Information processing method and device, equipment and storage medium

Abstract

The embodiment of the invention provides an information processing method and device, equipment and a storage medium. The method comprises the following steps: identifying one or more target text items corresponding to predetermined type information from first cue word information indicating a target task; and respectively replacing one or more target text items in the first cue word information with one or more predetermined marks to generate second cue word information. The predetermined mark indicates the attribute information of the target text item replaced by the predetermined mark. And based on the second cue word information, utilizing a machine learning model to execute the target task to obtain a task execution result. In this way, normal execution of the task can be ensured. Meanwhile, the predetermined type of information is prevented from being acquired or leaked by an unauthorized third party in the task execution process, and the safety of user information is improved.

Description

Technical Field

[0001] Example embodiments of the present disclosure generally relate to the field of computers, and more particularly, to methods, apparatuses, devices, and computer-readable storage media for information processing. Background Art

[0002] With the development of information technology, various terminal devices can provide various services to people in work and life. Applications that provide services can be deployed in terminal devices. Terminal devices present corresponding content and interact with users through the user interface of applications to meet various needs of users. Terminal devices or applications can provide users with digital assistant functions to support better interaction with users. During the interaction, users can send messages to digital assistants and put forward requirements to digital assistants. Summary of the invention

[0003] In a first aspect of the present disclosure, a method for information processing is provided. The method includes: identifying one or more target content items corresponding to predetermined type information from first prompt word information indicating a target task, the first prompt word information being generated based on a user input requesting the target task; replacing the one or more target content items in the first prompt word information with one or more predetermined tags, respectively, to generate second prompt word information, a predetermined tag in the one or more predetermined tags indicating attribute information of the target content item replaced by the predetermined tag; and executing the target task using a machine learning model based on the second prompt word information to obtain a task execution result.

[0004] In a second aspect of the present disclosure, a device for information processing is provided. The device includes: an identification module configured to identify one or more target content items corresponding to predetermined type information from first prompt word information indicating a target task, wherein the first prompt word information is generated based on a user input requesting the target task; a replacement module configured to replace one or more target content items in the first prompt word information with one or more predetermined tags, respectively, to generate second prompt word information, wherein a predetermined tag in the one or more predetermined tags indicates attribute information of the target content item replaced by the predetermined tag; and an execution module configured to execute the target task using a machine learning model based on the second prompt word information to obtain a task execution result.

[0005] In a third aspect of the present disclosure, an electronic device is provided. The device includes at least one processing unit; and at least one memory, the at least one memory is coupled to the at least one processing unit and stores instructions for execution by the at least one processing unit. When the instructions are executed by the at least one processing unit, the device executes the method of the first aspect.

[0006] In a fourth aspect of the present disclosure, a computer-readable storage medium is provided, wherein a computer program is stored on the computer-readable storage medium, and the computer program can be executed by a processor to implement the method of the first aspect.

[0007] It should be understood that the contents described in this content section are not intended to limit the key features or important features of the embodiments of the present disclosure, nor are they intended to limit the scope of the present disclosure. Other features of the present disclosure will become easily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0008] The above and other features, advantages and aspects of the embodiments of the present disclosure will become more apparent with reference to the following detailed description in conjunction with the accompanying drawings. In the accompanying drawings, the same or similar reference numerals represent the same or similar elements, wherein:

[0009] Figure 1 A schematic diagram showing an example environment in which embodiments according to the present disclosure may be implemented;

[0010] Figure 2 A schematic diagram showing an example architecture for information processing according to some embodiments of the present disclosure;

[0011] Figure 3 A schematic diagram showing an example interface for information processing according to some embodiments of the present disclosure;

[0012] Figure 4 A flowchart illustrating an example process for information processing according to some embodiments of the present disclosure;

[0013] Figure 5 A schematic structural block diagram showing an example apparatus for information processing according to some embodiments of the present disclosure; and

[0014] Figure 6 A block diagram of an electronic device capable of implementing various embodiments of the present disclosure is shown. DETAILED DESCRIPTION

[0015] Embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although certain embodiments of the present disclosure are shown in the accompanying drawings, it should be understood that the present disclosure can be implemented in various forms and should not be construed as being limited to the embodiments set forth herein. On the contrary, these embodiments are provided to provide a more thorough and complete understanding of the present disclosure. It should be understood that the drawings and embodiments of the present disclosure are only for exemplary purposes and are not intended to limit the scope of protection of the present disclosure.

[0016] It should be noted that the titles of any sections / subsections provided herein are not restrictive. Various embodiments are described throughout this article, and any type of embodiment may be included under any section / subsection. In addition, the embodiments described in any section / subsection may be combined in any manner with any other embodiments described in the same section / subsection and / or different sections / subsections.

[0017] In the description of the embodiments of the present disclosure, the term "including" and similar terms should be understood as open inclusion, that is, "including but not limited to". The term "based on" should be understood as "based at least in part on". The term "one embodiment" or "the embodiment" should be understood as "at least one embodiment". The term "some embodiments" should be understood as "at least some embodiments". Other explicit and implicit definitions may be included below. The terms "first", "second", etc. may refer to different or the same objects. Other explicit and implicit definitions may be included below.

[0018] The embodiments of the present disclosure may involve user data, data acquisition and / or use, etc. These aspects are subject to the corresponding laws, regulations and relevant provisions. In the embodiments of the present disclosure, all data collection, acquisition, processing, processing, forwarding, use, etc. are carried out on the premise that the user knows and confirms. Accordingly, when implementing each embodiment of the present disclosure, the type, scope of use, usage scenario, etc. of the data or information that may be involved should be informed to the user and the user's authorization should be obtained in an appropriate manner in accordance with the relevant laws and regulations. The specific notification and / or authorization method can vary according to the actual situation and application scenario, and the scope of the present disclosure is not limited in this respect.

