generated by a generative model based on the first query information, controlling the attribute of the generative interactive interface to a second attribute; in response to a confirmation command, controlling the attribute of the generative interactive interface to a third attribute; and after adding the second watermark content to the second response information, controlling the attribute of the generative interactive interface to a fourth attribute.

[0227] Optionally, the processor may also execute program code that performs the following steps: in response to a confirmation instruction, or after adding a second watermark to the second response information, archives the second query information to obtain an archive record; summarizes the archive record and the second response information, or summarizes the archive record and the target response information to obtain copyright information; and stores the copyright information.

[0228] Optionally, the processor may also execute program code for the following steps: sending a second query message to a cloud server and receiving a second response message returned by the cloud server, wherein the generation model is deployed in the cloud server; obtaining the generation model by calling a preset interface and generating the second response message based on the second query message using the generation model; sending the second query message to the cloud server, receiving the first semantic analysis result corresponding to the second query message returned by the cloud server, determining the target data corresponding to the first semantic analysis result using the generation model, sending the target data to the cloud server, receiving the second semantic analysis result corresponding to the target data returned by the cloud server, and generating the second response message based on the first and second semantic analysis results using the generation model.

[0229] In this embodiment, a first verification result is obtained by verifying the first response information displayed in the generative interactive interface. The first verification result is used to characterize whether the content of the first response information is safe, and the first response information is used to characterize the information that responds to the first query information input in the generative interactive interface. If the first verification result indicates that the content of the first response information is unsafe, and the first response information is generated by the generative model based on the first query information, the first query information is determined based on the data stored in the blockchain. The blockchain is used to store data corresponding to different query information and data corresponding to the corresponding response information, and the corresponding response information is generated by the generative model based on different query information. Based on the first query information, the original training data of the generative model is adjusted to obtain target training data, whereby the target training data is used to adjust the model parameters of the generative model. It is noteworthy that by using blockchain to store the data corresponding to different query information and the corresponding response information, when it is determined that the content of the first response information is insecure, and the first response information is generated by the generative model based on the first query information, the reason for the insecurity of the content of the first response information can be accurately located based on the first query information, and the parameters of the generative model can be adjusted. This improves the security of content generation using large models and solves the technical problem of low security when using large models for content generation.

[0230] Those skilled in the art will understand that the structure shown in the figure is for illustrative purposes only, and the computer terminal may also be a smartphone (such as an Android phone, an iOS phone, etc.), a tablet computer, a handheld computer, a mobile internet device (MID), a PAD, and other terminal devices. Figure 17 This does not limit the structure of the aforementioned electronic device. For example, computer terminal A may also include components that are more... Figure 17 The more or fewer components shown (such as network interfaces, display devices, etc.), or having the same Figure 17 The different configurations shown.

[0231] Those skilled in the art will understand that all or part of the steps in the various methods of the above embodiments can be implemented by a program instructing the hardware related to the terminal device. The program can be stored in a computer-readable storage medium, which may include: flash drive, read-only memory (ROM), random access memory (RAM), disk or optical disk, etc.

[0232] Example 15

[0233] Embodiments of this application also provide a storage medium. Optionally, in this embodiment, the storage medium can be used to store the program code executed by the data processing method provided in Embodiment 1.

[0234] Optionally, in this embodiment, the storage medium may be located in any computer terminal in a group of computer terminals in a computer network, or in any mobile terminal in a group of mobile terminals.

[0235] Optionally, in this embodiment, the storage medium is configured to store program code for performing the following steps: verifying the first response information displayed in the generative interactive interface to obtain a first verification result, wherein the first verification result is used to characterize whether the content of the first response information is secure, and the first response information is used to characterize the information responding to the first query information input in the generative interactive interface; if the first verification result indicates that the content of the first response information is insecure, and the first response information is information generated by the generative model based on the first query information, determining the first query information based on the data stored in the blockchain, wherein the blockchain is used to store data corresponding to different query information and data corresponding to the corresponding response information, and the corresponding response information is information generated by the generative model based on different query information; adjusting the original training data of the generative model based on the first query information to obtain target training data, wherein the target training data is used to adjust the model parameters of the generative model.

[0236] The sequence numbers of the embodiments in this application are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.

[0237] In the above embodiments of this application, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0238] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. The device embodiments described above are merely illustrative; for example, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection of units or modules may be electrical or other forms.

[0239] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0240] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0241] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as a USB flash drive, read-only memory (ROM), random access memory (RAM), portable hard drive, magnetic disk, or optical disk.

[0242] The above description is only a preferred embodiment of this application. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of this application, and these improvements and modifications should also be considered within the scope of protection of this application.

