Data processing method and related device

Abstract

The invention discloses a data processing method and a related device, and the method comprises the steps: obtaining a resource event related to a target virtual resource in response to the display of a resource page for the target virtual resource, and generating a first to-be-determined problem related to the resource event through the resource event and a matched problem template; obtaining a second to-be-determined question which is generated through the first large model and is related to the resource event; a target question is determined from the first to-be-determined question and the second to-be-determined question, the target question is used for being displayed in a large model starting entry of the resource page, and the large model starting entry is used for displaying a session interface for performing question-answer interaction with a second large model based on the trigger operation, a target question used for asking the second large model is input in the session interface. According to the method, the user is attracted to ask questions by using the questions, and the question-answer interaction willingness of the user is improved. The question and answer interaction frequency of the user and the large model is increased, and waste of interaction resources is avoided.

Description

Technical Field

[0001] This application relates to the field of data processing, and in particular to a data processing method and related apparatus. Background Technology

[0002] Currently, artificial intelligence (AI) assistants are commonly used in various software applications. Users can ask questions to AI assistants, and the AI ​​assistants will provide answers. The interaction between users and AI assistants is essentially an interaction between users and a large model, which can generate answers based on the questions asked by users.

[0003] Taking virtual resource provision software as an example, this type of software offers users various virtual resources. For instance, in stock trading software, the virtual resource is stocks. The resource page of this type of software displays a wealth of information about the virtual resources, such as changes in their value (e.g., stock price fluctuations). Users need to extract useful information from this vast amount of data to aid their decision-making. Therefore, users often engage in conversations with AI assistants in the chat interface to obtain useful information.

[0004] When interacting with the AI ​​assistant, users can manually input questions, but this incurs operational and cognitive costs, resulting in lower-quality questions. The software can also provide pre-set questions using templates, but these templates typically have a uniform question structure, leading to monotonous and inflexible questions that reduce user engagement and ultimately decrease the number of interactions between the user and the AI ​​assistant. Summary of the Invention

[0005] To address the aforementioned technical problems, this application provides a data processing method and related apparatus that attracts users to ask questions, thereby increasing their willingness to engage in question-and-answer interactions. It also increases the frequency of user interactions with the large model, preventing the waste of interaction resources.

[0006] The embodiments of this application disclose the following technical solutions:

[0007] In a first aspect, this application provides a data processing method, the method comprising:

[0008] In response to displaying a resource page for a target virtual resource, resource events related to the target virtual resource are obtained, the resource events including at least one of the following: the resource value of the target virtual resource changes to a preset condition, or the subject behavior of the resource subject of the target virtual resource.

[0009] Using the resource event and the matching question template, a first undetermined question related to the resource event is generated; and a second undetermined question related to the resource event is obtained by generating it through the first large model.

[0010] The target question is determined from the first pending question and the second pending question. The target question is displayed in the large model launch entry on the resource page. The large model launch entry is used to display a conversation interface for question-and-answer interaction with the second large model based on a trigger operation. The target question for asking the second large model has been entered in the conversation interface.

[0011] Secondly, this application provides a data processing apparatus, the apparatus comprising:

[0012] The first acquisition unit is configured to acquire resource events related to the target virtual resource in response to displaying a resource page for the target virtual resource. The resource events include at least one of the following: the resource value of the target virtual resource changes to a preset condition, or the subject behavior of the resource subject of the target virtual resource.

[0013] The second acquisition unit is configured to generate a first pending question related to the resource event using the resource event and a matching question template; and to acquire a second pending question related to the resource event generated by the first large model.

[0014] A determining unit is configured to determine a target question from the first pending question and the second pending question. The target question is to be displayed in the large model launch entry on the resource page. The large model launch entry is to display a conversation interface for question-and-answer interaction with the second large model based on a trigger operation. The target question for asking the second large model has been entered in the conversation interface.

[0015] Thirdly, this application provides a computer device, which includes a processor and a memory:

[0016] The memory is used to store computer programs;

[0017] The processor is configured to execute the method according to the computer program.

[0018] Fourthly, this application provides a computer-readable storage medium for storing a computer program that, when executed by a computer device, implements the method described thereon.

[0019] Fifthly, this application provides a computer program product including a computer program that, when run on a computer device, causes the computer device to perform the method.

[0020] As can be seen from the above technical solution, in order to facilitate user interaction with the large model through question-and-answer sessions, it is necessary to ensure that the questions input into the large model are sufficiently relevant to the scenario and more diverse. Therefore, the server responds by displaying a resource page for the target virtual resource and obtaining resource events related to the target virtual resource. Resource events include at least one of the following: changes in the resource value of the target virtual resource reaching preset conditions, or the main behavior of the resource subject of the target virtual resource. Thus, resource events can more comprehensively and accurately reflect the current status of the target virtual resource, and the questions generated based on the resource events can be more relevant to the actual scenario, making the questions more targeted and accurate. In addition, to make the questions more diverse, this application generates questions in two ways: one is to generate a first pending question related to the resource event through the resource event and a matching question template; the other is to generate a second pending question related to the resource event through the first large model. The first generation method focuses more on the text matching between the resource event and the question template, thus ensuring that the expression of the generated first pending question is more standardized and reducing errors. The second generation method focuses more on utilizing the event understanding ability of the large model to generate the second pending question, and the expression of the second pending question is richer and more diverse. These two types of pending questions complement each other, ensuring that the questions satisfy both standardized and diverse expression requirements. Next, the server determines the target question from these two types of pending questions to be displayed in the large model's startup entry point on the resource page. Users can then trigger the large model's startup entry point to enter a dialogue interface for question-and-answer interaction with the second large model. In this dialogue interface, the target question to be asked of the second large model is already entered, thus enabling user-secondary question-and-answer interaction with the second large model regarding the target question. In summary, this application, targeting virtual resources, obtains richer questions through two question generation methods. In particular, the additional use of large model-generated questions ensures that the questions are sufficiently relevant to current hot topics and diverse, avoiding rigidity and attracting users to ask questions, thereby increasing users' willingness to engage in question-and-answer interaction. It also increases the number of user-secondary question-and-answer interactions with the large model, avoiding waste of interaction resources. Attached Figure Description

[0021] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0022] Figure 1 A schematic diagram illustrating the implementation environment of a data processing method provided in this application embodiment;

[0023] Figure 2A flowchart illustrating a data processing method provided in an embodiment of this application;

[0024] Figure 3 A schematic diagram of a first prompt word provided for an embodiment of this application;

[0025] Figure 4 A schematic diagram illustrating yet another first prompt word provided in an embodiment of this application;

[0026] Figure 5 A schematic diagram illustrating yet another first prompt word provided in an embodiment of this application;

[0027] Figure 6 A schematic diagram of a third prompt word provided for an embodiment of this application;

[0028] Figure 7 This application provides a schematic diagram of a process for generating a problem using a first large model, as illustrated in an embodiment of the present application.

[0029] Figure 8 This application provides a schematic diagram illustrating the problem generated using a first large model, as part of an embodiment of the present application.

[0030] Figure 9 A schematic diagram illustrating a first unresolved problem provided in an embodiment of this application;

[0031] Figure 10 A flowchart illustrating the generation of a first undetermined problem is provided for an embodiment of this application.

[0032] Figure 11 This application provides a schematic diagram of a processing flow for a target problem.

[0033] Figure 12 This is a schematic diagram illustrating the maintenance of a problem template as provided in an embodiment of this application.

[0034] Figure 13 This application provides a schematic diagram of a process for generating questions using a question template, as illustrated in an embodiment of the present application.

[0035] Figure 14 A schematic diagram illustrating a multi-factor combination of a problem generation method provided in an embodiment of this application;

[0036] Figure 15 This is a schematic diagram of the overall flow of a data processing method provided in an embodiment of this application;

[0037] Figure 16 A schematic diagram of a data processing apparatus provided in an embodiment of this application;

[0038] Figure 17 A structural diagram of a terminal device provided in an embodiment of this application;

[0039] Figure 18 This is a structural diagram of a server provided in an embodiment of this application. Detailed Implementation

[0040] The embodiments of this application will now be described with reference to the accompanying drawings.

[0041] As described in the background section, taking stock-related software as an example of virtual resource provision software, the resource pages of such software display a large amount of information about virtual resources (stocks), making it difficult for ordinary users to quickly extract important decision-making signals. For instance, the individual stock news page may have a list of individual stock news items displaying an information flow, allowing users to browse stock news highlights by clicking, but it lacks an efficient summary format. Another example is the candlestick chart area, used to present stock price fluctuations, but users cannot quickly obtain the reasons for these fluctuations from the price increases or decreases or candlestick chart trends. Yet another example is the financial page, which contains data results and announcements related to stock financial reports, but users cannot easily access key performance information with a single click.

[0042] Therefore, while users can utilize AI assistants within software to obtain important information, current AI assistants generally rely on users to ask questions manually, incurring operational and cognitive costs, and resulting in relatively poor question quality. Some software can also provide pre-set questions via templates, but these templates typically have a uniform question structure, leading to monotonous and rigid questions that fail to provide users with a consistently engaging experience. Furthermore, these pre-set questions are usually confined to the conversation page with the AI, requiring users to enter the conversation page to view them. This reduces the usage rate of pre-set questions, resulting in fewer interactions between users and the AI ​​assistant, especially in the financial and securities fields where users need to ask questions that are more context-specific and varied.

[0043] Based on this, this application, targeting virtual resources, obtains a richer set of questions through two question generation methods. In particular, it adds the use of a large model to generate questions, ensuring that the questions are relevant to current hot topics and diverse, avoiding rigidity and attracting users to ask questions, thus increasing users' willingness to interact with the question-and-answer process. It also increases the number of user interactions with the large model, avoiding waste of interaction resources.

[0044] The data processing method provided in this application can be implemented using a computer device, which can be a terminal device or a server. The server can be an independent physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server providing cloud computing services. Terminal devices include, but are not limited to, mobile phones, computers, intelligent voice interaction devices, smart home appliances, vehicle terminals, aircraft, and extended reality (XR) devices. The terminal device and the server can be directly or indirectly connected via wired or wireless communication, and this application does not impose any limitations on this connection.

[0045] It is understood that in the specific implementation of this application, user information such as target virtual resources, resource events, various triggering operations, target accounts and other related data are involved. When the above embodiments of this application are applied to specific products or technologies, user permission or consent is required, and the collection, use and processing of related data must comply with the relevant laws, regulations and standards of the relevant countries and regions.

[0046] The data processing method provided in this application can be applied in at least one of the following scenarios. Of course, these scenarios are just examples, and the method can also be applied to other scenarios, which are not limited here.

