Data processing system and method based on multi-agent and large model

Abstract

This application discloses a data processing system and method based on multi-agent and large model. The method includes: generating category feature information corresponding to each target data information after information processing, processing each target data information separately, and obtaining target processing results.

Description

Technical Field

[0001] This application relates to the field of computers, specifically to a data processing system and method based on multiple agents and large models. Background Technology

[0002] With the development of computer technology, the development of Large Language Models (LLM) has exploded, and the ways of processing information are also increasing. For example, information is processed through large language models to output the results that users need. However, the current LLM interaction methods are relatively poorly suited for multi-task scenarios. Summary of the Invention

[0003] This application provides a data processing system and method based on multiple agents and large models.

[0004] In a first aspect, embodiments of this application provide a method, including:

[0005] Obtain information to be processed;

[0006] The information to be processed is subjected to information generation processing to obtain at least one target data information, and the category feature information corresponding to each target data information is determined.

[0007] The target data information is processed according to the category feature information to obtain the target processing result.

[0008] Secondly, embodiments of this application also provide a system, including:

[0009] The acquisition module is used to acquire information to be processed.

[0010] The determination module is used to perform information generation processing on the information to be processed to obtain at least one target data information, and to determine the category feature information corresponding to each of the target data information;

[0011] The processing module is used to process the target data information according to the category feature information to obtain the target processing result.

[0012] Thirdly, embodiments of this application also provide an electronic device, which includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the steps in the information processing method described above.

[0013] Fourthly, embodiments of this application also provide a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps in the information processing method described above.

[0014] Fifthly, embodiments of this application also provide a computer program product or computer program, which includes computer instructions stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the computer device to perform the methods provided in various optional implementations of embodiments of this application. Attached Figure Description

[0015] To more clearly illustrate the technical solutions in this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0016] Figure 1 This is a schematic diagram of a scenario for the information processing method provided in an embodiment of this application;

[0017] Figure 2 This is a flowchart illustrating the information processing method provided in an embodiment of this application;

[0018] Figure 3 This is another flowchart of the information processing method provided in the embodiments of this application;

[0019] Figure 4 This is a flowchart of the film and television recommendation model provided in the embodiments of this application;

[0020] Figure 5 This is a schematic diagram of the structure of the information processing system provided in this application;

[0021] Figure 6 This is a schematic diagram of the structure of the electronic device provided in the embodiments of this application. Detailed Implementation

[0022] The technical solutions of this application will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0023] This application provides an information processing method, system, electronic device, and computer-readable storage medium. Specifically, the electronic device includes terminal devices or servers. The terminal devices include, but are not limited to, desktop computers, laptops, tablets, mobile phones, televisions, etc. The server can be an independent physical server, a server cluster composed of multiple physical servers, or a distributed system. It can also be a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, content delivery networks (CDNs), and big data and artificial intelligence platforms. The server can be directly or indirectly connected via wired or wireless communication.

[0024] Please see Figure 1 Taking the execution of an information processing method by a terminal device as an example, the execution of the same information processing method by the server can be understood by referring to the steps of the terminal device executing the same information processing method. The specific execution process of the terminal device executing the information processing method is as follows:

[0025] Terminal device 10 acquires information to be processed, performs information generation processing on the information to be processed to obtain at least one target data information, determines the category feature information corresponding to each target data information, and processes each target data information according to each category feature information to obtain the target processing result.

[0026] It is understandable that by generating multiple target data information corresponding to the information to be processed, determining the category feature information corresponding to each target data information, and processing each target data information according to each category feature information, information can be processed in a multi-task manner, thereby improving the applicability of information processing to multi-task scenarios.

[0027] The following sections provide detailed descriptions of each example. It should be noted that the order in which the embodiments are described is not intended to limit the priority of the embodiments.

[0028] Please see Figure 2 , Figure 2 This is a flowchart illustrating the information processing method provided in an embodiment of this application. The specific flow of this information processing method can be as follows:

[0029] 101. Obtain information to be processed.

[0030] Among them, the information to be processed is the information that needs to be processed, such as search terms entered by the user for the search content, dialogue text entered for the dialogue, voice command information entered for the control of the device, etc., which are not limited in the embodiments of the present invention.

[0031] In this embodiment, the information processing method is illustrated using a smart TV as an example. The information to be processed is input by the user into the smart TV for processing. Examples include search terms for movies / TV shows, and adjustments to the smart TV's basic parameters. These basic parameters are key indicators describing the smart TV's performance and characteristics. Setting or adjusting these parameters changes the smart TV's performance, functions, or user experience. These basic parameters include, but are not limited to, brightness, volume, sharpness, color, and contrast parameters.

[0032] Understandably, to facilitate user interaction with smart TVs, the pending information may contain multiple pieces of information, requests, or commands that require processing by the smart TV. For example, the pending information "Please play me a suspense movie from the top 250 on Douban, preferably from Europe or America, and turn down the brightness and increase the volume after opening it" contains both the request "Please play me a suspense movie from the top 250 on Douban, preferably from Europe or America" ​​and the request "Turn down the brightness and increase the volume after opening it".

[0033] 102. Perform information generation processing on the information to be processed to obtain at least one target data information, and determine the category feature information corresponding to each target data information.

[0034] It is understood that the above information generation and processing can be achieved by decomposing the information to be processed using an artificial intelligence model (such as a large model), or by forming the information to be processed using logical reasoning of an artificial intelligence model (such as a large model). This application does not limit the specific implementation of the information generation and processing.

[0035] Correspondingly, the target data information is obtained through decomposition or logical reasoning of the information to be processed. Taking the example of obtaining target data information through decomposition of the information to be processed, the target data information is a part of the information to be processed. This embodiment of the application decomposes the information to be processed into multiple target data information, making it easier to clarify the various requirements and commands contained in the information to be processed, and facilitating the improvement of the processing efficiency and accuracy of the information to be processed by processing each requirement or command. For example, the information to be processed, "Please play me a suspense movie from the top 250 on Douban, preferably from Europe or America, and turn down the brightness and increase the volume after opening it," can be decomposed into target data information of "Please play me a suspense movie from the top 250 on Douban, preferably from Europe or America" ​​and "Turn down the brightness and increase the volume after opening it."

[0036] In this embodiment of the application, the decomposition of the information to be processed can be achieved through syntactic analysis methods or tools (such as Stanford CoreNLP, spaCy, NLTK, TextBlob, or OpenNLP), semantic analysis methods or tools (such as CoreNLP, AllenNLP, spaCy-srl), writing regular expressions, or machine learning (such as CRF, BiLSTM-CRF, Seq2Seq), etc.