[0019] In this specification and the embodiments, if personal information processing is involved, it will be processed on the premise of having a legal basis (such as obtaining the consent of the subject of personal information, or it is necessary to perform a contract, etc.), and will only be processed within the scope of regulations or agreements. If a user refuses to process personal information other than the necessary information for basic functions, it will not affect the user's use of basic functions.

[0020] Example Environment

[0021] Figure 1 1 is a schematic diagram of an example environment 100 in which embodiments of the present disclosure can be implemented. Figure 1 As shown, example environment 100 may include electronic device 110 .

[0022] In this example environment 100, the electronic device 110 can run an application 120 that provides information query services for the user 140. The application 120 can be any appropriate type of application for information query, and its examples can include but are not limited to: a digital assistant or other appropriate applications. The user 140 can interact with the application 120 via the electronic device 110 and / or its attached device. In some embodiments, the application 120 can provide or be configured with a language model to provide services to the user 140. In some embodiments, the third-party model 160 can be a machine learning model, a deep learning model, a learning model, a neural network, etc. In some embodiments, the third-party model 160 may include a language model, such as a large language model (LLM). The large language model can have question-answering capabilities by learning from a large amount of corpus. The third-party model 160 can also be based on other appropriate models. Although shown as being independent of the server 130, the third-party model 160 can run on the server 130, or other remote servers. In some embodiments, the application 120 can be used to determine whether the feedback of the third-party model 160 meets the user's expectations. The third-party model 160 refers to a model provided by other applications or platforms independent of the application 120.

[0023] exist Figure 1 In the environment 100 of the embodiment of the present invention, if the application 120 is in an active state, the electronic device 110 can present an interface 150 for supporting the acquisition of user input and presenting the interaction result to the user 140 through the application 120. The interface 150 may include, for example, a conversation interface between the user 140 and the application 120. The user 140 may send user input indicating a target task to the application 120 through the conversation interface, etc. During the interaction, the application 120 provides the acquired user input to the third-party model 160 to obtain the response of the third-party model 160 to the user input. Subsequently, the application 120 provides services to the user 140 based on the response of the third-party model 160. The interface 150 may be used to present a response to the user input. In some embodiments, depending on the configuration of the application 120, the interactive message with the application 120 may include a message in a multimodal form, such as a text message (e.g., a natural language text), a voice message, an image message, a video message, and the like.

[0024] In some embodiments, the application 120 may be associated with a corresponding database, in which data or information required for the application 120 to answer user interaction information is stored. Exemplarily, the application 120 may obtain information indicated by the user from a database connected to the application 120 (e.g., a database for storing data of the user 140, or a knowledge base for storing historical interaction information between the user 140 and the application 120) in response to user input. The application 120 may provide the acquired operation data to the third-party model 160, so that the machine learning model can provide corresponding services to the user according to the user output.

[0025] In some embodiments, the electronic device 110 communicates with the server 130 to provide services for the application 120. The electronic device 110 can be any type of mobile terminal, fixed terminal or portable terminal, including a mobile phone, a desktop computer, a laptop computer, a notebook computer, a netbook computer, a tablet computer, a media computer, a multimedia tablet, a handheld computer, a portable game terminal, a VR / AR device, a personal communication system (PCS) device, a personal navigation device, a personal digital assistant (PDA), an audio / video player, a digital camera / camcorder, a positioning device, a television receiver, a radio receiver, an e-book device, a game device, or any combination of the foregoing, including accessories and peripherals of these devices or any combination thereof. In some embodiments, the electronic device 110 can also support any type of interface for the user (such as a "wearable" circuit, etc.).

[0026] The server 130 may be an independent physical server, or a server cluster or distributed system composed of multiple physical servers, or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, content distribution networks, and big data and artificial intelligence platforms. The server 130 may include, for example, a computing system / server, such as a mainframe, an edge computing node, a computing device in a cloud environment, etc. The server 130 may provide background services for the application 120 that supports information query in the electronic device 110.

[0027] A communication connection may be established between the server 130 and the electronic device 110. The communication connection may be established in a wired manner or a wireless manner. The communication connection may include, but is not limited to, a Bluetooth connection, a mobile network connection, a Universal Serial Bus (USB) connection, a Wireless Fidelity (WiFi) connection, etc., and the embodiments of the present disclosure are not limited in this respect. In the embodiments of the present disclosure, the server 130 and the electronic device 110 may implement signaling interaction through the communication connection between the two.

[0028] It should be understood that the structure and function of the various elements in the environment 100 are described for exemplary purposes only and do not imply any limitation on the scope of the present disclosure.

[0029] As briefly mentioned above, a terminal device or application can provide services (such as information query, text processing, etc.) to users (e.g., corporate users) through a third-party model. In the process of calling a third-party model, the terminal device or application provides information corresponding to the user input to the third-party model to use the third-party model to provide services to the user. However, in some scenarios where a third-party model is called (e.g., in scenarios for corporate users), user information (e.g., internal corporate information, etc.) may be intercepted or accessed by unauthorized third parties, resulting in information leakage.

[0030] In view of this, an embodiment of the present disclosure provides an information processing scheme. In the scheme, one or more target content items corresponding to predetermined type information are identified from the first prompt word information indicating the target task. The one or more target content items in the first prompt word information are replaced with one or more predetermined tags respectively to generate second prompt word information. The predetermined tag in the one or more predetermined tags indicates the attribute information of the target content item replaced by the predetermined tag. Based on the second prompt word information, the target task is executed using a machine learning model to obtain a task execution result.

[0031] It will be more clearly understood through the following description that according to the scheme of the present disclosure, the target content item corresponding to the predetermined type information in the first prompt word information is first replaced with a predetermined mark to generate the second prompt word information. The predetermined type information can be personal identity information or information that is easy to expose the user's personal data, and the predetermined mark indicates the attribute information of the target content item it replaces. In this way, the second prompt word information used to perform the target task does not include such predetermined type information, while retaining the original attribute information. In this way, the normal execution of the task can be guaranteed, while preventing the predetermined type information from being obtained or leaked by an unauthorized third party during the execution of the task. Improve the security of user information.