[0047] Scenario 1: Asking questions on financial software

[0048] In this scenario, the virtual resource-providing software is stock-related software. Stock-related software is used by users to obtain stock-related information, providing data such as stock price fluctuations and the company's actions. Because this information is often technical and difficult to understand, the software can provide users with pre-set stock-related questions, such as "Why did the stock price of xxx stock plummet?" Users can directly ask these questions to the software's AI assistant to obtain the important information they want from the AI ​​assistant's answers, helping them quickly extract crucial information from technical details.

[0049] Scenario 2: Summarizing in news and information software

[0050] In this scenario, the virtual resource provider software is a news app, which offers users the latest news and information. Due to the large volume and rapid updates of news, such as continuous follow-up reports on the same news story, the news app can provide pre-set, news-related questions. These questions could include phrases like "What are the details of the event related to news item xx?" or "What caused the fire in city xx?" Users can directly ask these pre-set questions to the AI ​​assistant, helping them obtain a concise summary of events from a large volume of news.

[0051] Of course, this method can also be applied to other scenarios. For example, in shopping apps, users can not only select and order products, but also ask questions about them. The app's page can display pre-set product-related questions, such as "What is the price trend of product xx?" or "What are the main selling points of product xx?" Similarly, in gaming apps, users can be provided with questions related to game items and scenarios, such as "How do I obtain product xx?" or "What is the difficulty level of completing game scenario xx?" Furthermore, in short-video social media apps, users can be provided with questions related to content mentioned in short videos, such as "What is the story behind xx mentioned in the video?"

[0052] refer to Figure 1 The diagram shown illustrates the implementation environment of a data processing method provided in this embodiment. This data processing method can be applied to... Figure 1 The implementation environment shown includes server 100, terminal device 200, and database 300.

[0053] A client can be deployed on the terminal device 200. The client can run on the terminal device through a browser, or it can run on the terminal device as a standalone application (APP) or a mini-program.

[0054] Server 100 and terminal device 200 can communicate via a communication network. This communication network uses standard communication technologies and / or protocols, typically the Internet, but can also be any network, including but not limited to Bluetooth, local area network (LAN), metropolitan area network (MAN), wide area network (WAN), mobile, private network, or any combination of virtual private network. In some embodiments, customized or dedicated data communication technologies may be used to replace or supplement the aforementioned data communication technologies.

[0055] In light of the above implementation environment, the data processing method will be explained in detail. Taking stock market software as an example, when a user triggers the icon of the stock market software installed on terminal device 200, terminal device 200 can display a resource page for the target virtual resource (step A1), such as... Figure 1 The target virtual resource 102 is stock X, and the resource page 101 is the page corresponding to stock X. Terminal device 200 sends the displayed resource page to server 100 so that server 100 can provide the target information to terminal device 200. In response to displaying the resource page, server 100 retrieves resource events related to the target virtual resource from database 300 (step A2).

[0056] Resource events include at least one of the following: a change in the resource value of a target virtual resource reaching a preset condition, or a behavior of the entity that owns the resource. For example, a change in resource value reaching a preset condition might be a stock price of X hitting its daily limit, while the entity's behavior might be the release of the latest financial results by the company owning stock X. Resource events can more comprehensively and accurately reflect the current status of the target virtual resource, and subsequent questions generated based on these events can be more relevant to real-world scenarios, making the questions more targeted and accurate. Figure 1 The document also shows some resource events, which may or may not be related to stock X.

[0057] To enrich the variety of questions, this application generates questions in two ways: step A3, where server 100 generates a first pending question related to the resource event using the resource event and a matching question template, and obtains a second pending question related to the resource event generated by the first large model. For example, if the question template is "[Name] stock [Date] rose, what are the reasons?", then the first pending question would be "What are the reasons for the rise of stock X today?". Another example is "How did the funds flow after stock X hit its daily limit?" The first generation method focuses more on the textual matching between the resource event and the question template, thus ensuring a more standardized expression of the generated first pending question and reducing errors. The second generation method focuses more on utilizing the event understanding capabilities of the large model to generate the second pending question, resulting in a richer and more diverse expression of the second pending question. These two types of pending questions complement each other, enabling the questions to satisfy both standardized and diverse expression requirements.

[0058] Step A4: Server 100 determines the target problem from the first and second pending problems. Step A5: Server 100 sends the target problem to terminal device 200, and terminal device 200 displays the target problem in the large model startup entry on the resource page. (See reference...) Figure 1 The entry point for launching the large model is located in the middle of the page, and the target question 103 is displayed in the middle of the page.

[0059] When a user is interested in the target question and wants to ask a question, an action can be triggered at the large model's startup entry point, such as... Figure 1 When a user clicks the entry point, the terminal device 200 displays a conversation interface 104 for question-and-answer interaction with the second-largest model. The conversation interface 104 displays the target question 103 that has been entered to ask the second-largest model, and the second-largest model can answer it.

[0060] In summary, this application, targeting virtual resources, yielded a richer set of questions through two question generation methods. In particular, the additional use of a large model to generate questions ensured that the questions were relevant to current hot topics and varied, avoiding rigidity and attracting users to ask questions, thus increasing their willingness to engage in question-and-answer interaction. It also increased the frequency of user interaction with the large model, preventing the waste of interaction resources.

[0061] refer to Figure 2 The diagram shown is a flowchart of a data processing method provided in an embodiment of this application. The method can be executed by a computer device. In this embodiment, the computer device is a server as an example for illustration. The method includes steps 201-203.

[0062] Step 201: In response to displaying the resource page for the target virtual resource, obtain resource events related to the target virtual resource.

[0063] Virtual resources refer to virtualized resources existing on the internet. They are non-material resources that can meet specific user needs. Virtual resources can be virtual items, virtual assets, etc. Examples of virtual resources include stocks, news, and game items. A target virtual resource is a specific virtual resource, specifically the virtual resource displayed on a terminal device. For example, when a user opens the resource page for stock X in stock trading software, the target virtual resource is stock X.

[0064] A resource page is a page used to display information about virtual resources. It provides a detailed overview of the virtual resource and can be examples such as a details page or a news page. The display of a resource page for a target virtual resource can be triggered by user actions, such as opening software or searching for stock X, which will cause the terminal device to display the resource page for the target virtual resource.

[0065] When a terminal device displays a resource page for a target virtual resource, the server can acquire resource events related to the target resource. A resource event refers to an event that occurs surrounding a virtual resource; it is a specific behavior or resource change related to the virtual resource. A resource event can include at least one of the following: the resource value of the target virtual resource changes to meet a preset condition, or the behavior of the resource entity of the target virtual resource. A resource event is an event that affects a virtual resource. For example, resource events can include events such as stock price limits (limit up / down), new highs / lows, the lifting of price limits, company dividends, shareholder increases / decreases, earnings releases, and stock issuances.

[0066] Virtual resources have resource values, which reflect their worth. Examples include stock prices, news popularity, and the virtual value of game items. Preset conditions refer to the conditions that a resource value must meet beforehand. For example, preset conditions might include a stock price fluctuation reaching a predetermined range, a stock price reaching a new historical high / low, or a stock price hitting its daily price limit. As an example, a resource event could be a stock price fluctuation exceeding 3%.

[0067] Virtual resources have corresponding resource entities, which refer to the parties to which the virtual resources belong, such as the issuer of stocks. The actions of a resource entity refer to relevant behaviors that affect the virtual resources, such as causing a decline in stock prices. Examples of such actions include a company issuing an annual report, shareholders increasing or decreasing their holdings, distributing dividends, and issuing new shares.

[0068] In short, the server can obtain resource events related to the target virtual resource, so that the resource events can more comprehensively and accurately reflect the current status of the target virtual resource. Subsequent questions generated based on the resource events can be more in line with the actual scenario, and the questions are more targeted and accurate.

[0069] Step 202: Generate a first pending question related to the resource event using the resource event and the matching question template; and obtain a second pending question related to the resource event generated by the first major model.

[0070] To make the generated questions more diverse, this application generates questions in two ways: one is to generate a first undetermined question related to the resource event using the resource event and a question template that matches the resource event; the other is to generate a second undetermined question related to the resource event using the first major model.

[0071] The question template is a pre-configured template used to generate questions. The question template standardizes the format of the question; for example, a question template might be "[Name] What is the reason for the rise in stock [Date]?" Since resource events are diverse—for example, events related to changes in resource values, events related to subject behavior, and other events—it is necessary to match question templates from a large number of question templates based on resource events. Then, the question templates can be populated (or rendered) and improved based on the resource events to generate the first pending question related to the resource event. Pending questions refer to questions generated by the server for display on terminal devices; the first pending question refers to the question generated through template matching.

[0072] The first large model refers to the large model used to generate the problem, such as various existing large models. The first large model has powerful understanding capabilities, so resource events can be pre-input into it, allowing the first large model to generate related second undetermined problems based on these resource events. The second undetermined problems are the problems generated through the large model.

[0073] In short, the first generation method focuses more on the textual matching between resource events and question templates, thus ensuring that the generated first pending question is more standardized and reduces errors. The second generation method focuses more on utilizing the event understanding capabilities of the large model to generate the second pending question. The expression of the second pending question is richer and more diverse; that is, the question phrasing and questioning style of the second pending question output by the first large model will be more varied. These two types of pending questions complement each other, enabling the questions to satisfy both standardized expression and expressive diversity.

[0074] Furthermore, since the server can simultaneously generate the first pending question and obtain the second pending question, if an anomaly occurs in obtaining the second pending question, such as an error in the question generation process of the first model, or the generated second pending question has become invalid, the server can use the first pending question obtained by template matching to determine the target question that needs to be displayed based on the first pending question. This avoids the target question not being displayed for a long time, causing users to be unable to use the service, thereby improving the user's questioning experience.

[0075] Step 203: Determine the target problem from the first undetermined problem and the second undetermined problem.

[0076] Since both the first and second pending problems are related to resource events of the target virtual resource, the server can determine the target problem from these two types of pending problems. The target problem refers to the problem provided by the user for the user to use, and the number of target problems is not limited.

[0077] Specifically, the target question is displayed in the large model launch entry on the resource page. The large model launch entry is the entry point used to launch the large model for question-and-answer interaction. Based on a trigger operation, the large model launch entry displays a conversation interface for question-and-answer interaction with the second large model, where the target question to be asked to the second large model has already been entered.

[0078] In other words, users can trigger an operation to activate the main model. The terminal device, in response, displays a conversation interface. This conversation interface is the interface for question-and-answer interaction with the second main model, which is the main model used to answer questions. The second main model can be the same as the first main model or a different model. Furthermore, in the conversation interface, users do not need to repeatedly enter the target question; the target question is directly displayed, and the second main model can directly answer it.