[0037] Since different types of information often require different processing methods—for example, processing film and television recommendation information requires filtering films and television shows through content matching or feature matching before outputting them to users, while processing basic settings information requires generating control commands and adjusting the device's basic parameters—this embodiment of the application, by determining the category feature information of each target data information, helps to control the processing of each target data information based on the category feature information, thereby improving the efficiency and accuracy of information processing.

[0038] The category feature information represents the category or type corresponding to the content of the target data information. For example, category feature information includes, but is not limited to, movie / TV recommendation type, basic settings type, shopping recommendation type, or music recommendation type. For instance, if the target data information involves movie / TV content, then the category feature information includes movie / TV recommendation type; if the target data information involves adjusting the basic parameters of a smart TV, then the category feature information includes basic settings type, etc. It can be understood that in some scenarios, this category feature information is also called domain feature information, that is, it represents the domain involved in the target data information. For example, the target data information for movie / TV recommendations corresponds to the movie / TV recommendation domain, the target data information for basic parameter settings corresponds to the basic settings domain, or the target data information for shopping recommendations corresponds to the shopping recommendation domain, etc.

[0039] In this embodiment, the category feature information corresponding to the target data information can be obtained through keyword extraction and recognition, or through machine learning model recognition and classification. For example, through recognition analysis, the target data information "Please play me suspense movies from the top 250 on Douban, European and American" corresponds to the movie recommendation type, and the target data information "Turn the brightness down a bit and turn up the volume after opening" corresponds to the basic settings type.

[0040] 103. Process the target data information according to the feature information of each category to obtain the target processing result.

[0041] By processing target data based on the feature information of each category, the accuracy of processing target data is improved, thereby improving the accuracy of processing the information to be processed.

[0042] In summary, the embodiments of this application generate multiple target data information corresponding to the information to be processed, determine the category feature information corresponding to each target data information, and process each target data information according to each category feature information, thereby realizing information processing in a multi-task manner and improving the applicability of information processing to multi-task scenarios.

[0043] Specifically, in the embodiments of this application, a machine learning-based information recognition model, a deep learning-based artificial intelligence model, or a large model can be used to generate information from the information to be processed, thereby obtaining multiple target data information corresponding to the information to be processed. Correspondingly, when generating information from the information to be processed based on the information recognition model, prompting engineering can be combined to guide the information recognition model in processing the information to be processed, so as to fully utilize the model's capabilities and improve the matching degree between the model's processing results and the user's expectations. That is, optionally, in some embodiments of this application, the step "generating information from the information to be processed to obtain at least one target data information" includes:

[0044] Confirm the second prompt message;

[0045] Based on the second prompt information, the information to be processed is generated and processed through an information recognition model to obtain at least one target data information.

[0046] Among them, the information recognition model is the named entity recognition model, such as the large model.

[0047] The second prompt information indicates what operations the information recognition model needs to perform and controls the output format of the information recognition model. This second prompt information includes, but is not limited to, text-based information.

[0048] In this embodiment of the application, the second prompt information can be pre-configured. For example, the content that the information recognition model needs to process is determined based on business requirements, and the second prompt information is pre-configured based on the content, or the second prompt information is generated by filling in text descriptions related to the content on the prompt template based on the content.

[0049] Accordingly, in the specific data processing process, the second prompt information is extracted from the database based on the current processing requirements. For example, if the current processing requirement is information generation processing, the second prompt information for information generation processing is extracted from the database that stores the second prompt information. For example, the second prompt information includes information for decomposing the information to be processed or information for logical reasoning about the information to be processed.

[0050] In some cases, a binding relationship can be established between the second prompt message and the information recognition model. When the information recognition model is invoked to perform the information generation and processing task, the second prompt message will be automatically retrieved from the database.

[0051] Taking the information generation and processing, which includes the decomposition of information to be processed, as an example, the second prompt information includes decomposed prompt information. The step "based on the second prompt information, performing information generation and processing on the information to be processed through an information recognition model to obtain at least one target data information" includes:

[0052] Based on the decomposed prompt information, the target object information targeting at least one category feature is identified from the information to be processed through an information recognition model;

[0053] Furthermore, by using an information recognition model, the information to be processed is generated into at least one target data information according to the target object information.

[0054] Here, category features refer to the category of information, which is based on the field or scenario involved. For example, category features include the field of film and television recommendation, the field of music recommendation, or the field of basic parameter settings, or the scenario of film and television recommendation, the scenario of music recommendation, or the scenario of basic parameter settings. Target object information is a component of the information to be processed; it is information within the information to be processed that possesses specific category features. For example, if the information to be processed is in text form, then target object information includes film and television content keyword information for the film and television recommendation field, music content keyword information for the music recommendation field, or basic parameter keyword information for the basic parameter settings field, etc.

[0055] In this process, by identifying the target object information for category features contained in the information to be processed, it is helpful to generate multiple target data information corresponding to the information to be processed based on the target object information. For example, for keyword information in various fields, the information to be processed can be decomposed into multiple target data information based on the keyword information to achieve the generation of target data information.

[0056] Specifically, the decomposition prompt information is a portion of the second prompt information. This decomposition prompt information is used to instruct or guide the information recognition model to perform information generation and processing tasks. This decomposition prompt information can be configured with a unique number or tag during storage. Based on the need to decompose the information to be processed, the decomposition prompt information is retrieved from the database based on the number or tag. For example, the decomposition prompt information could be: "A command to the smart TV has been received, stored in {sentence}. It may contain multiple requirements, and the commands in {sentence} need to be split into individual requirements and stored in {}."

[0057] Optionally, following the decomposition prompt, the information recognition model utilizes its own information processing capabilities to decompose the information to be processed into multiple target data information. For example, given the input: {sentence:[“Please play me a suspense movie from the top 250 on Douban, preferably from Europe or America. After opening it, turn down the brightness and turn up the volume.”]}, the information recognition model, guided by the decomposition prompt, will decompose the information to be processed into the target data information “Please play me a suspense movie from the top 250 on Douban, preferably from Europe or America” and the target data information “After opening it, turn down the brightness and turn up the volume.”

[0058] Furthermore, the second prompt information can also guide the information recognition model to analyze the category feature information of each target data information obtained from the decomposition, and determine the category feature information corresponding to each target data information. That is, optionally, in some embodiments of this application, the step "determining the category feature information corresponding to each target data information" includes:

[0059] Based on the category identification prompts in the second prompt information, the category feature information corresponding to the target data information is identified through the information identification model.