[0032] Various example implementations of the solution are described in detail below in conjunction with the accompanying drawings.

[0033] Figure 2 A schematic diagram of an example architecture 200 for information processing according to some embodiments of the present disclosure is shown. The actions or operations described below with reference to the target application 210 can be implemented by the electronic device 110, or by the server 130, or by the server 130 and the electronic device 110 in coordination. In some embodiments, the user 140 obtains the services provided by the third-party model through the application 120 running at the electronic device 110. The user 140 provides the user input 220 indicating the target task to the target application 210. The user input 220 can be in the form of natural language (such as text input or voice input, etc.). The target application 210 runs at the electronic device 110 and interacts with the user through the interface 150. Figure 3FIG. 3 is a schematic diagram showing an example interface 300 for information processing according to some embodiments of the present disclosure. Figure 3 As shown, multiple applications are running in the terminal device 110, including at least a first application 320-1, a second application 320-2, a third application 320-3 and a fourth application 320-4, which can be individually or collectively referred to as applications 320. Figure 3 As shown, in response to the user 140 selecting a certain application (eg, clicking, etc.), the terminal device presents the user with an interface 320 corresponding to the application. The application is the target application 210. The target application 210 can obtain the user input 220 provided by the user 140 through the interface 320.

[0034] like Figure 2 As shown, user input 220 is provided to target application 210. Target application 210 performs information processing operations on user input 220. User input 220 indicates the target task that the user wants to perform (e.g., answering questions, text processing, chart generation, and information query, etc.). User input 220 may include user data, such as personal information such as the user's identity, address, and contact information. User input 220 may include internal knowledge of the enterprise, etc. In some embodiments, target application 210 may perform target tasks based on data related to the user to provide high-quality services to the user. For example, target application 210 may determine information about objects related to the user (such as historical browsing records, user preference information, etc.) based on user input 220 and provide services to the user. Target application 210 may generate a response to user input 220 based on internal knowledge of the enterprise related to user input 220.

[0035] In some embodiments, the target application 210 may determine context information 223 related to the user input 220 based on the user input 220. The context information 223 indicates an object related to the user input 220. Different user inputs 220 correspond to different objects. Exemplarily, if the user input 220 indicates a text processing task, the object related to the user input 220 may be a reference document (e.g., a user uploaded document, internal knowledge of an enterprise, etc.), a text processing template, etc. If the user input 220 indicates a commodity recommendation task, the object related to the user input 220 may be user preference information, a user account balance, etc. The object indicated by the context information 223 is not limited here. In some embodiments, the object may correspond to the user input 220, for example, it may correspond to the content included in the user input 220. Exemplarily, if the user input 220 includes "My name is Li XX, and my address is B Street, City A", then the objects corresponding to the user input 220 include "Li XX", "City A" and "B Street". In some embodiments, the object may correspond to the user who provides the user input 220. For example, the user input 220 is provided by user X (user ID is 000x1) of the target application 210 , and the object may correspond to user X. User X is the target user 221 corresponding to the user input 220 .

[0036] In some embodiments, objects related to the user may be stored in a database 222 of a target application 210. The target application 210 first determines the identity of a target user 221 corresponding to the user input 220. Subsequently, based on the identity of the target user 221, an object corresponding to the target user and the content included in the user input 220 is determined in the database of the target application 210. The target application 210 first obtains relevant information about the user (e.g., the user's name, email address, address, etc.) from the database of the application. These data have been stored in the database during the user's registration or use of the application 320. The target application 210 uses the information in the database to help identify predetermined types of information, thereby enhancing the accuracy of privacy protection. The target application 210 generates context information 223 by combining the database information with the results of entity recognition, and can better identify predetermined types of information related to the application scenario. For example, for a social media application, the user's personal information, published content, and even the user's interaction records may all be information that needs to be processed. In a shopping application, the user's order records, payment information, and address information may be the main information that needs to be processed. Subsequently, the target application 210 generates first prompt word information 230 based on the user input 220 and context information 223 related to the user input 220 .

[0037] The first prompt word information 230 may be provided to the information processing platform 240. The information processing platform 240 identifies one or more target content items 243 corresponding to the predetermined type information from the first prompt word information 230. Exemplarily, the content items may include text, voice, image or video, etc. The predetermined type information may include various appropriate types of information that need to be protected. For example, the predetermined type information may be personal identity information or information that is easy to expose the user's personal data. For another example, the predetermined type information may include information that needs to be protected, such as knowledge within an organization (e.g., within an enterprise). In different application scenarios (e.g., different applications 320 or using a certain application to perform different tasks), the predetermined type information may also be different. For example, in the first application 320-1, "user name" is the predetermined type information that needs to be processed, and the predetermined type information includes "user name". In the second application 320-2 (non-real-name application), it is no longer necessary to process the user name. Exemplarily, the information processing platform 240 may first determine the information processing strategy corresponding to each application 320, and then determine the predetermined type information according to the corresponding information processing strategy.

[0038] In some embodiments, the information processing platform 240 can identify the target text item 243 corresponding to the predetermined type of information through pattern matching 241-1. Exemplarily, the information processing platform 240 can identify the target text item 243 in the first prompt word information 230 through a regular expression. For example, a regular expression is used to match the post-order text of "my address is" in the first prompt word information 230, and the post-order text is identified as a target text item 243 of the address type. In this way, the target text item 243 in the first prompt word information 230 can be detected through pattern matching, a low-latency, low-discrimination method.

[0039] In some embodiments, the information processing platform 240 can identify the target text item 243 through the entity recognition model. Exemplarily, the first prompt word information 230 is provided to the entity recognition model to obtain the output of the entity recognition model. Subsequently, based on the output of the entity recognition model, the target text item 243 in the first prompt word information 230 is determined. In this way, the target text item 243 that needs to be processed in the first prompt word information 230 can be more accurately identified.