[0079] As an example, the resource page provides a "Ask AI" toolbar (the entry point for launching the main model). The toolbar also displays the target question. Clicking on the target question will bring up a half-screen pop-up (a conversation page), allowing the second main model to automatically answer the question. Users can also close the conversation interface themselves without affecting their continued browsing of the resource page.

[0080] In short, to facilitate timely Q&A interaction between users and the large model, a large model launch entry can be configured on the displayed resource page. This eliminates the need for users to exit the resource page and re-enter the Q&A session interface to interact with the large model, thus meeting users' needs to ask questions at any time.

[0081] In summary, this application, targeting virtual resources, generates a richer set of questions through two question generation methods. In particular, the addition of a large-scale model ensures that the questions are relevant to current trends and diverse, avoiding rigidity and providing users with a variety of questions. Users no longer need to manually input questions, and the displayed questions are no longer monotonous. When browsing resource pages, users only need to click the large-scale model activation entry to invoke a second large-scale model to answer the target question. This establishes a multi-stage interconnected mechanism from "resource event triggering - large-scale model question generation - one-stop scenario configuration," solving the problems of users not knowing how to ask questions when faced with complex and diverse resource events, and the aesthetic fatigue caused by the high repetitiveness of template-preset questions. This application encourages users to ask questions, increasing their willingness to engage in Q&A interaction. It also increases the number of Q&A interactions between users and the large-scale model, avoiding waste of interaction resources. This method can be applied to various scenarios, such as financial scenarios, helping users to ask precise questions and obtain important information.

[0082] Taking the stock market and financial scenarios as an example, from the user's perspective, this application can proactively summarize and extract recent hot topics for individual stocks, providing users with content interpretation based on market data. It can also help users summarize and simplify large amounts of information, quickly finding results from complex data, saving time spent searching for information. By generating questions through AI connectivity, the application allows for more diverse question angles, providing a comprehensive understanding of a stock. From a business perspective, this application simultaneously solves the problems of multi-scenario management and reuse, improving operational efficiency.

[0083] The following section will elaborate on this application from three aspects: the generation of pending questions, the generation of questions based on event triggers, and the display of target questions.

[0084] I. Generation of Undetermined Problems

[0085] 1.1 Generation of the Second Undetermined Problem

[0086] To improve the grammatical standardization of the second undetermined question generated by the first major model, in one possible implementation, the second undetermined question is determined as follows: a first prompt word is generated based on the resource event and the grammatical requirements corresponding to the resource event; based on the first prompt word, a second undetermined question related to the resource event is generated through the first major model, wherein the question grammar and grammatical requirements of the second undetermined question are related.

[0087] Because there are many types of resource events, corresponding sentence structure requirements can be determined for different resource events. Sentence structure requirements refer to the phrasing requirements for questions generated based on resource events, that is, the requirements for how the questions are expressed. For example, the sentence structure requirement is that the generated questions should be formatted similarly to questions such as "Why has the stock price of xxx continued to decline?"

[0088] The server generates a first prompt word based on the resource event and the corresponding sentence structure requirements. A prompt word is a hint used to guide the main model's thinking process; the first prompt word specifically guides the main model on how to generate relevant second undetermined questions based on the resource event. In practical applications, the first prompt word may include the resource event and the sentence structure requirements, which are illustrated using positive sample examples.

[0089] The server can input the first prompt word into the first main model. Guided by the first prompt word, the first main model can generate relevant second pending questions based on resource events. Since the first prompt word reflects the sentence structure requirements for the generated questions, the sentence structure of the generated second pending questions is related to the sentence structure requirements, and the sentence structure of the second pending questions meets the requirements. As an example, the sentence structure of the second pending questions is similar to that of the questions in the positive sample examples.

[0090] In summary, by inputting the first prompt word into the first model, the first prompt word can standardize the question sentence of the generated second pending question. This not only enriches the question sentence of the second pending question but also ensures that the question sentence is more standardized and regulated, thereby improving the accuracy of the second pending question and making the second pending question sufficiently relevant to the resource event.

[0091] The following example uses three different resource events to illustrate the first prompt word.

[0092] Example 1

[0093] The resource event related to the target virtual resource is that if the price fluctuation of stock X exceeds a preset range within a certain period, the corresponding first prompt word can be found here. Figure 3 As shown, Figure 3 This is a schematic diagram of a first prompt word provided in an embodiment of this application. The first prompt word is specifically:

[0094] As an ordinary stock investor, please follow the procedure below and ask your one most pressing question.

[0095] Execution process:

[0096] 1. From {{search_result}}<text_content> Obtain recent events that have caused abnormal stock price fluctuations in stock X, ensuring that the issues represent the most pressing questions that are of general concern to investors.

[0097] 2. Combining {{search_result}}<text_content> This refers to an internal professional financial database. The price fluctuation data used in your answer must be selected from this database. Values ​​must be precisely rounded to integers before being applied to the question. Data within {{content}} is prohibited.

[0098] Output requirements:

[0099] 1. Ensure sentences are fluent and the word count is within 18 characters. Do not exceed this limit, and do not output any other content.

[0100] 2. Please do not directly ask questions about the stock's fundamentals or funding situation.

[0101] Example of outputting a positive sample:

[0102] Why did stock xxx, whose net profit increased by 351%, fall by 21% on the 20th?

[0103] Can xxx's AI-powered stock trading service support a 36% increase in 60 days?

[0104] The stock price fluctuation event is as follows:

[0105] {{content}}

[0106] The relevant information is as follows:

[0107] {{search result}}

[0108] In the first prompt, stock X is the target virtual resource, the positive sample example is the sentence structure requirement, the stock fluctuation event is the resource event related to the target virtual resource, and the relevant data is the event automatically searched by the first model to help generate the second undetermined question. In subsequent Examples 2 and 3, the sentence structure requirements and resource events can be found here; they will not be repeated here.

[0109] Example 2

[0110] The resource event related to the target virtual resource is that the stock price of stock X reaches a new high / low within a certain period of time, for example, a new high / low in the past 20 days, the past 60 days, or an all-time high / low. The corresponding first prompt keyword can be found in the reference section. Figure 4 As shown, Figure 4 This is a schematic diagram illustrating yet another type of first prompt word provided in an embodiment of this application. The first prompt word is specifically:

[0111] As an ordinary stock investor, please raise your one most pressing question, considering the unusual stock price movements and related events.

[0112] Output requirements:

[0113] 1. The question must represent the most core concerns that general investors are generally concerned about recently.

[0114] 2. Ensure sentences are fluent and the word count is within 18 characters. Do not exceed this limit, and do not output any other content.

[0115] 3. Questions must be related to unusual stock price fluctuations. Please do not directly inquire about stock fundamentals or funding dynamics.

[0116] 4. The question must include the time frame of the price fluctuation during the stock's unusual activity, such as a new 60-day high or a new historical low.

[0117] The stock price fluctuation event is as follows:

[0118] {{content}}

[0119] The relevant information is as follows:

[0120] {{search result}}

[0121] Example of outputting a positive sample:

[0122] Why did Company xxx's order volume hit a 60-day low?

[0123] Can Company xxx's AI business support breaking its all-time daily record high?

[0124] Is the record high in xxx company's stock price due to a breakthrough in its core business?

[0125] Example of outputting negative samples:

[0126] Can the stock price of Company XXX maintain its new 60-day high?

[0127] Example 3

[0128] The resource event related to the target virtual resource is when the daily price fluctuation of stock X exceeds a preset threshold, such as a fluctuation exceeding 3% or 5%. The corresponding first prompt word can be found here. Figure 5 As shown, Figure 5 This is a schematic diagram illustrating another type of first prompt word provided in an embodiment of this application. The first prompt word is specifically:

[0129] As an ordinary stock investor, please raise one question that you are most concerned about, based on the stock price fluctuations and related events.

[0130] Output requirements:

[0131] 1. The question must represent the most core concerns that general investors are currently most concerned about.

[0132] 2. Ensure the sentences are fluent, include the stock name, and keep the number of characters to 15. Do not exceed this limit, and do not output any other content.

[0133] 3. Questions must be related to unusual stock price fluctuations. Please do not directly ask about stock fundamentals, funding dynamics, or other related issues.

[0134] 4. Your question must be based on the most relevant and up-to-date information in {{search_result}} that is relevant to the time of the event within {{content}}. Please prioritize relevant information from the day the event occurred or 1-2 days prior, and ignore older information that predates the event in {{content}}.

[0135] The stock price fluctuation event is as follows:

[0136] {{content}}

[0137] The relevant information is as follows:

[0138] {{search result}}

[0139] Example of outputting a positive sample:

[0140] Why is the stock price of Company XXX fluctuating despite positive order volume?

[0141] What happens to the capital flow after Company xxx's stock hits its daily limit?

[0142] What are the reasons for the continuous decline in the stock price of Company xxx?

[0143] To ensure the accuracy of the syntax requirements corresponding to resource events, in one possible implementation, the method may further include determining the event type of the resource event; and determining the syntax requirements corresponding to the resource event based on the event type.

[0144] The event type refers to the category to which the resource event belongs. For example, an event type could be a resource value change reaching a range of fluctuations, a resource value reaching a new high or low, or a resource value change reaching a fluctuation threshold—corresponding to the three examples above. Of course, event types can also include other types, such as those related to the entity's actions. For instance, an event type could be a company releasing financial reports or announcements, or a company issuing new shares. Further specific limitations on event types are not provided here.

[0145] Because different event types lead to different angles from which questions are asked about resource events, the server can determine the corresponding sentence structure requirements based on the event type. For example, when the event type is a resource value change reaching a range of fluctuations, the sentence structure requirement is to demonstrate a connection to the range of fluctuations, such as "Why did stock xxx, whose net profit increased by 351%, fall by 21% on the 20th?" Another example is when the event type is a resource value change reaching a fluctuation threshold, the sentence structure requirement is to demonstrate a connection to historical highs / lows, such as "Was the historical high of company xxx's stock related to a breakthrough in its core business?"

[0146] In summary, by determining the corresponding sentence structure requirements based on the event type of the resource event, the generated question structure of the second pending question can better match the event type of the resource event. The second pending question is more targeted to the resource event, ensuring that the second pending question is more relevant to the resource event, and the question focuses on a more precise angle.

[0147] To improve the efficiency of obtaining the second undetermined problem generated by the first large model, in one possible implementation, the second undetermined problem is determined as follows: in response to determining that a resource event related to the target virtual resource has occurred, a second undetermined problem related to the resource event is generated by the first large model and stored in the first database. Then, in step 202, obtaining the second undetermined problem related to the resource event generated by the first large model may include obtaining the second undetermined problem from the first database.