[0060] The category identification prompt information is part of the second prompt information. This category prompt information is used to prompt or guide the information recognition model to perform a category identification task on the target data information, so as to analyze the category feature information of each target data information through the information processing capability of the information recognition model itself. For example, in this embodiment of the application, the category identification prompt information includes: for each requirement (target data information) decomposed in {}, the type information involved in each requirement is identified respectively, and {catogory} displays the type information that you can select. If several requirements involve the same type information, the multiple requirements are concatenated in order and stored in a []. Wherein, {catogory}: {: "Movie Recommendation", "Shopping Recommendation", "Basic Settings"}.

[0061] You need to present your results in the following format:

[0062] {Example format:

[0063] Movie / TV Recommendations: [“”]

[0064] Shopping Recommendation: [“”]

[0065] Basic settings: [“”]}.

[0066] Accordingly, after analyzing the type information of each target data information, the information recognition model adds each target data information to the [] corresponding to its type information. For the target data information "Please play me a suspense movie from the top 250 on Douban, preferably European or American" and "Turn down the brightness and turn up the volume after opening," obtained through the information generation process, the information recognition model outputs the following according to the instructions of the above-mentioned category recognition prompts:

[0067] Movie recommendation: ["Please show me suspense movies from Douban's top 250, preferably from Europe and America."]

[0068] Shopping Recommendation: [“”]

[0069] Basic settings: ["After turning it on, turn the brightness down a bit and turn the volume up"]}.

[0070] Furthermore, to improve the efficiency and accuracy of processing various target data information, this application embodiment also considers the correlation between various target data information, such as the dependency during information processing, that is, whether the processing of the current target data information depends on the processing result of the previous target data information, so as to process each target data information in an orderly manner according to the dependency. That is, optionally, in some embodiments of this application, before the step "processing each target data information according to the feature information of each category to obtain the target processing result", the method further includes:

[0071] Based on the timing recognition prompts in the second prompt information, the processing timing information between each target data information is identified through the information recognition model.

[0072] The timing recognition prompt information is part of the second prompt information. This timing recognition prompt information is used to prompt or instruct the information recognition model to perform the task of recognizing the processing timing between various target data information. Based on the guidance of this timing recognition prompt information and combined with the information processing capabilities of the information recognition model itself, the processing timing information between various target data information is analyzed. This processing timing information is also the dependency or correlation between target data information, indicating the processing order between various target data information.

[0073] For example, in this embodiment of the application, the timing identification prompt information includes: analyzing {sentence} to determine whether there is a dependency relationship between requirements belonging to different domains. The dependency relationship specifically refers to: if task A is a prerequisite for task B, then B depends on A; if task A and task B are not related, then they are independent of each other.

[0074] Please enter the domain name within the curly braces {}. Output your results in the following format:

[0075] {{“”} depends on {“”}}.

[0076] In summary, by constructing a prompting engineering approach, the information recognition model is guided to perform information decomposition, category recognition, and time-series recognition tasks based on decomposed prompting information, category recognition prompting information, and time-series recognition prompting information, respectively. This allows the information recognition model to continuously output the expected results according to the design of the prompting engineering approach, thereby improving the accuracy of the analysis of the information to be processed and facilitating the subsequent processing of the content contained in the information.

[0077] To better understand the content of the second prompt, the decomposed prompt, category identification prompt, and timing identification prompt are combined into a single complete prompt, namely:

[0078] A command for the smart TV has been received and is stored in {sentence}. It may contain multiple requirements. The task requires three steps to complete:

[0079] 1) Divide {sentence} into individual requirements and store them in {}.

[0080] 2) Identify the domain involved in each requirement. {catogory} displays the domain categories you can select. If several requirements involve the same domain, concatenate them in order and store them in a [] array. {catogory}: {“Movie Recommendations”, “Shopping Recommendations”, “Basic Settings”}.

[0081] Your results should be displayed in the following format:

[0082] {Example format:

[0083] Movie / TV Recommendations: [“”]

[0084] Shopping Recommendation: [“”]

[0085] Basic settings: [“”]}.

[0086] 3) Analyze the {sentence} to determine if there are dependencies between requirements from different domains. A dependency specifically means: if task A is a prerequisite for task B, then B depends on A; if task A and task B are unrelated, then they are independent. Please fill in the domain name within the {}. Output your results in the following format:

[0087] {{“”} depends on {“”}}.

[0088] The above are examples of the second prompt information. Steps 1, 2, and 3 correspond to the decomposition prompt information, category identification prompt information, and timing identification prompt information, respectively. Please refer to the relevant documentation for further details. Figure 3 As shown, Figure 3 This is another flowchart of the information processing method provided in this application embodiment. For the information to be processed, "Please play me a suspense movie from the top 250 on Douban, preferably from Europe or America. After opening it, turn down the brightness and turn up the volume," after processing by the second information-enhancing guidance information recognition model, the following result is obtained:

[0089] Movie recommendation: ["Please show me suspense movies from Douban's top 250, preferably from Europe and America."]

[0090] Shopping Recommendation: [“”]

[0091] Basic settings: ["After turning it on, turn the brightness down a bit and turn the volume up"]}.

[0092] The options to "lower the brightness" and "increase the volume" depend on "playing suspense movies from Douban's top 250, preferably from Europe and America". In other words, the basic settings rely on movie and TV recommendations, meaning that the smart TV's basic settings are adjusted only after the movie and TV recommendations are output.

[0093] Understandably, guided by the second prompt, the above instructions were successfully segmented into one movie / TV recommendation task, zero shopping recommendation tasks, and two basic setup tasks; the model determined that the basic setup tasks depended on the movie / TV recommendation tasks. Therefore, subsequent processing of each target data information can proceed according to this dependency relationship, i.e., processing time-series information.

[0094] Understandably, with the development of computer technology, the application scenarios of large language models are constantly increasing. Users input information into large language models to obtain desired results. For example, users input text questions into large language models to wait for them to answer, or input search information into large language models to wait for them to output search results. However, due to the huge number of parameters in large language models, they require enormous computing resources. Therefore, large language models are usually deployed in the cloud. This leads to the need to upload private data to the cloud in some scenarios, and for terminal devices, the task domains involved are relatively fixed. Using large language models with a large number of parameters will lead to a waste of computing resources. Furthermore, large language models are designed for processing tasks in a broad range of domains, making it difficult to guarantee the accuracy of processing tasks in specific domains.

[0095] Therefore, in this embodiment, a target processing model with a small number of parameters is used to solve the problem of task processing accuracy in a single domain. For target data information with feature information of each category, the target data information corresponding to that category of feature information is processed according to the target processing model corresponding to that category of feature information. That is, optionally, in some embodiments of this application, the step "processing each target data information according to each category of feature information to obtain the target processing result" includes:

[0096] Determine the target processing model corresponding to each category of feature information;

[0097] Each target processing model processes the target data information corresponding to that model to obtain the target processing result.