Claims

1. A data processing method, characterized in that, include: The first response information displayed in the generative interactive interface is verified to obtain a first verification result, wherein the first verification result is used to characterize whether the content of the first response information is safe, and the first response information is used to characterize the information that responds to the first query information input in the generative interactive interface; If the first verification result indicates that the content of the first response information is insecure, and the first response information is generated by the generation model based on the first query information, the first query information is determined based on the data stored in the blockchain. The blockchain is used to store data corresponding to different query information and data corresponding to the corresponding response information. The corresponding response information is generated by the generation model based on the different query information. Based on the first query information, the original training data of the generative model is adjusted to obtain target training data. The target training data is used to adjust the model parameters of the generative model. The target training data is obtained by adding the first query information and the first response information to the original training data, or by cleaning the original training data.

2. The method according to claim 1, characterized in that, Based on the data stored in the blockchain, the first query information is determined, including one of the following: Based on the data corresponding to the first response information, read the data corresponding to the first query information stored in the blockchain, and determine that the data corresponding to the first query information is the first query information; The data corresponding to the first response information is determined, and the data corresponding to the first query information stored in the blockchain is read based on the data corresponding to the first response information. The first query information is then read from a preset storage space based on the data corresponding to the first query information, wherein the preset storage space is used to store the data corresponding to the different query information and the different query information.

3. The method according to claim 1, characterized in that, Based on the first query information, the original training data of the generative model is adjusted to obtain the target training data, including: The first query information is verified to obtain a second verification result, wherein the second verification result is used to characterize whether the first query information causes the content of the first response information to be insecure; If the second verification result indicates that the first query information causes the content of the first response information to be insecure, the first query information and the first response information are used as negative samples, and the negative samples are added to the original training data to obtain the target training data. If the second verification result indicates that the first query information did not cause the content of the first response information to be unsafe, the original training data is cleaned to obtain the target training data.

4. The method according to claim 1, characterized in that, The method further includes: The first reply information is detected to obtain the detection result corresponding to the first reply information, wherein the detection result is used to characterize whether the first watermark content is added to the first reply information; If the detection result indicates that the first watermark content is added to the first reply information, the first watermark content added to the first reply information is extracted. Based on the content of the first watermark, determine whether the first response information is generated using the generation model based on the first query information.

5. The method according to claim 1, characterized in that, The method further includes: Obtain the second query information displayed in the generative interactive interface; The generation model is used to generate a second response based on the second query information; The second response information is displayed in the generative interactive interface, and the data corresponding to the second query information and the data corresponding to the second response information are stored in the blockchain.

6. The method according to claim 5, characterized in that, Displaying the second response information in the generative interactive interface, and storing the data corresponding to the second query information and the data corresponding to the second response information in the blockchain, including: The generative interactive interface displays abbreviated information corresponding to the second response information, wherein the amount of information in the abbreviated information is less than the amount of information in the second response information; The generative interactive interface outputs a prompt message, which is used to prompt whether to store the data corresponding to the second query message and the data corresponding to the second response message in the blockchain; In response to a confirmation command applied to the generative interactive interface, the second response information is displayed in the generative interactive interface, and the data corresponding to the second query information and the data corresponding to the second response information are stored in the blockchain. The confirmation command is used to indicate confirmation of storing the data corresponding to the second query information and the data corresponding to the second response information in the blockchain.

7. The method according to claim 6, characterized in that, If the confirmation instruction is not received within a preset time period, the second response information shall not be displayed in the generative interactive interface, and the data corresponding to the second query information and the data corresponding to the second response information shall not be stored in the blockchain.

8. The method according to claim 5, characterized in that, The method further includes: Obtain the second watermark content corresponding to the second reply information; The target reply information is obtained by adding the second watermark content to the second reply information using a hidden watermark method. The target response information is displayed in the generative interactive interface.

9. The method according to claim 7 or 8, characterized in that, The method further includes: When the first response information is generated by a generative model based on the first query information, the attribute of the generative interactive interface is controlled to be a first attribute, wherein the attribute includes at least one of the following: the interface color of the generative interactive interface, the interface size of the generative interactive interface, the font of the content displayed in the generative interactive interface, the content color of the displayed content, and the content size of the displayed content. If the first response information is not generated using a generative model based on the first query information, the attribute of the generative interactive interface is controlled to be the second attribute. In response to the confirmation command, the attribute of the generative interactive interface is controlled to be a third attribute; After adding the second watermark content to the second reply information, the attribute of the generative interactive interface is controlled to be the fourth attribute.

10. The method according to claim 7 or 8, characterized in that, The method further includes: In response to a confirmation command, or after adding a second watermark to the second reply information, the second query information is archived to obtain an archived record; The archived records and the second response information are summarized, or the archived records and the target response information are summarized, to obtain copyright information; The copyright information shall be preserved.