[0148] In other words, when the server determines that a resource event related to the target virtual resource has occurred, it can input the resource event into the first major model, which then generates the second undetermined problem. Specifically, determining that a resource event has occurred can mean that the server obtains the resource event from the internet.

[0149] The second undetermined question can be stored in the first database, so that when a user opens a resource page for the target virtual resource, the second undetermined question already generated by the first model can be directly retrieved from the first database.

[0150] In short, once the server determines that a resource event related to the target virtual resource has occurred, it can generate a second pending question using the first major model. Then, when the user opens the resource page of the target virtual resource, the server responds to the display of the resource page by obtaining the resource event, and generates a first pending question based on the resource event. At the same time as generating the first pending question, the already generated second pending question is also obtained.

[0151] In summary, once a resource event related to the target virtual resource is determined to have occurred, a second pending problem is generated in advance. This allows the second pending problem to be retrieved directly from the first database when it is subsequently used, without waiting for the first large model to generate it on the spot. This improves the efficiency of obtaining the second pending problem, reduces the impact of the time difference in generating the second pending problem by the first large model on the problem retrieval, solves the limitation of the time difference in generating problems by the large model, and enables rapid retrieval of the second pending problem.

[0152] To ensure the full and effective utilization of the first major model to generate the problem, in one possible implementation, resource events include first-level events and second-level events. The importance of second-level events is higher than that of first-level events. The importance of events is positively correlated with the rate of change of the resource value of the target virtual resource, or the importance of events is positively correlated with the difficulty of the subject's behavior. The second undetermined problem is determined as follows: in response to resource events related to the target virtual resource that have second-level events, the second undetermined problem related to the second-level events is generated through the first major model.

[0153] Since there are many resource events related to the target virtual resource, different levels can be assigned to distinguish them. Resource events belonging to different levels have different levels of importance; the higher the level, the more important the resource event. That is, resource events can include first-level events and second-level events, with second-level events having higher importance than first-level events.

[0154] Specifically, when a resource event involves a change in the value of a target virtual resource that reaches a threshold condition, the greater the rate of change in the resource value, the greater the importance of the resource event. That is, the rate of change in the resource value of the target virtual resource in a second-level event is higher than the rate of change in the resource value of the target virtual resource in a first-level event. Here, the rate of change in resource value refers to the degree of change in the value of the virtual resource, which can be reflected through range fluctuations, historical highs / lows, and whether a fluctuation threshold has been reached.

[0155] When a resource event is a behavior of a resource subject targeting a virtual resource, a higher difficulty level indicates a greater event importance. In other words, the difficulty of a behavior occurring in a second-level event is higher than the difficulty of a behavior occurring in a first-level event. Behavior difficulty refers to the difficulty for the resource subject to perform the behavior; the greater the difficulty, the less likely the behavior is to occur. For example, the difficulty of a company publishing its annual report is higher than the difficulty of a company publishing a promotional message.

[0156] Furthermore, more important second-level events can be input into the first main model to obtain second undetermined problems related to the second-level events. For first-level events, since their importance is relatively low, they may not be input into the first main model to generate problems; however, they may also be input into the first main model, which is not limited here.

[0157] In summary, by generating second-level undetermined problems based on second-level events of higher importance, problems can be generated using specific resource events. This enables precise utilization of the limited resources of the first-level model. First-level events of lower importance do not need to be input into the first-level model and do not occupy model computing resources. This raises the usage threshold of the first-level model and enables targeted utilization of the first-level model.

[0158] To further enhance the richness of the first major model generation problem, in one possible implementation, the method may further include, in response to the timed task satisfying the triggering condition, if there are no resource events related to the target virtual resource, generating a third undetermined problem related to the target virtual resource through the first major model based on the trend prediction target of the target virtual resource, then step 203, determining the target problem from the first undetermined problem and the second undetermined problem, may include determining the target problem from the first undetermined problem, the second undetermined problem, and the third undetermined problem.

[0159] Scheduled tasks are tasks that are triggered at set times, meaning they are triggered according to a certain time cycle, such as triggering once a day. Scheduled tasks are used to generate questions. A scheduled task is considered to have reached its trigger time when its triggering conditions are met.

[0160] If the server responds to a scheduled task that meets the triggering conditions, but there are no resource events related to the target virtual resource at this time—meaning the change in the target virtual resource's value has not met the preset conditions, or the resource subject of the target virtual resource has not performed any action—then it is impossible to generate the first and second pending problems using resource events. In this case, the target can be predicted based on its trend, and a third pending problem can be generated using the first major model. This third pending problem is also related to the target virtual resource.

[0161] The trend prediction objective refers to the objective of predicting the future trend of a target virtual resource. This can be predicting changes in the resource value, predicting the behavior of the resource's subject, or other aspects of the target virtual resource, which will not be elaborated here. In other words, the timed task is more specifically a task that predicts the future trend of a target virtual resource, i.e., generating a forward-looking question. Specifically, the generated third undetermined question is a question asking about the future trend of the target virtual resource. For example, the third undetermined question could be, "Is it likely that stock xxx will continue its upward trend in the future?"

[0162] In summary, by triggering scheduled tasks, the first model can also generate prospective questions, namely the third undetermined questions, without needing to input resource events into the first model. This makes the types of questions generated by the model more diverse, covering not only current resource events but also predictions of future trends. This further improves the diversity and variability of the questions generated by the model, meeting the user's needs.

[0163] In practical applications, when the first model generates the third undetermined question, a third prompt word can be input into the first model. The third prompt word is used to guide the first model on how to generate the third undetermined question. The third prompt word can reflect the trend prediction target of the target virtual resource.

[0164] As an example, see reference Figure 6 The diagram shown is a schematic representation of a third prompt word provided in an embodiment of this application. The third prompt word is specifically:

[0165] You are an ordinary stock investor who has been navigating the Chinese A-share market for decades. Your investment experience tells you that in the short-term speculation of A-shares, huge "expectation gaps" and "imagination space" are often stronger catalysts for stock prices than certainty. You are good at chasing hot topics and understand that "stories" and "expectations" have a greater influence than "performance" and "facts" in the short term.

[0166] You must strictly follow all of the following steps internally:

[0167] I. Thinking and Judgment Process:

[0168] 1. Remove events from the search results that are more than two days later than the latest date. For events that meet the timeliness criteria, conduct an impact assessment and select the most promising candidate catalyst. If there are no trending events within the specified time frame, the event period can be extended to within 7 days. Your judgment logic should follow the following:

[0169] (1) If an event is currently a hot topic and causes stock price fluctuations, then pay attention to the event;

[0170] (2) If the negative / positive factors have already been fully priced in, meaning that short-term events have already been reflected in stock price fluctuations, then pay attention to recent major events;

[0171] (3) If a short-term event causes a sustained rise or fall in stock price, then pay attention to the subsequent impact of the event;

[0172] (4) If there are no major events in the near future and short-term events do not cause stock price fluctuations, then pay attention to events that may lead to future growth.

[0173] (5) The “short term” should not exceed 2 days. Events that have been in the past week or more are prohibited, as such events are likely to have been absorbed by the stock price.

[0174] 2. Among multiple real catalysts, prioritize the event with the most "storytelling," "imagination potential," and "expectation gap." Events with ambiguous impact should be given lower priority. Based on the final selected catalyst, construct a question. Before outputting the final answer, you must conduct a self-check. The checklist must meet the following three requirements:

[0175] (1) All your analysis and final questions must be based on verifiable public information.

[0176] (2) It is strictly forbidden to speculate, assume or fabricate any non-existent events.

[0177] (3) The catalyst selected must be closely related to the core business or theme of the target company.

[0178] II. Output Requirements:

[0179] 1. Construct your final choice into a question, with a limit of 18 characters. Do not exceed this limit.

[0180] 2. The output question does not contain any prefixes, explanations, or extraneous text.

[0181] 3. The questions are aimed at ordinary stock traders and need to be related to stock prices. Please ensure that the questions are clear, accurate, and easy to understand.

[0182] 4. The question must include the stock name, and abbreviations can be used, such as xxx stock.

[0183] Here, {{search_result}} represents relevant data, which are events automatically searched by the first model to help generate the third undetermined question, rather than resource events related to the target virtual resource.

[0184] Next, combined Figure 7 This paper describes the overall process of generating the second and third undetermined problems from the first major model. Figure 7This is a schematic diagram illustrating a process for generating a problem using a first large model, as provided in an embodiment of this application.

[0185] In response to the display of a resource page for a target virtual resource, the server can acquire resource events related to the target virtual resource. These events include changes in the target virtual resource's value reaching preset conditions, and the actions of the resource entity being queried. After acquiring resource events, questions can be generated using the first major model. Additionally, the triggering of scheduled tasks can also utilize the first major model to generate questions. Furthermore, when a resource value change reaches a preset condition, the server can record the change. After querying the resource entity's actions, invalid actions can be filtered out. Figure 7 These two steps are not shown in the diagram.

[0186] These three types of data are input into the first main model. The first main model parses the input data and rewrites the search terms based on the parsing results. The rewritten search terms are used to enable the first main model to automatically search for relevant information (e.g., {{search_result}}). The relevant information includes the company's financial reports, real-time stock quotes, and recent company news. The first main model obtains integrated suggestion words based on the parsing results and relevant resources. When the input data is a resource event related to the target virtual resource, the integrated suggestion word is the first suggestion word. When the input data is a scheduled task, the integrated suggestion word is the third suggestion word.

[0187] The first major model performs large-scale inference based on integrated prompts, yielding inference results. These results are then validated, and those that pass validation are designated as generated questions and stored in the database. Specifically, when the integrated prompt is the first prompt, the generated question becomes the second determined question; when the integrated prompt is the third prompt, the generated question becomes the third undetermined question.

[0188] refer to Figure 8 The diagram shown is a display illustration of a problem generated using the first major model, provided in an embodiment of this application. The unresolved problem generated by the first major model is displayed on the resource page 102 of the target virtual resource 101 (i.e., stock X). Figure 8 In (a), the pending question shown is the second pending question 201 generated based on the resource event. The resource event here is the abnormal stock price movement, and the second pending question 201 is specifically "What is the reason for the sharp drop in stock X in the past 5 days?"

[0189] exist Figure 8In (b), the pending problem shown is the third pending problem 202 generated based on a scheduled task. Specifically, the first major model can generate a forward-looking third pending problem based on the current target (i.e., the target virtual resource) and combined with network information (i.e., relevant data). The third pending problem may be, for example, a forward-looking prediction of future stock trends, or a related stock selection based on popular concepts / star industry chains. Figure 8 In (b), the third undetermined question 202 is, for example, “In the xx chain, which stocks have greater potential than stock X?”