[0098] In this context, target-processing models are specifically trained for a single domain or type of information processing task. Unlike large language models that are based on the entire dataset, target-processing models focus on a single domain or type of information processing task, resulting in higher accuracy. For example, target-processing models include movie / TV recommendation models, infrastructure models, shopping recommendation models, and music recommendation models. In other words, target-processing models are domain-specific versions of large language models. These target-processing models are also sometimes referred to as agents.

[0099] Accordingly, each target data information is processed according to its corresponding target processing model, improving the accuracy of the target processing results. For example, for the target data information of "playing suspense movies from Douban's top 250, especially those from Europe and America", the film and television recommendation model searches for and plays movies and television shows that meet the requirements. For the target data information of "turning the brightness down" and "turning the volume up", the basic settings model is used to set the desired results.

[0100] It should be noted that, in this embodiment, each target processing model is deployed locally on the terminal. For example, the movie / TV recommendation model, shopping recommendation model, music recommendation model, and basic settings model commonly used in smart TVs are deployed locally on the smart TV, and each target processing model is called through different configured interfaces. Therefore, when users perform movie / TV searches or adjust basic settings through the target processing models on the smart TV, there is no need to upload private data to the cloud, ensuring the security of user data.

[0101] Optionally, since there may be dependencies between the various target data information, in order to improve the accuracy of processing the information to be processed in this application embodiment, the processing of each target data information is also controlled based on the dependencies between the various target data information. That is, optionally, in some embodiments of this application, the step "processing the target data information corresponding to each target processing model through each target processing model to obtain the target processing result" includes:

[0102] Determine the processing timing information between various target data;

[0103] Based on the processing sequence information between the target data, the target data corresponding to each target processing model is processed sequentially through each target processing model to obtain the target processing result.

[0104] Among them, the processing timing information refers to the order in which the target data information is processed. This processing timing information can be determined based on the steps described above, namely, "based on the timing identification prompt information in the second prompt information, and identifying the processing timing information between each target data information through the information identification model". That is, the processing timing information between each target data information is analyzed by reasoning through the information identification model. The method of guiding the information identification model based on the timing identification prompt information helps to guide the information identification model to effectively perform the reasoning task of analyzing and processing timing information, and improves the matching degree between the output result of the information identification model and the user's expectations.

[0105] Among them, the processing of timing information reflects the dependencies between various target data information. For example, basic settings such as brightness and volume need to be adjusted only after the recommended movies and TV shows are played. Therefore, the basic setting of "turning down the brightness and turning up the volume after turning it on" depends on the movie and TV show recommendation of "please play me a suspense movie from the top 250 on Douban, from Europe and America". So, the movie and TV show recommendation needs to be executed first, and the brightness and volume of the TV need to be adjusted according to the basic settings when the recommended movie and TV show results are obtained and played.

[0106] like Figure 3 As shown, the video recommendation model is first called to process the task of recommending movies, specifically "please play me suspense movies from the top 250 on Douban, preferably from Europe and America." After the task is successfully executed, the basic settings model is called to process the basic settings task, specifically "turn down the brightness and increase the volume after opening the app." Since the basic settings depend on the video recommendation model, if the video recommendation task fails, the basic settings model does not need to be called to process the basic settings task.

[0107] like Figure 3 As shown, assuming the information to be processed also includes shopping recommendations, music recommendations, and other needs or tasks, and there are no dependencies between these needs or tasks, then their respective target processing models are called in parallel, and their respective model processing results are obtained. The results of each model are then combined to obtain the target processing result; for example, the results of movie / TV show recommendations, basic settings, and music recommendations are combined into the target processing result.

[0108] In order to ensure that each target processing model correctly processes its corresponding target data information, this application embodiment includes prompt information to guide the operation of the target processing models, so that each target processing model can correctly process the target data information. That is, optionally, in some embodiments of this application, the step "processing the target data information corresponding to each target processing model in sequence according to the processing time information between each target data information to obtain the target processing result" includes:

[0109] Determine the first prompt information corresponding to each target processing model;

[0110] According to the processing sequence information between each target data information, each target processing model is called in sequence. For any target processing model, the target data information corresponding to the target processing model is processed according to the first prompt information corresponding to the target processing model to obtain the model processing result corresponding to the target processing model.

[0111] The target processing result is generated based on the model processing results corresponding to each target data information.

[0112] The first prompt information guides the target processing model in handling tasks, indicating which tasks the model needs to process. Each target processing model has its own corresponding first prompt information for its specific tasks. For example, the first prompt information for a target processing model for movie / TV recommendation might include selecting and playing movies / TV shows according to the instructions corresponding to the target data information. Similarly, the first prompt information for a target processing model for basic parameter settings might include adjusting the basic parameters of the smart TV according to the instructions corresponding to the target data information.

[0113] The first prompt information corresponding to each target processing model can be pre-configured based on the processing tasks of each target processing model. Furthermore, a binding relationship can be established between each target processing model and its respective first prompt information. When a target processing model is called to process the corresponding task, the first prompt information bound to that target processing model is called, and the target processing model is guided to execute the corresponding task based on the first prompt information.

[0114] In this process, the introduction of initial prompts helps the target processing model to correctly process the target data information in order to obtain results that meet the user's expectations.

[0115] In this context, the corresponding requirements can also be broken down into multiple parts based on the first prompt information, so as to process the corresponding tasks in a more granular way. That is, optionally, in some embodiments of this application, the first prompt information includes target prompt information, and the step "for any target processing model, process the target data information corresponding to the target processing model according to the first prompt information corresponding to the target processing model to obtain the model processing result" includes:

[0116] For any target processing model, extract at least one object information to be processed from the target data information corresponding to the target processing model based on the target prompt information corresponding to the target processing model.

[0117] The target processing model processes the information of each object to be processed in sequence to obtain the model processing result that matches the information of each object to be processed.

[0118] That is, the target prompt information is used to instruct the target processing model to extract multiple object information to be processed from the target data information. It can be understood that the first prompt information may also include other prompt information besides extracting multiple object information to be processed.

[0119] Extracting information about the objects to be processed from the target data can also be achieved by decomposing the target data. For example, the target data can be decomposed using a target processing model to obtain information about the objects to be processed. It can be understood that this information about the objects to be processed is more refined information obtained by decomposing the target data. Decomposing the target data facilitates more refined or step-by-step processing of the target data. For instance, for multiple objects to be processed after decomposition, each object can be processed progressively to obtain the final model processing result.

[0120] For example, for a movie recommendation model, the corresponding target prompt information could be:

[0121] You will be acting as a domain expert, tasked with breaking down a complex problem into several relatively simpler ones and storing them in a list for the target processing model to solve step-by-step. User questions will be provided in {user question: "}, and should be output in the following format:

[0122] {Secondary Question: [“”]}.