11. The method according to claim 5, characterized in that, The generation model generates a second response based on the second query information, including at least one of the following: Send the second query information to the cloud server and receive the second response information returned by the cloud server, wherein the generation model is deployed in the cloud server; The generated model is obtained by calling a preset interface, and the second response information is generated based on the second query information using the generated model; Send the second query information to the cloud server, receive the first semantic analysis result corresponding to the second query information returned by the cloud server, use the generative model to determine the target data corresponding to the first semantic analysis result, send the target data to the cloud server, receive the second semantic analysis result corresponding to the target data returned by the cloud server, and use the generative model to generate the second response information based on the first semantic analysis result and the second semantic analysis result.

12. A data processing method, characterized in that, include: Retrieve the first query information displayed in the generative interactive interface; A generative model is used to generate a first response based on the first query information. The model parameters of the generative model are adjusted based on the target training data. If the first query information makes the content of the first response information unsafe, the target training data is obtained by adding the first query information and the first response information to the original training data. If the first query information does not make the content of the first response information unsafe, the target training data is obtained by cleaning the original training data. The first response information is displayed in the generative interactive interface; The data corresponding to the first query information and the data corresponding to the first response information are stored in the blockchain. The blockchain is used to store the data corresponding to different query information and the data corresponding to the corresponding response information. The corresponding response information is generated by the generation model based on the different query information.

13. The method according to claim 12, characterized in that, Displaying the first response information in the generative interactive interface, and storing the data corresponding to the first query information and the data corresponding to the first response information in the blockchain, including: The generative interactive interface displays abbreviated information corresponding to the first response information, wherein the amount of information in the abbreviated information is less than the amount of information in the first response information; The generative interactive interface outputs a prompt message, which is used to prompt whether to store the data corresponding to the first query message and the data corresponding to the first reply message in the blockchain. In response to a confirmation command applied to the generative interactive interface, the first response information is displayed in the generative interactive interface, and the data corresponding to the first query information and the data corresponding to the first response information are stored in the blockchain. The confirmation command is used to indicate confirmation of storing the data corresponding to the first query information and the data corresponding to the first response information in the blockchain.

14. The method according to claim 13, characterized in that, If the confirmation instruction is not received within a preset time period, the first response information shall not be displayed in the generative interactive interface, and the data corresponding to the first query information and the data corresponding to the first response information shall not be stored in the blockchain.

15. The method according to claim 12, characterized in that, The method further includes: Obtain the first watermark content corresponding to the first reply information; The target reply information is obtained by adding the first watermark content to the first reply information using a hidden watermark method. The target response information is displayed in the generative interactive interface.

16. The method according to claim 12, characterized in that, The generation model generates a first response based on the first query information, including at least one of the following: Send the first query information to the cloud server and receive the first response information returned by the cloud server, wherein the generation model is deployed in the cloud server; The generated model is obtained by calling a preset interface, and the first response information is generated based on the first query information using the generated model; The first query information is sent to the cloud server, the first semantic analysis result corresponding to the first query information is received from the cloud server, the target data corresponding to the first semantic analysis result is determined using the generative model, the target data is sent to the cloud server, the first semantic analysis result corresponding to the target data is received from the cloud server, and the first response information is generated based on the first semantic analysis result and the first semantic analysis result using the generative model.

17. A data processing method, characterized in that, include: Inputting multimodal information in the dialog interface, wherein the type of multimodal information includes at least one of the following: text information containing character information, video frame information containing frame image information, and audio information; A generative model is used to generate response information corresponding to the multimodal information based on the multimodal information. The response information includes at least one of the following types: text information, image information, video information, and voice information. The model parameters of the generative model are adjusted based on the target training data. If the first query information causes the content of the first response information to be insecure, the target training data is obtained by adding the first query information and the first response information to the original training data. If the first query information does not cause the content of the first response information to be insecure, the target training data is obtained by cleaning the original training data. The data corresponding to the multimodal information and the data corresponding to the response information are stored in the blockchain. The blockchain is used to store the data corresponding to different query information and the data corresponding to the corresponding response information. The corresponding response information is generated by the generation model based on the different query information.

18. A data processing system, characterized in that, include: A front-end client is used to display a generative interactive interface and capture the first query information input in the generative interactive interface; A backend server, connected to the frontend client, is used to generate a first response based on the first query information using a generative model. The model parameters of the generative model are adjusted based on the target training data. If the first query information causes the content of the first response to be insecure, the target training data is obtained by adding the first query information and the first response to the original training data. If the first query information does not cause the content of the first response to be insecure, the target training data is obtained by cleaning the original training data. A blockchain, connected to the backend server, is used to store the data corresponding to the first query information and the data corresponding to the first response information.

19. An electronic device, characterized in that, include: Memory, which stores executable programs; A processor for running the program, wherein the program, when running, performs the method according to any one of claims 1 to 18.

20. A computer-readable storage medium, characterized in that, The computer-readable storage medium includes a stored executable program, wherein, when the executable program is executed, it controls the device on which the computer-readable storage medium is located to perform the method according to any one of claims 1 to 18.

Images