[0040] In some embodiments, multiple entity recognition models of different types may be used to determine the target text item 243. For example, the first prompt word information 230 may be processed using an internal entity recognition model 241-3 and an entity recognition model provided by a third party. The number of parameters of the third-party entity recognition model 241-2 exceeds that of the internal entity recognition model 241-3 (i.e., the performance of the third-party entity recognition model 241-2 is better than that of the internal entity recognition model 241-3). The internal entity recognition model 241-3 may be a model trained based on the application scenario of the target application 210 to further improve the accuracy of entity recognition. Exemplarily, the model parameters of the internal entity recognition model 241-3 may be updated based on training context information related to the user training input. The training context information indicates a training object related to the user training input. In some embodiments, multiple entity recognition models may correspond to different information processing strategies, respectively. Exemplarily, multiple entity recognition models may be trained in advance according to the information processing strategies corresponding to different applications 320. Subsequently, based on the correspondence between different entity recognition models and different applications, the corresponding entity recognition model is determined to process the first prompt word information 230. Thus, a plurality of target text items 243 are determined from the first prompt word information 230 .

[0041] In some embodiments, the internal entity recognition model 241-3 may determine the target text item 243 from the first prompt word information 230 based on a predetermined reference content item. The reference content item may be a historical content item identified as corresponding to a predetermined type of information during the historical information processing process.

[0042] In some embodiments, the content item (i.e., the entity in the first prompt word information 230) determined by the information processing platform 240 based on pattern matching and / or entity recognition model may not match the entity corresponding to the target application 210. Therefore, the information processing platform 240 may first determine the candidate content item. For example, the information processing platform 240 may determine the first candidate content item 242-1 from the first prompt word information 230 by performing pattern matching corresponding to the predetermined type of information on the first prompt word information. The information processing platform 240 may determine the second candidate content item 242-2 based on the first prompt word information 230 using the entity recognition model. Subsequently, the target content item 243 is determined based on the first candidate content item 242-1 and the second candidate content item 242-2. In this way, the information processing platform 240 only replaces the entity that matches the target application 210.

[0043] Exemplarily, the information processing policy corresponding to the target application 210 may indicate that only the user name needs to be processed. If the candidate content item is "XX City", this content item does not need to be processed. In addition, even if the candidate content item is a name, but this name is not a user of the target application 210 (for example, the name is not registered in the target application 210, or the candidate content item does not represent a name), the candidate content item is not a target content item 243. For example, the user input 220 includes the address "Li XX City" ("Li XX" is a name), and the candidate content item does not need to be processed. Therefore, the information processing platform 240 can determine whether any candidate content item matches the entity corresponding to the target application 210. If it matches, the candidate content item is determined as one of the target content items 243. In this way, based on the current scenario (that is, the usage scenario of the target application 210), multiple target content items 243 are determined from multiple first candidate content items 242-1 and second candidate content items 242-2.

[0044] The above-described methods of using pattern matching, third-party entity recognition model, and internal entity recognition model can be used alone or in combination. In some embodiments, if multiple methods among the above three methods are used, content items determined by different methods may be repeated. To this end, in some embodiments, the first candidate content item 242-1, the second candidate content item 242-2, and the third candidate content item 242-3 (obtained using the internal entity recognition model 241-3) can be deduplicated to filter out duplicate candidate content items. In this way, the target content item 243 is determined more accurately.

[0045] In some scenarios (e.g., business-oriented scenarios), a large amount of text may need to be processed, resulting in the inability to accurately and quickly determine the target content item 243. Therefore, in order to improve the efficiency of determining the target content item 243, an internal entity recognition model 241-3 based on a Bloom filter (e.g., a counting bloom filter, etc.) can be used to determine the target content item 243 from the first prompt word information 230. In some embodiments, the information processing platform 240 uses the historical content item 244 identified as corresponding to the predetermined type of information as a reference content item. Based on the reference content item, a content item matching the reference content item is detected from the first prompt word information 230. Subsequently, the content item matching the reference content item is determined as the first target content item. For example, if "Alice" is a reference content item, "Alice" in the first prompt word information 230 is determined as the target content item 243.

[0046] In some embodiments, the reference content item is an entity (i.e., target content item 243) that has been determined to correspond to the predetermined type of information. The information processing platform 240 can update (such as, add, delete, or replace, etc.) the reference content item stored in the Bloom filter based on the target content item 243 determined in the process of processing information. Exemplarily, in an information processing process, if the target content item 243 determined by the information processing platform 240 does not match the existing reference content item, the target content item 243 can be added as a new reference content item. In addition, the information processing platform 240 can delete the reference content item when any reference content item does not correspond to the target application 210. For example, the Bloom filter includes a reference content item "Alice" for the target application 210. After the database of the target application 210 is updated, "Alice" is no longer a user of the target application 210, and the reference content item "Alice" is deleted. In this way, the electronic device can quickly determine the target content item 243, thereby processing the user input 220.

[0047] In some embodiments, the information processing platform 240 replaces the target content items 243 in the first prompt word information 230 with predetermined tags, respectively, to generate the second prompt word information 250. The information processing platform 240 replaces the target content items 243 in the first prompt word information 230 with predetermined tags. Exemplarily, the electronic device can use a predetermined tag dictionary 245 to save the correspondence between the predetermined tags and the target content items 243. Subsequently, based on the correspondence between the predetermined tags and the target content items 243, the predetermined tags are used to replace each target content item 243 in the first prompt word information 230 to generate the second prompt word information 250. At the same time, in order to ensure the quality of service provided to the user, it is necessary to enable the second prompt word information 250 generated after the replacement to accurately represent the target task corresponding to the user input 220. Therefore, the predetermined tag can indicate the attribute information (for example, semantics) of the replaced target content item 243. For example, if the target content item 243 indicates the user's name, a predetermined tag indicating "name" (for example,<Name_1> If the target content item 243 indicates address information, the target content item 243 is replaced with a predetermined tag indicating "address" (e.g.,<Location_1> The electronic device may use other types of predetermined tags to replace the target content item 243, which is not limited here. The predetermined tag dictionary 245 may be implemented based on the Redis key-value storage system.