[0190] II. Generation of Event-Triggered Problems

[0191] Event-triggered issues refer to issues generated based on resource events and issue templates, without requiring the use of a large model for generation. These generated issues are the aforementioned first pending issues. The specific generation process and issue types of the first pending issues will be explained below.

[0192] To ensure the accuracy of the first pending question, in one possible implementation, step 202 involves generating a first pending question related to a resource event using the resource event and a matching question template. This can include: when the resource event includes a first type of event where the resource value change of the target virtual resource meets a preset condition, matching a first template related to the first type of event from the question template, and generating the first pending question using the first type of event and the first template; when the resource event includes a second type of event where the main body of the target virtual resource exhibits a behavior, matching a second template related to the second type of event from the question template, and generating the first pending question using the second type of event and the second template.

[0193] Since resource events include at least two types—one being the change in the resource value of the target virtual resource reaching a preset condition, and the other being the subject behavior of the resource subject of the target virtual resource—for ease of subsequent description, resource events where the change in the resource value of the target virtual resource reaches a preset condition are denoted as the first type of event, and resource events where the subject behavior of the resource subject of the target virtual resource is denoted as the second type of event, so as to match appropriate question templates for different types of resource events in the future.

[0194] In other words, when generating the first pending issue, it is necessary to match the corresponding issue template based on the type of resource event. When the resource event includes the first type of event, the first template related to the first type of event is matched from a large number of issue templates, and then the first pending issue corresponding to the first type of event is generated based on the first type of event and the first template.

[0195] When a resource event includes a second type of event, a second template related to the second type of event is matched from a large number of question templates. Based on the second type of event and the second template, a first undetermined question corresponding to the second type of event is generated. Here, the first template is the template that matches the first type of event, and the second template is the template that matches the second type of event.

[0196] In summary, by determining the matching question template based on the type of resource event, the question template and resource event can be more closely matched, and the question content of the question template is more consistent with the event content of the resource event, thereby improving the accuracy of the first pending question.

[0197] In one possible implementation, the first template includes a question template that asks why a preset condition was met, and the second template includes a question template that asks about the impact of the behavior.

[0198] Since the first type of event is a change in the resource value of the target virtual resource, in order to further improve the accuracy of the first pending question generated based on the first type of event, the first template that matches the first type of event can be used to inquire about the reason for the occurrence of the first type of event, that is, why the resource value reaches the preset condition, so that the first template will match the first type of event better.

[0199] As an example, see reference Figure 9 The diagram shown illustrates a first unresolved question provided in an embodiment of this application. On the resource page 102 of the target virtual resource 101 (i.e., stock X), the first unresolved question 203 obtained through template matching is displayed. Figure 9 In (a), the first question to be determined, 203, is a question used to inquire about the reason for the abnormal stock price movement. The first question to be determined, 203, is "X stock fell by 10% in 5 days due to rumors about AI hardware".

[0200] Similarly, since the second type of event is the subject behavior of the resource subject of the target virtual resource, in order to improve the accuracy of the first pending question generated based on the second type of event, the second template matched with the second type of event can be used to inquire about the impact of the subject behavior.

[0201] Continue to refer to Figure 9 As shown, in Figure 9 In (b), the first question to be determined, 203, is a question used to inquire about the impact of the company's actions. Specifically, the first question to be determined, 203, is "Why did the abnormal activity occur after the executives of Company X reduced their holdings?"

[0202] In summary, when generating the corresponding first pending question based on the type of resource event, if the resource event is a change in resource value, the first template used to generate the first pending question is a template that asks for the reason for the change. If the resource event is a behavior of the resource subject, the second template used to generate the first pending question is a template that asks for the impact caused by the behavior. This makes the question template's questioning logic for resource events more conventional and normal, thereby further improving the matching accuracy of the question template.

[0203] When a resource event includes both a first type of event and a second type of event, in order to improve the accuracy of the generated first pending question, in one possible implementation, step 202, generating a first pending question related to the resource event using the resource event and a matched question template, may include: when the resource event includes both a first type of event and a second type of event, matching a first template related to the first type of event and a second template related to the second type of event from the question templates, and generating a composite question as the first pending question using the first type of event and the first template, as well as the second type of event and the second template.

[0204] In other words, when a resource event simultaneously includes a change in resource value (i.e., the first type of event) and an action taken by the resource entity (i.e., the second type of event), the corresponding templates for each event can be determined from the question templates, namely the first template and the second template. Then, based on the first type of event and the matching first template, and the second type of event and the matching second template, a composite question is generated, which is then designated as the first undetermined question. A composite question is one that asks for both the cause and the effect. For example, a composite question could be, "What are the reasons for the continuous decline in the stock price of stock X, and what adverse effects will it have subsequently?"

[0205] In summary, when a resource event includes two types of events, the first undetermined problem generated is a composite problem, which further enhances the richness of the problem and allows the first undetermined problem to more comprehensively fit the resource event.

[0206] To improve the efficiency of generating the first undetermined problem, in one possible implementation, the method may further include storing the resource event in a second database in response to determining that a resource event related to the target virtual resource has occurred; then step 201, obtaining the resource event related to the target virtual resource, may include obtaining the resource event related to the target virtual resource from the second database.

[0207] In other words, once the server determines that a resource event related to the target virtual resource has occurred, it can store the resource event in a second database. This second database is used to store resource events. Therefore, when a user opens a resource page, the server, in response to the display of the resource page, can directly retrieve the resource event related to the target virtual resource from the second database, without needing to retrieve it from the internet.

[0208] In summary, by storing resource events related to the target virtual resource and retrieving them directly from the second database when needed, the server greatly improves the efficiency of resource event retrieval. As a result, when a user opens a resource page, the server can quickly obtain the resource events to rapidly generate the first pending issue, thus improving the issue generation efficiency.

[0209] Next, refer to Figure 10 The diagram shown is a flowchart illustrating a process for generating a first undetermined problem according to an embodiment of this application.

[0210] When the server determines that a resource event has occurred, such as a change in the resource value of a target virtual resource reaching a preset condition, or a subject of a resource activity, it can perform data validation and store the validated resource events in the database (i.e., the second database). Next, in response to the display of the resource page, the server can retrieve the resource events from the database, match the corresponding question template based on the type of the resource event, and render the question template according to the resource event, thereby generating the first pending question. Furthermore, when a resource value change reaches a preset condition, the server can record the change in resource value. After querying the subject of the resource activity, it can also filter invalid activities. Figure 10 These two steps are not shown in the diagram.

[0211] III. Display of the Target Problem

[0212] Next, we will explain in detail how to display the target issues on the resource page, including the logic of issue filtering, priority sorting, and issue update display before displaying the target issues.

[0213] To improve the accuracy of the second model's response to the input target question, in one possible implementation, the method may further include: determining a second cue word to indicate the question response based on the target question; and inputting the second cue word related to the target question into the second model.

[0214] The second prompt word refers to a prompt word used to guide the second large model on how to answer the question. The server can determine the corresponding second prompt word based on the target question, thus guiding the second large model on how to accurately answer the target question. Therefore, when a user wants to ask the target question displayed at the large model's startup entry point, they can trigger an operation on the large model's startup entry point. In response to this trigger operation, the server can input the second prompt word associated with the target question into the second large model.

[0215] In short, when a user triggers the main model's startup, the terminal device displays a conversation interface, allowing the user to interact with the second main model through question-and-answer dialogue. When the conversation interface is open, the target question is automatically displayed, eliminating the need for manual input. On the server side, the target question and associated second prompts are input into the second main model, enabling it to output an accurate answer. On the terminal device side, the second main model's answer to the target question is displayed in the conversation interface. In other words, the second prompts are used to assist the second main model in answering and may not be displayed in the conversation interface.

[0216] In this way, when using the second model to answer the target question, a second prompt word associated with the target question can be input into the second model, so that the second model can answer the question under the guidance of the second prompt word, avoiding the situation where the second model answers randomly or with poor quality. The second prompt word can effectively constrain the answering logic of the second model, thereby improving the accuracy of the answer to the target question.

[0217] To ensure the timeliness of the displayed target problem, in one possible implementation, step 203, determining the target problem from the first and second undetermined problems, may include: identifying expired events relative to the display time from resource events based on the display time of the resource page; removing the undetermined problems corresponding to the expired events from the first and second undetermined problems, and then determining the target problem.

[0218] The display time of a resource page refers to the moment when the terminal device displays the resource page, specifically the moment the user opens the resource page. Resource events typically have a certain time-sensitivity. If a resource event occurs more than a certain period of time ago, it means that the resource event has occurred a long time ago, and its timeliness is very low, with little reference value at present. In this case, the resource event is an expired event. An expired event is a resource event that has become invalid and no longer has any reference value; questions generated based on expired events are no longer reliable.

[0219] As an example, a resource event might be that the change in the value of a target virtual resource reached an all-time high last year. Since resource values ​​(such as stock prices) are updated too frequently, this resource event is no longer relevant and is therefore considered an expired event. Another example is that a resource event might be the release of shareholder changes last month by the entity responsible for the resource. Because this resource event occurred a long time ago and is already reflected in the stock price changes during that period, it has almost no impact on future stock prices and is also considered an expired event.

[0220] As the server responds to the display of the resource page, it acquires resource events and generates questions based on these events. Since the resource events occur at different times, the timeliness of the generated questions also varies. Some questions are more time-sensitive, while others are less time-sensitive or have even become invalid.

[0221] To ensure the timeliness of the displayed target problem, the acquired resource events need to be filtered for timeliness, identifying expired events. This allows for the subsequent determination of the corresponding expired problem based on expired events, and the identification of the target problem from the remaining expired issues. Based on this, the server can filter resource events from a large number of acquired resource events according to the display time of the resource page, selecting those that have become expired relative to the display time. Since some of the generated first and second pending problems are based on expired events, and some are based on non-expired issues, after identifying expired events, the pending problems corresponding to expired events can be removed from these two categories. The remaining pending problems are the pending problems corresponding to non-expired events, and the target problem identified from the remaining pending problems is also a non-expired problem.

[0222] As an example, a resource event may have a validity period of 3 days. However, if more than 3 days have passed between the time the event occurred and the time the resource page is displayed, then the resource event is considered expired, and the corresponding pending issue will not be displayed as a target issue. In another example, if the resource page is displayed at 12:00, and a resource event that occurred at 11:00 has been overridden or overturned by another resource event that occurred at 11:30, then the resource event that occurred at 11:00 is considered expired, and the corresponding pending issue will not be displayed as a target issue.

[0223] In summary, when determining the target question from the pending questions, if the resource event corresponding to the pending question is an expired event, the pending question will not be used as the target question. This ensures that the target question is not expired, avoids the target question becoming outdated, which could reduce users' desire to ask questions or provide incorrect guidance to users, and ensures that the displayed target question is highly real-time.