[0123] The above are the target prompt information corresponding to the film and television recommendation model in this application embodiment. Accordingly, in this application embodiment, for the film and television recommendation task or target data information of "Please play me suspense movies from the top 250 on Douban, European and American", the corresponding secondary questions (i.e., the information of the objects to be processed) are as follows:

[0124] Level 2 question: ["Retrieve all movies from the Douban Top 250 list."]

[0125] "Select films from the list that belong to the suspense genre."

[0126] "Select films from the list that belong to the European and American regions."

[0127] Correspondingly, the three secondary problems obtained after the division can be processed in sequence. For example, the target processing model first filters out all the movies in the Douban Top 250 list, then filters out the suspense movies from all the movies in the Douban Top 250 list, and finally filters out the movies from Europe and America from the selected suspense movies.

[0128] Accordingly, in the embodiments of this application, in order to make the target processing results better meet the user's needs and improve user satisfaction, the target data information can also be processed in conjunction with the user's historical records. That is, optionally, in some embodiments of this application, the step "processing each object information to be processed sequentially through the target processing model to obtain the model processing result matching each object information to be processed" includes:

[0129] Determine the historical processing feature information corresponding to the target processing model. The historical processing feature information is constructed based on the historical processing results of the target processing model in historical time.

[0130] The target processing model processes the information of each object to be processed and the historical processing feature information in sequence to obtain the model processing result that matches the information of each object to be processed and the historical processing feature information.

[0131] The historical processing feature information reflects the characteristics of historical processing results, such as user preferences for searching movies and TV shows, the frequency of basic settings adjustments, and habits. This historical processing feature information is constructed based on the historical timeframe and the user's acceptance of the final return result from the target processing model. In other words, if the user accepts the final return result from the target processing model, this round of interaction will be refined and described, incorporating the user's preferences and stored in the memory area. In subsequent tasks, the target processing model can retrieve this memory, making it easier to output content that matches the user's needs.

[0132] In this context, it can be understood that historical time refers to the time prior to the current time. The current time is the time when the information to be processed, the target data information to be processed, or the object information to be processed is being processed. For example, if the information to be processed or the target data information to be processed is being processed at the current moment, then the historical time is the moment prior to the current moment, such as the previous moment, five minutes ago, or one month ago.

[0133] It is understandable that when combining historical processing feature information with the analysis model processing results, the historical processing feature information is essentially added as the information of the objects to be processed, so that the target processing model outputs the model processing results that match the information of each object to be processed. That is, optionally, in some embodiments of this application, the step "processing each object information and historical processing feature information sequentially through the target processing model to obtain the model processing results that match each object information and historical processing feature information" includes:

[0134] Historical processing characteristics are identified as information of the objects to be processed;

[0135] The target processing model sequentially matches the information of each object to be processed, and matches the next object to be processed based on the processing result of the current object to be processed, until all object information is processed, and the model processing result matching each object information is obtained.

[0136] In this context, matching refers to the process of analyzing the relationship or similarity between information and the object to be processed. For example, the target processing model compares the relationship or similarity between each object to be matched and the object to be processed. This analysis process is performed by the target processing model. Relationships and similarities can each be set with corresponding thresholds to distinguish the degree of closeness or similarity. The object to be matched, once successfully matched, becomes the model's processing result.

[0137] For example, taking the information of the objects to be processed as keywords for filtering film and television works, historical processing feature information is also added as keywords, and suitable film and television works are matched through each keyword. In the process of matching the results, matching can be performed step by step based on the information of each object to be processed, such as filtering based on each keyword to obtain the final film and television works. For example, for the information to be processed, "Please play me a suspense movie from the top 250 on Douban, from Europe and America, turn down the brightness and turn up the volume after opening it," "Douban Top 250 list," "suspense," and "Europe and America" ​​are all corresponding information of the objects to be processed, i.e., keywords. In this embodiment, the historical processing feature information "director Kubrick" is used as information of the objects to be processed or a keyword, and matched with film and television works along with "Douban Top 250 list," "suspense," and "Europe and America." For example, after obtaining all the movies in the Douban Top 250 list based on the keyword "Douban Top 250 list", the historical processing feature information "director Kubrick" is used as the object information or keyword to be processed. It is then combined with the keywords "suspense" and "European and American" to obtain the movies in the Douban Top 250 list that belong to the suspense category, are European and American, and are directed by Kubrick.

[0138] After each round of interaction, the goal processing model receives user actions, such as accepting feedback or asking questions again, corresponding to task success and failure, respectively. The goal processing model records the questions and results of successful tasks, a process called "reflection."

[0139] Optionally, in this embodiment, an example template can be provided so that the target processing model can "reflect" according to the example template, wherein the example template is as follows:

[0140] You are a logger. You should record your interactions with users in the following format. Please consult the knowledge base and record the tags of the content selected by the user in the "Reflection" section.

[0141] User question: {"Please play me suspense movies from the top 250 on Douban, preferably from Europe and America"}.

[0142] The result of my execution: {Retrieved [The Shawshank Redemption, The Invisible Guest, ...]}.

[0143] User feedback: {[Select "Invisible Guest"]}.

[0144] Reflection: {"The Invisible Guest", Tags: Region - Spain, Year - 2016, Director - Oriol Paulo}.

[0145] Correspondingly, historical processing features are more refined "reflections," also known as advanced reflections or high-level reflections. These statistics include the frequency of tags associated with the films and television shows viewed by the user, and automatically generated summary descriptions of the user's preferences. This is also implemented using a suggestion project, such as suggestion templates.

[0146] You need to analyze the past twenty "reflections" and complete two tasks. First, sort the tags by frequency; second, generate a paragraph summarizing the deeper connections between the films and television shows mentioned.

[0147] Frequency ranking: {eg: "Suspense": 5 times; "Thriller": 4 times; "Western" and "European and American": 3 times; ...,}.

[0148] Advanced reflection: {eg: "Most of these films come from Europe and America, with dramatic plots and grand scenes."}

[0149] Basic reflections only need to be temporarily stored until higher-level reflections are generated after 20 rounds of interaction. High-level reflections are stored for a long time, which allows for long-term memory of user preferences.

[0150] As user input length and historical records increase, the running speed of large language models or target processing models decreases. It is necessary to clear user history records to ensure immediate feedback, thus preventing the long-term retention of user habits and hindering the improvement of recommended content quality. This application's embodiment addresses this by storing refined historical interaction results—historical processing feature information—and combining this information to process target data, thereby improving the accuracy of target data processing and the model's results. Furthermore, since only refined historical processing feature information needs to be stored, the amount of data required for storage is reduced, lowering the storage burden.