[0048] In some embodiments, the information processing platform 240 performs the target task based on the second prompt word information 250. The information processing platform 240 can provide the second prompt word information 250 to other applications, platforms or third-party models 160. The third-party model 160 performs the target task based on the provided second prompt word information 250. Exemplarily, the third-party model 160 can be a language model. The second prompt word information 250 is provided to the third-party model 160 to obtain the execution result 260 of the third-party model 160. The execution result 260 may include text. In the process of the third-party model 160 performing the target task, the third-party model 160 only processes the replaced prompts and does not contact the predetermined type of information in the first prompt word information 230. In this way, the security of user information is improved.

[0049] The third-party model 160 cannot obtain the predetermined type information in the first prompt word information 230, so the execution result 260 output by the third-party model 160 may include one or more predetermined tags. The execution result 260 output by the third-party model 160 may be provided to the information processing platform 240. Subsequently, based on the correspondence between the target content item 243 and the predetermined tag, the information processing platform 240 replaces the predetermined tag included in the execution result 260 with the target content item 243 corresponding to the predetermined tag to generate an updated execution result 270. For example,<Name_1> The tag is replaced with "Alice". Subsequently, the information processing platform 240 provides the updated execution result 270 to the target application 210, so as to present the updated execution result 270 to the user through the target application 210 as a response to the user input 220. Figure 3 As shown, the information processing platform 240 can present the updated execution result 270 in the interface 320 corresponding to the target application 210. In some embodiments, the interface 320 also includes an entry 330 for or user input 220. The user can provide the user input 220 to the target application 210 through the entry 330. The user 140 can provide feedback on the updated execution result 270 to the target application 210 through the entry 330 (such as, evaluation of the updated execution result 270, update of the user input 220, etc.).

[0050] In some embodiments, in order to further improve the security of user information, the infrastructure layer and application layer of the information processing platform 240 can be implemented in a trusted execution environment (TEE) to ensure that all operations related to user information are performed in a secure and isolated environment.

[0051] Continue to refer Figure 2. In some embodiments, in order to further improve the security and reliability of the information processing process, the communicating sequential processes (CSP) can be removed from the user's trusted computing base (TCB). At the same time, the root file system of the container is additionally protected to prevent CSP from tampering with and injecting data into the mounted volume. In addition, an additional security layer is introduced to ensure that CSP cannot abuse permissions to access or modify confidential data. Exemplarily, for the target user related to the user input 220, a corresponding trusted execution environment (e.g., a container) can be created. Subsequently, the identification operation of the target content item 243 and the replacement operation of the target content item 243 are performed in the trusted execution environment. A tenant-defined execution policy is thereby implemented. The execution policy is embedded in an unmodifiable verification report when the system starts. The execution policy defines the operations that are allowed to be executed, including mounting / unmounting devices, receiving and processing original user prompts, interacting with third-party LLMs, and returning the final execution result 260. In this way, it is ensured that only authorized operations can be executed during system operation, thereby further reducing the potential attack surface.

[0052] At the same time, for the infrastructure layer of the information processing platform 240, Intel TDX (Trusted Domain Extensions) technology can be used to generate verification reports for container groups to ensure that the containers follow security policies during operation. By executing restricted policies, the security of the container life cycle is ensured. Furthermore, Kata-shim can start the information processing instance and perform key container operations in a secure virtual machine. This includes pulling images from the warehouse, starting pods and containers, and running user space code. This verification mechanism ensures the security and integrity of these operations, allowing container operations to run in a trusted environment, thereby ensuring the security of user data.

[0053] In some instances, the operations on the container can be limited, and only a specific set of operations is allowed, and any other operations are not supported. These operations may include, for example, mounting / unmounting devices (Mount / Unmount Devices), starting and stopping containers (Start / Stop Containers), receiving the original request of the user (Receive Raw Prompts from Users), sending masked requests to the third-party model 160 (Send Masked Prompts to Third-Party Large Language Models), processing and returning the original response of the third-party model 160 (Process and Return the LLM's Raw Response to the End User), etc. In some embodiments, it is first determined whether the operation instruction indicates one of multiple predetermined operations, and if one of the predetermined operations is indicated, the operation instruction is responded to. The above-mentioned execution policy defined by the tenant is embedded in the authentication report when the container is started, and cannot be modified during the operation of the system. In this way, it is ensured that the operation restrictions throughout the life cycle of the container are always protected, so that the execution policy cannot be bypassed or tampered by malicious attackers. The execution policy cannot be modified and can be verified during the system life cycle. In this way, the security of the container is enhanced, and the traceability and credibility of the system operation are also ensured. Exemplarily, the operations on the container supported by the execution policy are shown in Table 1.

[0054] Table 1

[0055]

[0056]

[0057] As shown in Table 1, in order to ensure the security of the trusted execution environment, it is necessary to perform identity authentication on the operation instructions for the trusted execution environment. In some embodiments, in response to the received operation instruction indicating one of a plurality of predetermined operations, an identity corresponding to the operation instruction is determined. Subsequently, based on the operation instruction for the trusted execution environment, an identity corresponding to the operation instruction is determined. If the identity matches the first user, the operation instruction is executed.