[0224] To increase users' desire to ask questions about the target question, in one possible implementation, step 203, determining the target question from the first and second undetermined questions, may include: sorting the first and second undetermined questions from high to low according to their respective question priorities to obtain a question sequence, wherein the question priority is determined based on the generation method and the corresponding resource event of the corresponding undetermined question; and selecting the target question from the question sequence.

[0225] Among them, the questions have corresponding question priorities, which refer to the degree to which the question is displayed. The higher the question priority, the more likely the question will be displayed first, that is, the greater the possibility that the question will be the target question.

[0226] Each problem in the first and second pending problems also has its own corresponding problem priority. The priority of a pending problem can be determined based on its corresponding generation method and resource event. Here, a pending problem can be either one of the first or the second pending problems. The generation method for a pending problem refers to the way it is generated. Generation methods can include template matching, large model generation, etc. As an example, the priority of a problem generated using the large model generation method can be higher than that generated using the template matching method.

[0227] Furthermore, the resource events used to generate pending issues also affect issue priority, such as the importance and real-time nature of the resource events. In short, issue priority can be determined by comprehensively considering both the method of issue generation and the resource events used to generate the issues.

[0228] After determining the priority of each pending problem, the pending problems in the first and second pending problems are sorted from highest to lowest priority to obtain a problem sequence. The closer a pending problem is to the head of the problem sequence, the higher its priority. The server can select a target problem from the problem sequence, specifically from the head of the sequence. For example, the pending problem ranked first in the sequence can be selected as the target problem.

[0229] In summary, by sorting multiple pending questions based on their priority and then selecting the target question from the sequence, we can ensure that the selected target question has a high priority, present better target questions to users, encourage users to ask questions about the target question, and improve the utilization of interactive resources.

[0230] To increase users' desire to ask questions about the target question when they repeatedly enter the same resource page, in one possible implementation, when the resource page is displayed through the target account, when the resource page is displayed for the nth time within a predetermined time period through the target account, the first target question is displayed in the large model startup entry of the resource page. The first target question is different from the target question displayed in the large model startup entry of the resource page when the resource page is displayed for the previous n-1 times. The first target question is the target question determined from the first undetermined question and the second undetermined question.

[0231] The resource page is displayed through the target account, which is the account used to log in to the terminal device. The scheduled duration refers to a pre-set duration, such as 1 hour. If the user opens the resource page for the nth time within the scheduled event using the target account, the target question displayed at the main model launch entry point on the resource page at this time will be the first target question, which is different from the target questions displayed in the previous n-1 times.

[0232] In other words, when the same user opens the resource page of the target virtual resource multiple times, the server can determine a new, previously undisplayed pending question from the first and second pending questions each time the resource page is displayed, and display it as the target question on the resource page.

[0233] As an example, when the same user opens the resource page for the first time, the target question displayed is "Why did stock X fall by 10% yesterday?" When the user opens the resource page for the second time, the target question displayed is updated to "What major moves has stock X made recently?"

[0234] In summary, when the same user opens the same resource page multiple times, the target question displayed on the resource page can be updated and changed, ensuring that the user sees a new target question each time they open the resource page, regardless of whether the user asked the target question in previous visits. By presenting the user with a new target question each time, the likelihood of the target question meeting the user's questioning needs is maximized, increasing the probability of the user asking the target question and thus promoting question-and-answer interaction between the user and the large model.

[0235] In order to further improve the probability of the target question displayed each time hitting the user's question needs when the same user opens the same resource page multiple times, in one possible implementation, when the target question is selected from the question sequence, the method may also include, when the resource page is displayed for the nth time within a predetermined time period through the target account, taking the pending question whose sort position is closest to the head of the question sequence and has not been displayed in the large model startup entry as the first target question.

[0236] In other words, if the target question displayed each time is determined from the question sequence, then the target question displayed for the nth time can be the pending question that is first in the question sequence and has not been displayed before. For example, the question sequence includes pending question Q1, pending question Q2, pending question Q3, and so on, from beginning to end. If the first two pending questions have already been displayed when the user opens the resource page the first two times, then pending question Q3 can be displayed on the resource page when the user opens it for the third time.

[0237] In summary, by selecting the most recent and undisplayed pending questions from the question sequence and displaying them on the resource page, we ensure that each displayed pending question is one that the user has not seen before and has the highest display priority. This increases the likelihood that the user will ask a question about the pending question, thereby further increasing the probability of user interaction with the large model.

[0238] Next, combined Figure 11 This section explains the display process for the target problem. (Reference) Figure 11 The diagram shown is a schematic representation of a processing flow for a target problem provided in an embodiment of this application, taking a stock as an example of the target virtual resource.

[0239] The server queries resource value changes to meet preset conditions, queries the main behavior of the resource entity, and thus obtains resource events. Based on these resource events, a second pending question is generated using the first major model. The server can also query the first pending question generated through template matching, thus obtaining the first and second pending questions. Then, the target virtual resource is matched against the stock market and stock type to obtain a corresponding question list. The pending questions in the question list are then assessed for expiration. Specifically, based on real-time data such as already triggered events for the stock and current market conditions, market anomaly verification and individual stock event verification are performed to identify expired events relative to the display time of the resource page.

[0240] Next, the pending issues corresponding to the expired events are removed from the issue list, leaving the remaining pending issues as valid issues. These valid issues are then sorted according to their priority to obtain an issue sequence, from which the target issue is selected for display. The server can also determine the corresponding second prompt word based on the target issue, so that when the user asks the target issue to the second main model, the second main model can provide a more accurate answer based on the target issue and the second prompt word.

[0241] The following section provides a detailed explanation of the configuration and maintenance of the problem template.

[0242] When configuring an issue template, no specific trigger event needs to be input beforehand. Trigger events can be such as resource events. The issue template can be adapted to all targets (i.e., virtual resources). The issue template can be displayed to users based on time settings or priority order. The issue template can be used to generate the first pending issue in conjunction with resource events. In addition, the issue template can also generate issues without combining resource events (i.e., without trigger events), such as generating issues about event summaries.

[0243] Specifically, question templates can summarize recent news, hot topics, institutional opinions, and future major events. For example, questions generated based on these templates might include: "What are the positive and negative news stories for stock xxx in the past 7 days?", "What major events are about to happen for stock xxx?", and "What are institutional expectations for the performance of stock xxx?". Furthermore, question templates can also be relevant to the current scenario, which refers to the specific resource page being viewed, such as a search page or a newsletter page. For example, on the holdings subpage of stock X's resource page, questions generated based on these templates might include: "How to analyze the top holdings of stock X?" A newsletter is a product feature name, a product format presented to the user.

[0244] To prevent the issue template from becoming outdated, the issue template can be maintained in a timely manner. In one possible implementation, the method may also include updating the issue template according to the issue generation logic. The issue generation logic includes at least one of querying the resource event itself, predicting the impact of the resource event, or the cause of the resource event.

[0245] The question generation logic refers to the logic behind question generation. This logic can involve inquiring about the resource event itself, predicting its potential impact, or investigating its causes. In other words, the server can update the question template from the perspective of inquiring about the resource event itself, for example, "[Name] What does a change in shareholders of a stock mean?" The server can also update the question template from the perspective of predicting the impact of the resource event, for example, "[Name] What impact does the annual report released by a stock have on its stock price?" Finally, the server can update the question template from the perspective of the causes of the resource event, for example, "[Name] What are the reasons for the rise in stock price on [Date]?"

[0246] In summary, by updating the problem template from the perspectives of the resource event itself, its impact, and the reasons for its occurrence, we can ensure that the updated problem template is more in line with human thinking logic, more realistic and accurate, and more real-time.

[0247] As an example, see reference Figure 12The diagram shown is a schematic diagram of an issue template maintenance method provided in this application embodiment. It illustrates information such as template identifier, issue template, coverage market, additional parameters, distribution scenario, issue priority, coverage target, front-end reporting scenario, effective platform, triggering condition, release status, and front-end reporting sub-scenario. By adjusting the parameters of the above information, the issue generation logic can be changed, thereby completing the update of the issue template.

[0248] Next, refer to Figure 13 The following explains how to use the question template. Figure 13 This is a schematic diagram illustrating a process for generating questions using a question template, as provided in an embodiment of this application.

[0249] When a user opens the resource page of the target virtual resource, the server responds to the display of the resource page by loading a pre-configured question template from the configuration center. The configuration and updating of this question template can be done manually by the configuration personnel or automatically by the server. The server matches the question templates obtained from the configuration center with the question templates corresponding to the resource event of the target virtual resource, thus determining the question template matching the resource event. The server then renders the question template, generating the first pending question. Next, the first pending question is filtered and rearranged. From the rearranged questions, the target question is determined. Furthermore, a second prompt word corresponding to the target question can be determined based on the target question. This allows the second model to accurately answer the target question when the user asks it later, guided by the second prompt word. The server then returns the target question to the terminal device and displays it in the main model startup entry on the resource page.

[0250] Combination Figure 14 and Figure 15 As shown, the data processing method provided in this application is described in its entirety.

[0251] refer to Figure 14 The diagram illustrates a multi-combination approach to question generation provided in this application, taking a financial investment scenario with stock X as the target virtual resource. Users primarily focus on three aspects: "What happened?", "Why / What does it mean?", and "What will happen next?". Questions can be asked on all three aspects, meaning the generated questions mainly cover these three areas.

[0252] Regarding "What happened," questions can be generated based on a configured question template, without requiring a triggering event. The question template can be used to summarize recent events, generating a question such as "What are the hot topics for stock X in the past 7 days that are worth paying attention to?". Furthermore, if a resource event occurs but hasn't yet caused any changes, questions about the prospects / probability of changes can be posed using a large model. That is, inputting the resource event of the target virtual resource into the first large model, the first large model outputs a second pending question. For example, the second pending question could be "Will the price reduction of product Y of stock X drag down its stock price?". Additionally, when a scheduled task meets its triggering conditions and no resource event occurs, a forward-looking third pending question can also be generated using the first large model.

[0253] Regarding "What will happen next?", questions can be generated based on question templates without requiring a trigger event. Question templates are used to summarize recent events, and the generated questions could be, for example, "What noteworthy events will occur in stock X?".

[0254] Regarding "why / what does it mean," we can retrieve the resource events that have occurred and generate questions based on the logic that triggered the events. In other words, we use the resource events and matching question templates to generate the first pending question. Specifically, if the resource event is a change in resource value reaching a preset condition, the matching question template is used to inquire about the cause; that is, there is an anomaly, so we ask for the reason. The generated first pending question might be, for example, "Why did stock X break its 60-day high today?" If the resource event is a principal action by the resource entity, the matching question template is used to inquire about the impact of the action; that is, there is an action, so we ask about the impact. The generated first pending question might be, for example, "What information does stock X's financial report reveal?"