[0151] It is understandable that each target data information is processed by calling its corresponding target processing model. In this embodiment, a film and television recommendation model is used as an example. Target data information of the film and television recommendation type is processed by calling the film and television recommendation model. The film and television playback resources on smart TVs mainly come from TV partner apps, such as iQiyi, Tencent Video, and Youku Video. A film and television knowledge base supporting keyword matching needs to be established in advance, requiring sufficient data reserves and complete classification tags for target processing model queries. In the film and television knowledge base, films and television shows are tagged by region, year, film and television type, awards, director, main actors, ratings, etc.

[0152] For example, please see Figure 4 , Figure 4 This is a flowchart of the film and television recommendation model provided in the embodiments of this application. Specifically:

[0153] 111. Received the target data message "Please play me suspense movies from the top 250 on Douban, preferably from Europe and America";

[0154] 112. Break down the task "Please play me suspense movies from the Douban Top 250 list, preferably from Europe and America" ​​into three steps: ["Get all movies from the Douban Top 250 list.", "Filter out suspense movies from the list.", "Filter out movies from Europe and America from the list."]

[0155] 113. Extract keywords, the keywords are: Douban Top 250 / Suspense / Western;

[0156] 114. Determine whether the keyword points to a specific movie or TV show title. If it does, proceed to step 115; otherwise, proceed to step 116.

[0157] 115. Call the partner's APP API;

[0158] 116. Determine whether the keyword belongs to the knowledge base tag. If yes, proceed to step 117; otherwise, proceed to step 118.

[0159] 117. Search the knowledge base (suspense / Western);

[0160] For example, a knowledge base can be queried by combining keywords with historical processing features. If a user likes Kubrick's movies, then Kubrick's movies can be filtered by keywords.

[0161] Specifically: retrieve the n most frequent tags (ranked by frequency) of the movies the user has watched as supplementary keywords. For example, if the user has frequently watched movies directed by "Kubrick", the target processing model supplements this tag and filters movies in the knowledge base by "suspense", "Western", and "Kubrick".

[0162] Alternatively, descriptions of user preferences (historical processing feature information) stored in the memory area can be used as prompts to input into the target processing model, assisting the model in querying the knowledge base; or, movies that meet the criteria can be ranked based on movie ratings.

[0163] 118. Calling external service APIs;

[0164] 119. Determine if relevant information is found. If yes, proceed to step 120; otherwise, proceed to step 121.

[0165] 120. Obtain a list of relevant movies and TV shows;

[0166] For example, you can call the Douban API to get the Douban top 250 list, store the movie titles in the format of ["The Shawshank Redemption", "Forrest Gump", ...] into the list, and then search for suspense and European and American movies step by step from the list.

[0167] 121. End the task or send an inquiry message to the user.

[0168] Understandably, user requests to watch movies or TV shows fall into two categories:

[0169] 1) Keywords that clearly point to the name of a movie or TV show, such as "I want to watch Peppa Pig".

[0170] The movie recommendation model can identify the movie title and directly enter the partner's APP to search for "Peppa Pig". If a resource is matched, it will jump to the interface; otherwise, it will return "No Peppa Pig resources available".

[0171] 2) Keywords without specific movie or TV show titles, such as "Please play me suspense movies from the top 250 on Douban, from Europe and America".

[0172] If the movie recommendation model cannot find results in the partner's app, it will search the knowledge base. However, the knowledge base does not contain real-time Douban ranking information, so an external service API needs to be called. Preset APIs include: Douban Movie API, TMDb API, IMDb API, etc. (If the API also fails to find relevant information, the task can be terminated or the user can be contacted).

[0173] In addition, other models such as the shopping recommendation model and the infrastructure model process their respective target data information in a similar manner, using knowledge bases and external service APIs.

[0174] In summary, the embodiments of this application generate multiple target data information corresponding to the information to be processed, determine the category feature information corresponding to each target data information, and process each target data information according to each category feature information, thereby realizing information processing in a multi-task manner and improving the applicability of information processing to multi-task scenarios.

[0175] By selecting a target processing model with fewer parameters and stronger domain specificity, this model can be deployed locally on the terminal, eliminating the need for users to upload private data during use and thus protecting user privacy. Furthermore, a more domain-specific target processing model also helps improve the processing quality of relevant information within that domain. For example, for smart TVs, only task domains related to the TV usage scenario can be deployed, reducing the number of target processing models deployed and contributing to more accurate processing of tasks within the TV usage scenario.

[0176] By statistically analyzing users' historical interaction results and saving refined historical processing feature information, it helps to retain users' long-term preference data, avoids the target processing model from periodically clearing users' historical records, which would lead to the inability to remember users' usage habits in the long term, and helps to improve the quality of interaction and feedback content.

[0177] Specifically, by using prompts to break down requirements into more granular information or needs, such as decomposing the information to be processed into multiple target data items, and then further decomposing the target data items into multiple objects to be processed, the system avoids the problem of multiple requirements being mixed into a single instruction, which makes the model difficult to understand and simplifies the model's processing difficulty. Simultaneously, by analyzing the dependencies between various target data items, the system facilitates the processing control of each target data item. This ensures that if the task corresponding to some target data items cannot be completed, it does not affect the execution of other tasks that do not have dependencies, thus maximizing the completion of user instructions.

[0178] To facilitate better implementation of the information processing method of this application, this application also provides an information processing system based on the above-described information processing method. The meaning of the third target term is the same as in the above-described information processing method; specific implementation details can be found in the description of the method embodiments.

[0179] Please see Figure 5 , Figure 5 This is a schematic diagram of the structure of the information processing system provided in this application, wherein the information processing system may include:

[0180] The acquisition module 301 is used to acquire information to be processed;

[0181] The determination module 302 is used to perform information generation processing on the information to be processed, obtain at least one target data information, and determine the category feature information corresponding to each target data information;

[0182] The processing module 303 is used to process the target data information according to the feature information of each category to obtain the target processing result.

[0183] Optionally, in some embodiments of this application, the target data information is processed according to the feature information of each category to obtain the target processing result, including:

[0184] Determine the target processing model corresponding to each category of feature information;

[0185] Each target processing model processes the target data information corresponding to that model to obtain the target processing result.

[0186] Optionally, in some embodiments of this application, the target data information corresponding to each target processing model is processed by each target processing model to obtain the target processing result, including:

[0187] Determine the processing timing information between various target data;

[0188] Based on the processing sequence information between the target data, the target data corresponding to each target processing model is processed sequentially through each target processing model to obtain the target processing result.