[0058] In summary, the present application provides an information processing solution that improves the security of user information by preventing predetermined types of information from being obtained or leaked by unauthorized third parties during the execution of tasks. The information processing platform 240 can identify and process target content items in user input and context information through advanced privacy protection components integrated in the system, combined with pattern matching and machine learning technology. By introducing this context-sensitive data classification method to determine the target content item, the detected user data is anonymized or replaced with a placeholder to ensure that user data can be protected during the data processing process. At the same time, the target content item of the response is replaced with a predetermined tag to further ensure that the data can still maintain the integrity and coherence of the context after being processed by an external large language model. Thereby ensuring that the output of the model is meaningful. After the third-party model generates the execution result, based on the correspondence between the predetermined tag and the target content item, the execution result of the third-party model is updated to restore the target content item to ensure the integrity and accuracy of the result.

[0059] Figure 4 A flow chart of an example process 400 of information processing according to some embodiments of the present disclosure is shown. For example, the process 400 can be implemented at the information processing platform 240, but this is merely exemplary.

[0060] like Figure 4 As shown, in block 410, the information processing platform 240 identifies one or more target content items corresponding to predetermined type information from first prompt word information indicating a target task, where the first prompt word information is generated based on user input requesting the target task.

[0061] In some embodiments, identifying one or more target content items corresponding to the predetermined type of information includes at least one of the following: determining at least a portion of the one or more target content items from the first prompt word information by performing pattern matching corresponding to the predetermined type of information on the first prompt word information, determining at least a portion of the one or more target content items based on the first prompt word information using one or more entity recognition models, or determining at least a portion of the one or more target content items from the first prompt word information based on one or more reference content items, wherein the one or more reference content items include historical content items identified as corresponding to the predetermined type of information.

[0062] In some embodiments, candidate content items are determined from prompt word information by pattern matching or utilizing one or more entity recognition models, and determining at least a portion of one or more target content items includes: determining whether the candidate content item matches one or more entities corresponding to a target application, in which the user input is issued; and in response to a given candidate content item matching the one or more entities, determining the candidate content item as one of the one or more target content items.

[0063] In some embodiments, at least a portion of one or more target text items is determined from the first prompt word information based on one or more reference content items: based on the one or more reference content items, content items matching the one or more reference content items are detected from the first prompt word information; and in response to detecting a content item matching a first reference content item from the one or more reference content items, the detected content item is determined as a first target content item from the one or more target content items.

[0064] In some embodiments, process 400 further includes: in response to a second target content item of the one or more target content items not matching the one or more reference content items, adding the second target content item as another reference content item.

[0065] In block 420 , the information processing platform 240 replaces one or more target content items in the first cue word information with one or more predetermined tags to generate second cue word information, wherein a predetermined tag in the one or more predetermined tags indicates attribute information of the target content item replaced by the predetermined tag.

[0066] In some embodiments, the attribute information includes the type of the replaced target content item and an identifier corresponding to the replaced target content item.

[0067] In box 430, the information processing platform 240 uses the machine learning model to execute the target task based on the second prompt word information to obtain a task execution result.

[0068] In some embodiments, process 400 also includes obtaining an execution result of the target task, the execution result including text; in response to the execution result including at least one predetermined tag among one or more predetermined tags, restoring the predetermined tag among the at least one predetermined tag to the target content item replaced by the predetermined tag to update the execution result; and presenting the updated execution result as a response to the user input.

[0069] In some embodiments, identification of at least one or more target content items and replacement of one or more target content items are performed in a trusted execution environment, the trusted execution environment is created for a target user related to the user input, and process 400 also includes: based on the operation instruction for the trusted execution environment, determining an identity corresponding to the operation instruction; and in response to the identity matching the first user, executing the operation instruction.

[0070] In some embodiments, determining the identity corresponding to the operation instruction includes: determining whether the operation instruction indicates one of a plurality of predetermined operations; and determining the identity corresponding to the operation instruction in response to the operation instruction indicating one of the plurality of predetermined operations.

[0071] In some embodiments, the first prompt word information is obtained by: determining context information related to the user input based on the user input, the context information indicating an object related to the user input; and generating the first prompt word information based on the context information and the user input.

[0072] In some embodiments, an entity recognition model used to determine one or more target content items is trained by updating model parameters of the entity recognition model based on training context information related to user training input, where the training context information indicates a training object related to the user training input.

[0073] Example devices and equipment

[0074] The embodiments of the present disclosure also provide corresponding devices for implementing the above methods or processes. Figure 5 1 shows a schematic structural block diagram of an example apparatus 500 for information processing according to some embodiments of the present disclosure. The apparatus 500 may be implemented as or included in the information processing platform 240. Each module / component in the apparatus 500 may be implemented by hardware, software, firmware, or any combination thereof.

[0075] like Figure 5 As shown, the device 500 includes an identification module 510, which is configured to identify one or more target content items corresponding to the predetermined type information from the first prompt word information indicating the target task, and the first prompt word information is generated based on the user input requesting the target task. The device 500 also includes a replacement module 520, which is configured to replace the one or more target content items in the first prompt word information with one or more predetermined tags to generate a second prompt word information, and the predetermined tag in the one or more predetermined tags indicates the attribute information of the target content item replaced by the predetermined tag. The device 500 also includes an execution module 530, which is configured to execute the target task using the machine learning model based on the second prompt word information to obtain a task execution result.

[0076] In some embodiments, the attribute information includes the type of the replaced target content item and an identifier corresponding to the replaced target content item.

[0077] In some embodiments, the identification module 510 is also configured to determine at least a portion of one or more target content items from the first prompt word information by performing pattern matching corresponding to predetermined type information on the first prompt word information, determine at least a portion of one or more target content items based on the first prompt word information using one or more entity recognition models, and determine at least a portion of the one or more target content items from the first prompt word information based on one or more reference content items, wherein the one or more reference content items include historical content items identified as corresponding to the predetermined type information.

[0078] In some embodiments, candidate content items are determined from the prompt word information by pattern matching or utilizing one or more entity recognition models, and the recognition module 510 is also configured to determine whether the candidate content item matches one or more entities corresponding to the target application, in which the user input is issued; and in response to the candidate content item matching the one or more entities, determine a given candidate content item as one of the one or more target content items.