[0255] refer to Figure 15 The diagram shown is a schematic representation of the overall flow of a data processing method provided in an embodiment of this application. When a user enters the resource page of a target virtual resource, the resource page is displayed. The display of the resource page indicates that the terminal device wants to query the target question, so that the target question is displayed at the main model startup entry point on the resource page, allowing the user to ask the target question to the secondary model.

[0256] To ensure that the displayed target questions are sufficiently diverse and attractive to users, this application focuses on the question generation service, i.e., how to generate target questions.

[0257] Problem generation can be mainly divided into two methods: one is based on a large model, and the other is based on a template. The large model method, which generates problems based on whether input resource events is required, can be further subdivided into two methods. Similarly, the template method, which also relies on input resource events, can be further subdivided into two methods. See subsequent explanations for details.

[0258] First, let's explain the problem of generating large models. Problems can be generated in advance using large models and stored in the first database. Problems can then be retrieved directly from the first database later.

[0259] When the large model requires resource events as input, these events related to the target virtual resource can be categorized into two main types: the first type occurs when the resource value of the target virtual resource changes to meet a preset condition; the second type occurs when the resource entity of the target virtual resource engages in a subjective action. For both types of resource events, the first large model can be used to generate a second undetermined problem related to the resource event.

[0260] To ensure the accuracy of the generated questions, the corresponding sentence structure requirements, or prompt word requirements, can be determined based on the event type of the resource events. A first prompt word is generated based on the resource events and sentence structure requirements, and this first prompt word is input into the first main model to generate the second pending question. Furthermore, the resource events used to generate the second pending question have a higher event importance.

[0261] When the large model does not require input resource events, the server can respond to the scheduled task meeting the triggering conditions and without resource events, predict the target based on the trend of the target virtual resource, and generate the third undetermined problem through the first large model.

[0262] Next, we will explain how to generate issues using templates. Operations and maintenance personnel can configure and maintain issue templates in the configuration center. To ensure rapid retrieval of resource events, once the server determines that a resource event has occurred, it can store the event in the database (i.e., the second database), from which subsequent resource events can be retrieved directly. The server can load issue templates from the configuration center, query resource events from the database, and generate the first pending issue based on the resource event and the matching issue template. Furthermore, issues related to the target virtual resource, such as summary issues, can be directly generated based on the issue template.

[0263] The server can query generated issues (including first pending issues, second pending issues, third pending issues, and issues generated directly using issue templates without triggering events). To ensure that issues are not expired, expired events relative to the display time of the resource page can be identified from resource events. The pending issues corresponding to the expired events are removed from the queried issues, and the target issues for display are determined from the remaining issues.

[0264] When identifying the target issue, the retrieved issues can be sorted from highest to lowest priority to obtain a problem sequence, from which the target issue can be determined. Furthermore, when the same user opens the same resource page multiple times, the target issue can be updated, with the first issue in the problem sequence that has not yet been displayed identified as the target issue each time.

[0265] To ensure the second major model, which handles the question-and-answer interaction, can accurately respond when a user asks a question about the target issue, a corresponding second prompt word can be assigned to the target question. This way, when a user triggers the target question, the main model service can be invoked, and the second prompt word and the target question can be input into the second major model, improving the accuracy of the answer.

[0266] In summary, this application, targeting virtual resources, yielded a richer set of questions through two question generation methods. In particular, the additional use of a large model to generate questions ensured that the questions were relevant to current hot topics and varied, avoiding rigidity and attracting users to ask questions, thus increasing their willingness to engage in question-and-answer interaction. It also increased the frequency of user interaction with the large model, preventing the waste of interaction resources.

[0267] refer to Figure 16 The diagram shown is a schematic of a data processing apparatus provided in an embodiment of this application. The data processing apparatus 1600 includes:

[0268] The first acquisition unit 1601 is configured to acquire resource events related to the target virtual resource in response to displaying a resource page for the target virtual resource. The resource events include at least one of the following: the resource value of the target virtual resource changes to a preset condition, or the subject behavior of the resource subject of the target virtual resource.

[0269] The second acquisition unit 1602 is used to generate a first pending question related to the resource event through the resource event and the matched question template; and to acquire a second pending question related to the resource event generated through the first large model.

[0270] The determining unit 1603 is used to determine a target question from the first pending question and the second pending question. The target question is used to be displayed in the large model launch entry of the resource page. The large model launch entry is used to display a conversation interface for question-and-answer interaction with the second large model based on a trigger operation. The target question for asking the second large model has been entered in the conversation interface.

[0271] In one possible implementation, the second acquisition unit is used for:

[0272] Generate a first prompt word based on the resource event and the sentence structure requirements corresponding to the resource event;

[0273] Based on the first prompt word, a second pending question related to the resource event is generated through the first large model. The question format of the second pending question is related to the format requirement.

[0274] In one possible implementation, the second acquisition unit is used for:

[0275] Determine the event type of the resource event;

[0276] The syntax requirement corresponding to the resource event is determined based on the event type.

[0277] In one possible implementation, the second acquisition unit is used for:

[0278] In response to determining that a resource event related to the target virtual resource has occurred, a second undetermined question related to the resource event is generated through the first large model and stored in the first database;

[0279] The process of obtaining the second undetermined question related to the resource event generated by the first major model includes:

[0280] Retrieve the second undetermined question from the first database.

[0281] In one possible implementation, the resource events include first-level events and second-level events, where the importance of the second-level events is higher than that of the first-level events. The importance of the events is positively correlated with the rate of change of the resource value of the target virtual resource, or, the importance of the events is positively correlated with the difficulty of the subject's behavior. The second acquisition unit is used for:

[0282] In response to resource events related to the target virtual resource that have a second-level event, a second undetermined question related to the second-level event is generated through the first large model.

[0283] In one possible implementation, the device further includes:

[0284] The first generation unit is used to respond to the timing task meeting the triggering condition, and if there is no resource event related to the target virtual resource, generate a third undetermined problem related to the target virtual resource through the first large model based on the trend prediction target of the target virtual resource.

[0285] The determining unit is used for:

[0286] The target problem is determined from the first undetermined problem, the second undetermined problem, and the third undetermined problem.

[0287] In one possible implementation, the second acquisition unit is used for:

[0288] When the resource event includes a first type of event in which the resource value change of the target virtual resource reaches a preset condition, a first template related to the first type of event is matched from the question template, and the first pending question is generated through the first type of event and the first template.

[0289] When the resource event includes a second type of event involving the principal behavior of the resource subject of the target virtual resource, a second template related to the second type of event is matched from the question template, and the first pending question is generated through the second type of event and the second template.

[0290] In one possible implementation, the first template includes a question template that asks why the preset condition was met, and the second template includes a question template that asks about the impact of the behavior.

[0291] In one possible implementation, the second acquisition unit is used for:

[0292] When the resource event includes the first type of event and the second type of event, a first template related to the first type of event and a second template related to the second type of event are matched from the question templates, and a composite question is generated as the first undetermined question using the first type of event and the first template, as well as the second type of event and the second template.

[0293] In one possible implementation, the device further includes:

[0294] The storage unit is used to store the resource event in the second database in response to determining that a resource event related to the target virtual resource has occurred;

[0295] The first acquisition unit is used for:

[0296] Retrieve resource events related to the target virtual resource from the second database.

[0297] In one possible implementation, the device further includes:

[0298] The first determining unit is configured to determine a second prompt word for indicating a response to the question based on the target question.

[0299] The input unit is used to input the second prompt word, which is associated with the target problem, into the second large model.

[0300] In one possible implementation, the determining unit is used for:

[0301] Based on the display time of the resource page, identify expired events relative to the display time from the resource events;

[0302] After removing the unresolved issues corresponding to the failed events from the first and second unresolved issues, the target issue is determined.

[0303] In one possible implementation, the determining unit is configured to:

[0304] Based on the priority of the first undetermined problem and the second undetermined problem respectively, the first undetermined problem and the second undetermined problem are sorted from high to low to obtain a problem sequence. The problem priority is determined according to the generation method of the corresponding undetermined problem and the corresponding resource event.

[0305] Select the target question from the sequence of questions.

[0306] In one possible implementation, when the resource page is displayed through a target account, when the resource page is displayed for the nth time within a predetermined time period through the target account, the first target question is displayed in the large model launch entry of the resource page. The first target question is different from the target question displayed in the large model launch entry of the resource page when the resource page is displayed for the previous n-1 times. The first target question is a target question determined from the first pending question and the second pending question.

[0307] In one possible implementation, when the target problem is selected from the sequence of problems, the apparatus further includes:

[0308] The first determining unit is used to, when the resource page is displayed for the nth time within a predetermined time period through the target account, select the pending question whose sorting position is closest to the head of the sequence and which has not been displayed in the large model startup entry as the first target question.

[0309] In one possible implementation, the device further includes:

[0310] The question template is updated according to the question generation logic, which includes at least one of querying the resource event itself, predicting the impact of the resource event, or the cause of the resource event.

[0311] This application also provides a computer device, including a terminal device or a server, in which the aforementioned data processing apparatus can be configured. The computer device will now be described in conjunction with the accompanying drawings.

[0312] If the computer device is a terminal device, please refer to Figure 17 As shown, this application provides a terminal device, taking a mobile phone as an example:

[0313] Figure 17 The diagram shown is a block diagram of a portion of the structure of a mobile phone provided in an embodiment of this application. (Reference) Figure 17 The mobile phone includes components such as a radio frequency (RF) circuit 1410, a memory 1420, an input unit 1430, a display unit 1440, a sensor 1450, an audio circuit 1460, a Wi-Fi module 1470, a processor 1480, and a power supply 1490. Those skilled in the art will understand that... Figure 17 The mobile phone structure shown does not constitute a limitation on the mobile phone and may include more or fewer components than shown, or combine certain components, or have different component arrangements.

[0314] The following is combined Figure 17 A detailed introduction to each component of a mobile phone:

[0315] The RF circuit 1410 can be used to receive and transmit signals during information transmission or calls. In particular, it receives downlink information from the base station and processes it with the processor 1480; in addition, it transmits uplink data to the base station.

[0316] The memory 1420 can be used to store software programs and modules. The processor 1480 executes various mobile phone functions and data processing by running the software programs and modules stored in the memory 1420. The memory 1420 may mainly include a program storage area and a data storage area. The program storage area may store the operating system, applications required for at least one function (such as sound playback function, image playback function, etc.), etc.; the data storage area may store data created according to the use of the mobile phone (such as audio data, phonebook, etc.). In addition, the memory 1420 may include high-speed random access memory, and may also include non-volatile memory, such as at least one disk storage device, flash memory device, or other volatile solid-state storage device.