[0189] Optionally, in some embodiments of this application, the target data information corresponding to each target processing model is processed sequentially according to the processing timing information between the target data information, to obtain the target processing result, including:

[0190] Determine the first prompt information corresponding to each target processing model;

[0191] According to the processing sequence information between each target data information, each target processing model is called in sequence. For any target processing model, the target data information corresponding to the target processing model is processed according to the first prompt information corresponding to the target processing model to obtain the model processing result corresponding to the target processing model.

[0192] The target processing result is generated based on the model processing results corresponding to each target data information.

[0193] Optionally, in some embodiments of this application, the first prompt information includes target prompt information. For any target processing model, the target data information corresponding to the target processing model is processed according to the first prompt information corresponding to the target processing model to obtain the model processing result, including:

[0194] For any target processing model, extract at least one object information to be processed from the target data information corresponding to the target processing model based on the target prompt information corresponding to the target processing model.

[0195] The target processing model processes the information of each object to be processed in sequence to obtain the model processing result that matches the information of each object to be processed.

[0196] Optionally, in some embodiments of this application, the target processing model sequentially processes the information of each object to be processed to obtain a model processing result matching the information of each object to be processed, including:

[0197] Determine the historical processing feature information corresponding to the target processing model. The historical processing feature information is constructed based on the historical processing results of the target processing model in historical time.

[0198] The target processing model processes the information of each object to be processed and the historical processing feature information in sequence to obtain the model processing result that matches the information of each object to be processed and the historical processing feature information.

[0199] Optionally, in some embodiments of this application, the target processing model sequentially processes the information of each object to be processed and the historical processing feature information to obtain the model processing result that matches the information of each object to be processed and the historical processing feature information, including:

[0200] Add historical processing feature information as information for objects to be processed;

[0201] The target processing model sequentially matches the information of each object to be processed, and matches the next object to be processed based on the processing result of the current object to be processed, until all object information is processed, and the model processing result matching each object information is obtained.

[0202] Optionally, in some embodiments of this application, information generation processing is performed on the information to be processed to obtain at least one target data information, including:

[0203] Confirm the second prompt message;

[0204] Based on the second prompt information, the information to be processed is generated and processed through an information recognition model to obtain at least one target data information.

[0205] Optionally, in some embodiments of this application, the second prompt information includes decomposed prompt information. Based on the second prompt information, information generation processing is performed on the information to be processed through an information recognition model to obtain at least one target data information, including:

[0206] Based on the decomposition prompt information, target object information targeting at least one category feature is identified from the information to be processed through an information recognition model;

[0207] Furthermore, by using an information recognition model, the information to be processed is generated into at least one target data information according to the target object information.

[0208] Optionally, in some embodiments of this application, determining the category feature information corresponding to each target data information includes:

[0209] Based on the category identification prompts in the second prompt information, the category feature information corresponding to the target data information is identified through the information identification model.

[0210] Optionally, in some embodiments of this application, before processing the target data information according to the feature information of each category to obtain the target processing result, the method further includes:

[0211] Based on the timing recognition prompts in the second prompt information, the processing timing information between each target data information is identified through the information recognition model.

[0212] In this embodiment of the application, the acquisition module 301 acquires the information to be processed, the determination module 302 performs information generation processing on the information to be processed to obtain at least one target data information, and determines the category feature information corresponding to each target data information. The processing module 303 processes each target data information according to each category feature information to obtain the target processing result.

[0213] In summary, the embodiments of this application generate multiple target data information corresponding to the information to be processed, determine the category feature information corresponding to each target data information, and process each target data information according to each category feature information, thereby realizing information processing in a multi-task manner and improving the applicability of information processing to multi-task scenarios.

[0214] In addition, this application also provides an electronic device, such as Figure 6 As shown, it illustrates the structural diagram of the electronic device involved in this application, specifically:

[0215] The electronic device may include components such as a processor 401 with one or more processing cores, a memory 402 with one or more computer-readable storage media, a power supply 403, and an input unit 404. Those skilled in the art will understand that... Figure 6 The electronic device structure shown does not constitute a limitation on the electronic device and may include more or fewer components than shown, or combine certain components, or have different component arrangements. Wherein:

[0216] The processor 401 is the control center of the electronic device, connecting various parts of the device via various interfaces and lines. It executes software programs and / or modules stored in the memory 402, and calls data stored in the memory 402, to perform various functions and process data. Optionally, the processor 401 may include one or more processing cores; preferably, the processor 401 may integrate an application processor and a modem processor, wherein the application processor mainly handles the operating system, user interface, and applications, and the modem processor mainly handles wireless communication. It is understood that the modem processor may not be integrated into the processor 401.

[0217] The memory 402 can be used to store software programs and modules. The processor 401 executes various functional applications and data processing by running the software programs and modules stored in the memory 402. The memory 402 may mainly include a program storage area and a data storage area. The program storage area may store the operating system, application programs 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 electronic device, etc. In addition, the memory 402 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. Accordingly, the memory 402 may also include a memory controller to provide the processor 401 with access to the memory 402.

[0218] The electronic device also includes a power supply 403 that supplies power to the various components. Preferably, the power supply 403 can be logically connected to the processor 401 through a power management system, thereby enabling functions such as charging, discharging, and power consumption management through the power management system. The power supply 403 may also include one or more DC or AC power supplies, recharging systems, power fault detection circuits, power converters or inverters, power status indicators, and other arbitrary components.

[0219] The electronic device may also include an input unit 404, which can be used to receive input digital or character information, and generate keyboard, mouse, joystick, optical or trackball signal inputs related to user settings and function control.

[0220] Although not shown, the electronic device may also include a display unit, etc., which will not be described in detail here. Specifically, in this embodiment, the processor 401 in the electronic device loads the executable files corresponding to the processes of one or more application programs into the memory 402 according to the following instructions, and the processor 401 runs the application programs stored in the memory 402, thereby implementing the steps in any of the information processing methods provided in this application, specifically:

[0221] This application embodiment obtains information to be processed, performs information generation processing on the information to be processed to obtain at least one target data information, determines the category feature information corresponding to each target data information, processes each target data information according to each category feature information, and obtains the target processing result.

[0222] In this embodiment, multiple target data information corresponding to the information to be processed are generated, and the category feature information corresponding to each target data information is determined. The target data information is then processed according to the category feature information, thereby realizing information processing in a multi-task manner and improving the applicability of information processing to multi-task scenarios.

[0223] Those skilled in the art will understand that all or part of the steps in the various methods of the above embodiments can be performed by instructions, or by instructions controlling related hardware. These instructions can be stored in a computer-readable storage medium and loaded and executed by a processor.

[0224] Therefore, this application provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of any of the information processing methods provided in this application.

[0225] The computer-readable storage medium may include: read-only memory (ROM), random access memory (RAM), disk or optical disk, etc.