[0079] In some embodiments, the identification module 510 is also configured to detect, based on the one or more reference content items, a content item that matches the one or more reference content items from the first prompt word information; and in response to detecting a content item that matches the first reference content item in the one or more reference content items, determine the detected content item as a first target content item in the one or more target content items.

[0080] In some embodiments, the apparatus 500 further includes a reference content item updating module configured to add the second target content item as another reference content item in response to the second target content item in the one or more target content items not matching the one or more reference content items.

[0081] In some embodiments, the device 500 also includes an execution result acquisition module, which is configured to obtain the execution result of the target task, the execution result including text; in response to the execution result including at least one predetermined tag among one or more predetermined tags, restoring the predetermined tag among the at least one predetermined tag to the target content item replaced by the predetermined tag to update the execution result; and presenting the updated execution result as a response to the user input.

[0082] In some embodiments, identification of at least one or more target content items and replacement of one or more target content items are performed in a trusted execution environment, the trusted execution environment is created for a target user related to the user input, and the device 500 also includes an operation instruction execution module, which is configured to determine an identity corresponding to the operation instruction based on the operation instruction for the trusted execution environment; and execute the operation instruction in response to the identity matching the first user.

[0083] In some embodiments, determining the identity corresponding to the operation instruction includes: determining whether the operation instruction indicates one of a plurality of predetermined operations; and determining the identity corresponding to the operation instruction in response to the operation instruction indicating one of the plurality of predetermined operations.

[0084] In some embodiments, the first prompt word information is obtained by: determining context information related to the user input based on the user input, the context information indicating an object related to the user input; and generating the first prompt word information based on the context information and the user input.

[0085] In some embodiments, an entity recognition model used to determine one or more target content items is trained by updating model parameters of the entity recognition model based on training context information related to user training input, where the training context information indicates a training object related to the user training input.

[0086] like Figure 6 As shown, the electronic device 600 is in the form of a general electronic device. The components of the electronic device 600 may include, but are not limited to, one or more processors or processing units 610, a memory 620, a storage device 630, one or more communication units 640, one or more input devices 660, and one or more output devices 660. The processing unit 610 may be an actual or virtual processor and is capable of performing various processes according to a program stored in the memory 620. In a multi-processor system, multiple processing units execute computer executable instructions in parallel to improve the parallel processing capability of the electronic device 600.

[0087] The electronic device 600 typically includes a plurality of computer storage media. Such media may be any accessible media that is accessible to the electronic device 600, including but not limited to volatile and non-volatile media, removable and non-removable media. The memory 620 may be a volatile memory (e.g., registers, caches, random access memory (RAM)), a non-volatile memory (e.g., a read-only memory (ROM), an electrically erasable programmable read-only memory (EEPROM), flash memory), or some combination thereof. The storage device 630 may be a removable or non-removable medium, and may include a machine-readable medium, such as a flash drive, a disk, or any other medium, which may be capable of being used to store information and / or data and may be accessed within the electronic device 600.

[0088] The electronic device 600 may further include additional removable / non-removable, volatile / non-volatile storage media. Figure 6 As shown in , a disk drive for reading or writing from a removable, non-volatile disk (e.g., a "floppy disk") and an optical drive for reading or writing from a removable, non-volatile optical disk may be provided. In these cases, each drive may be connected to the bus (not shown) by one or more data media interfaces. The memory 620 may include a computer program product 625 having one or more program modules that are configured to perform various methods or actions of various embodiments of the present disclosure.

[0089] The communication unit 640 implements communication with other electronic devices through a communication medium. Additionally, the functions of the components of the electronic device 600 can be implemented with a single computing cluster or multiple computing machines that can communicate through a communication connection. Therefore, the electronic device 600 can operate in a networked environment using a logical connection with one or more other servers, a network personal computer (PC), or another network node.

[0090] The input device 650 may be one or more input devices, such as a mouse, a keyboard, a tracking ball, etc. The output device 660 may be one or more output devices, such as a display, a speaker, a printer, etc. The electronic device 600 may also communicate with one or more external devices (not shown) through the communication unit 640 as needed, such as a storage device, a display device, etc., communicate with one or more devices that allow a user to interact with the electronic device 600, or communicate with any device that allows the electronic device 600 to communicate with one or more other electronic devices (e.g., a network card, a modem, etc.). Such communication may be performed via an input / output (I / O) interface (not shown).

[0091] According to an exemplary implementation of the present disclosure, a computer-readable storage medium is provided, on which computer-executable instructions are stored, wherein the computer-executable instructions are executed by a processor to implement the method described above. According to an exemplary implementation of the present disclosure, a computer program product is also provided, which is tangibly stored on a non-transitory computer-readable medium and includes computer-executable instructions, and the computer-executable instructions are executed by a processor to implement the method described above.

[0092] Various aspects of the present disclosure are described herein with reference to the flowcharts and / or block diagrams of the methods, devices, equipment, and computer program products implemented according to the present disclosure. It should be understood that each box in the flowchart and / or block diagram and the combination of each box in the flowchart and / or block diagram can be implemented by computer-readable program instructions.

[0093] These computer-readable program instructions can be provided to a processing unit of a general-purpose computer, a special-purpose computer, or other programmable data processing device, thereby producing a machine, so that when these instructions are executed by the processing unit of the computer or other programmable data processing device, a device that implements the functions / actions specified in one or more boxes in the flowchart and / or block diagram is generated. These computer-readable program instructions can also be stored in a computer-readable storage medium, and these instructions cause the computer, programmable data processing device, and / or other equipment to work in a specific manner, so that the computer-readable medium storing the instructions includes a manufactured product, which includes instructions for implementing various aspects of the functions / actions specified in one or more boxes in the flowchart and / or block diagram.