[0317] The input unit 1430 can be used to receive input numeric or character information, and to generate key signal inputs related to user settings and function control of the mobile phone. Specifically, the input unit 1430 may include a touch panel 1431 and other input devices 1432.

[0318] The display unit 1440 can be used to display information input by the user or information provided to the user, as well as various menus of the mobile phone. The display unit 1440 may include a display panel 1441.

[0319] The mobile phone may also include at least one sensor 1450, such as a light sensor, a motion sensor, and other sensors.

[0320] Audio circuitry 1460, speaker 1461, and microphone 1462 provide an audio interface between the user and the mobile phone.

[0321] WiFi is a short-range wireless transmission technology. Through the WiFi module 1470, mobile phones can help users send and receive emails, browse web pages, and access streaming media, providing users with wireless broadband internet access.

[0322] The processor 1480 is the control center of the mobile phone. It connects to various parts of the mobile phone through various interfaces and lines. It performs various functions of the mobile phone and processes data by running or executing software programs and / or modules stored in the memory 1420 and calling data stored in the memory 1420.

[0323] The phone also includes a power supply 1490 (such as a battery) that powers the various components.

[0324] In this embodiment, the processor 1480 included in the terminal device is also used to execute the steps in the methods of the various embodiments of this application.

[0325] If the computer device is a server, this application embodiment also provides a server; please refer to [link to relevant documentation]. Figure 18 As shown, Figure 18 This is a structural diagram of a server 1500 provided in an embodiment of this application. The server 1500 can vary significantly due to different configurations or performance. It may include one or more central processing units (CPUs) 1522 (e.g., one or more processors) and memory 1532, and one or more storage media 1530 (e.g., one or more mass storage devices) for storing application programs 1542 or data 1544. The memory 1532 and storage media 1530 can be temporary or persistent storage. The program stored in the storage media 1530 may include one or more modules (not shown in the diagram), each module including a series of instruction operations on the server. Furthermore, the CPU 1522 may be configured to communicate with the storage media 1530 and execute the series of instruction operations in the storage media 1530 on the server 1500.

[0326] Server 1500 may also include one or more power supplies 1526, one or more wired or wireless network interfaces 1550, one or more input / output interfaces 1558, and / or one or more operating systems 1541, such as Windows Server. TM Mac OS X TM Unix TM Linux TMFreeBSD TM etc.

[0327] The steps performed by the server in the above embodiments can be based on Figure 18 The server structure shown.

[0328] In addition, this application embodiment also provides a storage medium for storing a computer program for executing the method provided in the above embodiment.

[0329] This application also provides a computer program product including a computer program, which, when run on a computer device, causes the computer device to perform the method provided in the above embodiments.

[0330] Those skilled in the art will understand that all or part of the steps of the above method embodiments can be implemented by hardware related to program instructions. The aforementioned program can be stored in a computer-readable storage medium. When the program is executed, it performs the steps of the above method embodiments. The aforementioned storage medium can be at least one of the following media: read-only memory (ROM), RAM, magnetic disk or optical disk, and other media that can store computer programs.

[0331] In the embodiments of this application, the terms "module" or "unit" refer to a computer program or part of a computer program that has a predetermined function and works with other related parts to achieve a predetermined goal, and can be implemented wholly or partially using software, hardware (such as processing circuitry or memory), or a combination thereof. Similarly, a processor (or multiple processors or memory) can be used to implement one or more modules or units. Furthermore, each module or unit can be part of an overall module or unit that includes the functionality of that module or unit.

[0332] It should be noted that the various embodiments in this specification are described in a progressive manner, and the same or similar parts between the various embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, for the device and system embodiments, since they are basically similar to the method embodiments, the description is relatively simple, and the relevant parts can be referred to the description of the method embodiments. The device and system embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and 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 modules can be selected to achieve the purpose of the solution in this embodiment according to actual needs. Those skilled in the art can understand and implement this without creative effort.

[0333] The above description is merely one specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Moreover, based on the implementation methods provided in the above aspects, this application can be further combined to provide more implementation methods.

Claims

1. A data processing method, characterized in that, The method includes: In response to displaying a resource page for a target virtual resource, resource events related to the target virtual resource are obtained, the resource events including at least one of the following: the resource value of the target virtual resource changes to a preset condition, or the subject behavior of the resource subject of the target virtual resource. Using the resource event and the matching question template, a first undetermined question related to the resource event is generated; and a second undetermined question related to the resource event is obtained by generating it through the first large model. The target question is determined from the first pending question and the second pending question. The target question is displayed in the large model launch entry on the resource page. The large model launch entry is used to display a conversation interface for question-and-answer interaction with the second large model based on a trigger operation. The target question for asking the second large model has been entered in the conversation interface.

2. The method according to claim 1, characterized in that, The second undetermined issue was determined in the following manner: Generate a first prompt word based on the resource event and the sentence structure requirements corresponding to the resource event; Based on the first prompt word, a second pending question related to the resource event is generated through the first large model. The question format of the second pending question is related to the format requirement.

3. The method according to claim 2, characterized in that, The method further includes: Determine the event type of the resource event; The syntax requirement corresponding to the resource event is determined based on the event type.

4. The method according to claim 1, characterized in that, The second undetermined issue was determined in the following manner: In response to determining that a resource event related to the target virtual resource has occurred, a second undetermined question related to the resource event is generated through the first large model and stored in the first database; The process of obtaining the second undetermined question related to the resource event generated by the first major model includes: Retrieve the second undetermined question from the first database.

5. The method according to claim 1, characterized in that, The resource events include first-level events and second-level events. The importance of second-level events is higher than that of first-level events. The importance of the events is positively correlated with the rate of change of the resource value of the target virtual resource, or the importance of the events is positively correlated with the difficulty of the subject's behavior. The second undetermined problem is determined in the following way: In response to resource events related to the target virtual resource that have a second-level event, a second undetermined question related to the second-level event is generated through the first large model.

6. The method according to claim 1, characterized in that, The method further includes: In response to the timed task meeting the triggering condition, if there are no resource events related to the target virtual resource, a third undetermined problem related to the target virtual resource is generated through the first large model based on the trend prediction target of the target virtual resource. Determining the target problem from the first undetermined problem and the second undetermined problem includes: The target problem is determined from the first undetermined problem, the second undetermined problem, and the third undetermined problem.

7. The method according to any one of claims 1-6, characterized in that, The step of generating a first pending question related to the resource event using the resource event and the matching question template includes: When the resource event includes a first type of event in which the resource value change of the target virtual resource reaches a preset condition, a first template related to the first type of event is matched from the question template, and the first pending question is generated through the first type of event and the first template. When the resource event includes a second type of event involving the principal behavior of the resource subject of the target virtual resource, a second template related to the second type of event is matched from the question template, and the first pending question is generated through the second type of event and the second template.

8. The method according to claim 7, characterized in that, The first template includes a question template that asks why the preset conditions were met, and the second template includes a question template that asks about the impact of the behavior.

9. The method according to claim 7, characterized in that, The step of generating a first pending question related to the resource event using the resource event and the matching question template includes: When the resource event includes the first type of event and the second type of event, a first template related to the first type of event and a second template related to the second type of event are matched from the question templates, and a composite question is generated as the first undetermined question using the first type of event and the first template, as well as the second type of event and the second template.

10. The method according to claim 7, characterized in that, The method further includes: In response to determining that a resource event related to the target virtual resource has occurred, the resource event is stored in a second database; The acquisition of resource events related to the target virtual resource includes: Retrieve resource events related to the target virtual resource from the second database.

11. The method according to any one of claims 1-6, characterized in that, The method further includes: Determine a second prompt word to indicate the response to the question based on the target question; Input the second prompt word related to the target problem into the second large model.

12. The method according to any one of claims 1-6, characterized in that, Determining the target problem from the first undetermined problem and the second undetermined problem includes: Based on the display time of the resource page, identify expired events relative to the display time from the resource events; After removing the unresolved issues corresponding to the failed events from the first and second unresolved issues, the target issue is determined.

13. The method according to any one of claims 1-6, characterized in that, Determining the target problem from the first undetermined problem and the second undetermined problem includes: Based on the priority of the first undetermined problem and the second undetermined problem respectively, the first undetermined problem and the second undetermined problem are sorted from high to low to obtain a problem sequence. The problem priority is determined according to the generation method of the corresponding undetermined problem and the corresponding resource event. Select the target question from the sequence of questions.

14. The method according to claim 1 or 13, characterized in that, When the resource page is displayed through the target account, when the resource page is displayed for the nth time within a predetermined time period through the target account, the first target question is displayed in the large model launch entry of the resource page. The first target question is different from the target question displayed in the large model launch entry of the resource page when the resource page is displayed for the previous n-1 times. The first target question is the target question determined from the first pending question and the second pending question.

15. The method according to claim 14, characterized in that, When the target problem is selected from the sequence of problems, the method further includes: When the resource page is displayed for the nth time within a predetermined time period through the target account, the pending question that is closest to the head of the sequence and has not been displayed in the large model startup entry is taken as the first target question.

16. The method according to claim 1, characterized in that, The method further includes: The question template is updated according to the question generation logic, which includes at least one of querying the resource event itself, predicting the impact of the resource event, or the cause of the resource event.

17. A data processing apparatus, characterized in that, The device includes: The first acquisition unit is configured to acquire resource events related to the target virtual resource in response to displaying a resource page for the target virtual resource. The resource events include at least one of the following: the resource value of the target virtual resource changes to a preset condition, or the subject behavior of the resource subject of the target virtual resource. The second acquisition unit is configured to generate a first pending question related to the resource event using the resource event and a matching question template; and to acquire a second pending question related to the resource event generated by the first large model. A determining unit is configured to determine a target question from the first pending question and the second pending question. The target question is to be displayed in the large model launch entry on the resource page. The large model launch entry is to display a conversation interface for question-and-answer interaction with the second large model based on a trigger operation. The target question for asking the second large model has been entered in the conversation interface.

18. A computer device, characterized in that, The computer device includes a processor and memory: The memory is used to store computer programs; The processor is configured to perform the method according to any one of claims 1-16 according to the computer program.

19. A computer-readable storage medium, characterized in that, The computer-readable storage medium is used to store a computer program that, when executed by a computer device, performs the method described in any one of claims 1-16.

20. A computer program product comprising a computer program, which, when run on a computer device, causes the computer device to perform the method of any one of claims 1-16.

Images