[0094] Computer-readable program instructions can be loaded onto a computer, other programmable data processing apparatus, or other device so that a series of operational steps are performed on the computer, other programmable data processing apparatus, or other device to produce a computer-implemented process, so that the instructions executed on the computer, other programmable data processing apparatus, or other device implement the functions / actions specified in one or more boxes in the flowchart and / or block diagram.

[0095] The flow chart and block diagram in the accompanying drawings show the possible architecture, function and operation of the system, method and computer program product according to multiple implementations of the present disclosure. In this regard, each square box in the flow chart or block diagram can represent a part of a module, program segment or instruction, and a part of a module, program segment or instruction includes one or more executable instructions for realizing the logical function of the specification. In some implementations as replacements, the function marked in the square box can also occur in a sequence different from that marked in the accompanying drawings. For example, two continuous square boxes can actually be executed substantially in parallel, and they can sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each square box in the block diagram and / or flow chart, and the combination of the square boxes in the block diagram and / or flow chart can be realized by a special hardware-based system that performs the function or action of the specification, or can be realized by a combination of special hardware and computer instructions.

[0096] The above descriptions of various implementations of the present disclosure are exemplary, non-exhaustive, and not limited to the disclosed implementations. Many modifications and variations will be apparent to those of ordinary skill in the art without departing from the scope and spirit of the described implementations. The selection of terms used herein is intended to best explain the principles of the implementations, practical applications, or improvements to the technology in the market, or to enable other persons of ordinary skill in the art to understand the various implementations disclosed herein.

Claims

1. An information processing method, comprising: identifying one or more target content items corresponding to predetermined type information from first prompt word information indicating a target task, wherein the first prompt word information is generated based on user input requesting the target task; Replacing the one or more target content items in the first prompt word information with one or more predetermined tags respectively to generate second prompt word information, wherein a predetermined tag in the one or more predetermined tags indicates attribute information of the target content item replaced by the predetermined tag; as well as Based on the second prompt word information, the target task is executed using a machine learning model to obtain a task execution result.

2. The method according to claim 1, wherein identifying one or more target content items corresponding to the predetermined type of information comprises at least one of the following: determining at least a portion of the one or more target content items from the first cue word information by performing pattern matching corresponding to the predetermined type information on the first cue word information, Based on the first prompt word information, using one or more entity recognition models, determining at least a portion of the one or more target content items, or At least a portion of the one or more target content items is determined from the first cue word information based on one or more reference content items, wherein the one or more reference content items include a historical content item identified as corresponding to the predetermined type of information.

3. The method according to claim 2, wherein determining candidate content items from the prompt word information by the pattern matching or by using the one or more entity recognition models, and determining at least a portion of the one or more target content items comprises: determining whether the candidate content item matches one or more entities corresponding to a target application in which the user input is issued; as well as In response to the candidate content item matching the one or more entities, the candidate content item is determined as one of the one or more target content items.

4. The method according to claim 2, wherein determining at least a portion of the one or more target text items from the first cue word information based on one or more reference content items comprises: Based on one or more reference content items, detecting content items matching the one or more reference content items from the first prompt word information; as well as In response to detecting a content item that matches a first reference content item among the one or more reference content items, the detected content item is determined as a first target text item among the one or more target text items.

5. The method according to claim 2, further comprising: In response to a second target content item among the one or more target content items not matching the one or more reference content items, the second target content item is added as another reference content item.

6. The method according to claim 1, further comprising: Obtaining an execution result of the target task, wherein the execution result includes text; In response to the execution result including at least one predetermined mark among the one or more predetermined marks, restoring the predetermined mark among the at least one predetermined mark to the target content item replaced by the predetermined mark, so as to update the execution result; as well as The updated execution result is presented in response to the user input.

7. The method of claim 1 , wherein at least the identification of the one or more target content items and the replacement of the one or more target content items are performed in a trusted execution environment, the trusted execution environment being created for a target user associated with the user input, and The method further comprises: Based on the operation instruction for the trusted execution environment, determining an identity corresponding to the operation instruction; as well as In response to the identity identifier matching the first user, executing the operation instruction.

8. The method according to claim 7, wherein determining the identity corresponding to the operation instruction comprises: determining whether the operation instruction indicates one of a plurality of predetermined operations; as well as In response to the operation instruction indicating one of the plurality of predetermined operations, an identity corresponding to the operation instruction is determined.

9. The method according to claim 1, wherein the first prompt word information is obtained by: Based on the user input, determining context information related to the user input, the context information indicating an object related to the user input; and The first prompt word information is generated based on the context information and the user input.

10. The method according to claim 2, wherein the entity recognition model used to determine the one or more target content items is trained by: Based on training context information related to the user training input, model parameters of the entity recognition model are updated, and the training context information indicates a training object related to the user training input.

11. The method according to claim 1, wherein the attribute information comprises a type of the replaced target content item and an identifier corresponding to the replaced target content item.

12. An information processing device, comprising: an identification module configured to identify one or more target content items corresponding to predetermined type information from first prompt word information indicating a target task, wherein the first prompt word information is generated based on a user input requesting the target task; a replacement module, configured to replace the one or more target content items in the first prompt word information with one or more predetermined tags, respectively, to generate second prompt word information, wherein a predetermined tag in the one or more predetermined tags indicates attribute information of the target content item replaced by the predetermined tag; as well as The execution module is configured to execute the target task based on the second prompt word information using a machine learning model to obtain a task execution result.

13. An electronic device, comprising: at least one processing unit; as well as At least one memory, the at least one memory being coupled to the at least one processing unit and storing instructions for execution by the at least one processing unit, the instructions causing the electronic device to perform the method according to any one of claims 1 to 11 when executed by the at least one processing unit.

14. A computer-readable storage medium having a computer program stored thereon, wherein the computer program can be executed by a processor to implement the method according to any one of claims 1 to 11.

Images