[0226] Since the instructions stored in the computer-readable storage medium can execute the steps in any of the information processing methods provided in this application, the beneficial effects that any of the information processing methods provided in this application can achieve can be realized, as detailed in the preceding embodiments, and will not be repeated here.

[0227] The above provides a detailed description of an information processing method, system, electronic device, and computer-readable storage medium provided in this application. Specific examples have been used to illustrate the principles and implementation methods of the present invention. The description of the above embodiments is only for the purpose of helping to understand the method and core ideas of the present invention. At the same time, those skilled in the art will recognize that there will be changes in the specific implementation methods and application scope based on the ideas of the present invention. Therefore, the content of this specification should not be construed as a limitation of the present invention.

Claims

1. A method, characterized in that, include: Obtain information to be processed; The information to be processed is subjected to information generation processing to obtain at least one target data information, and the category feature information corresponding to each target data information is determined. The target data information is processed according to the category feature information to obtain the target processing result.

2. The method according to claim 1, characterized in that, The step of processing the target data information according to the category feature information to obtain the target processing result includes: Determine the target processing model corresponding to each of the aforementioned category feature information; The target data information corresponding to each target processing model is processed by each target processing model to obtain the target processing result.

3. The method according to claim 2, characterized in that, The step of processing the target data information corresponding to each target processing model through each target processing model to obtain the target processing result includes: Determine the processing timing information between each of the target data information; According to the processing timing information between the target data information, the target data information corresponding to each target processing model is processed sequentially through each target processing model to obtain the target processing result.

4. The method according to claim 3, characterized in that, The step of processing the target data information corresponding to each target processing model sequentially through the processing time sequence information between the target data information to obtain the target processing result includes: Determine the first prompt information corresponding to each of the target processing models; According to the processing timing information between the target data information, each target processing model is called sequentially, and for any target processing model, the target data information corresponding to the target processing model is processed according to the first prompt information corresponding to the target processing model to obtain the model processing result corresponding to the target processing model. The target processing result is generated based on the model processing result corresponding to each of the target data information.

5. The method according to claim 4, characterized in that, The first prompt information includes target prompt information; For any of the target processing models, the target data information corresponding to the target processing model is processed according to the first prompt information corresponding to the target processing model to obtain the model processing result, including: For any of the target processing models, at least one object information to be processed is extracted from the target data information corresponding to the target processing model based on the target prompt information corresponding to the target processing model. The target processing model processes the information of each object to be processed sequentially to obtain the model processing result that matches the information of each object to be processed.

6. The method according to claim 5, characterized in that, The step of processing each of the object information to be processed sequentially through the target processing model to obtain the model processing result matching each of the object information to be processed includes: Determine the historical processing feature information corresponding to the target processing model, wherein the historical processing feature information is constructed based on the historical processing results of the target processing model in historical time. The target processing model processes each of the object information to be processed and the historical processing feature information in sequence to obtain the model processing result that matches each of the object information to be processed and the historical processing feature information.

7. The method according to claim 6, characterized in that, The step of processing each of the object information to be processed and the historical processing feature information sequentially through the target processing model to obtain the model processing result matching each of the object information to be processed and the historical processing feature information includes: The historical processing feature information is identified as the object information to be processed; The target processing model sequentially matches the information of each object to be processed, and matches the next object to be processed based on the processing result of the current object to be processed, until all the object to be processed information is processed, and the model processing result matching each object to be processed information is obtained.

8. The method according to claim 1, characterized in that, The process of generating information from the information to be processed to obtain at least one target data information includes: Confirm the second prompt message; Based on the second prompt information, the information to be processed is processed by an information recognition model to obtain at least one target data information.

9. The method according to claim 8, characterized in that, The second prompt information includes decomposed prompt information. The step of generating information based on the second prompt information and using an information recognition model to process the information to be processed to obtain at least one target data information includes: Based on the decomposition prompt information, the target object information for at least one category feature is identified from the information to be processed using the information recognition model; Furthermore, the information to be processed is generated into at least one target data information according to the target object information through the information recognition model.

10. The method according to claim 8, characterized in that, Before processing the target data information according to the category feature information to obtain the target processing result, the method further includes: Based on the timing recognition prompt information in the second prompt information, the processing timing information between each of the target data information is identified through the information recognition model.

11. A system, characterized in that, include: The acquisition module is used to acquire information to be processed. The determination module is used to perform information generation processing on the information to be processed to obtain at least one target data information, and to determine the category feature information corresponding to each of the target data information; The processing module is used to process the target data information according to the category feature information to obtain the target processing result; Optionally, the processing module processes the target data information according to the category feature information to obtain the target processing result, including: Determine the target processing model corresponding to each of the aforementioned category feature information; The target data information corresponding to each target processing model is processed by each target processing model to obtain the target processing result; Optionally, the processing module processes the target data information corresponding to each target processing model through each target processing model to obtain the target processing result, including: Determine the processing timing information between each of the target data information; According to the processing timing information between the target data information, the target data information corresponding to each target processing model is processed sequentially through each target processing model to obtain the target processing result; Optionally, the processing module processes the target data information corresponding to each target processing model sequentially through the processing timing information between the target data information, to obtain the target processing result, including: Determine the first prompt information corresponding to each of the target processing models; According to the processing timing information between the target data information, each target processing model is called sequentially, and for any target processing model, the target data information corresponding to the target processing model is processed according to the first prompt information corresponding to the target processing model to obtain the model processing result corresponding to the target processing model. Generate target processing results based on the model processing results corresponding to each of the target data information; Optionally, the first prompt information includes target prompt information. The processing module, for any target processing model, processes the target data information corresponding to that target processing model based on the first prompt information corresponding to that target processing model to obtain a model processing result, including: For any of the target processing models, at least one object information to be processed is extracted from the target data information corresponding to the target processing model based on the target prompt information corresponding to the target processing model. The target processing model processes the information of each object to be processed sequentially to obtain the model processing result that matches the information of each object to be processed. Optionally, the processing module processes each of the object information to be processed sequentially through the target processing model to obtain a model processing result matching each of the object information to be processed, including: Determine the historical processing feature information corresponding to the target processing model, wherein the historical processing feature information is constructed based on the historical processing results of the target processing model in historical time. The target processing model processes each of the object information to be processed and the historical processing feature information in sequence to obtain the model processing result that matches each of the object information to be processed and the historical processing feature information. Optionally, the processing module processes each of the object information to be processed and the historical processing feature information sequentially through the target processing model to obtain a model processing result that matches each of the object information to be processed and the historical processing feature information, including: Add the historical processing feature information as the object information to be processed; The target processing model sequentially matches the information of each object to be processed, and matches the next object to be processed based on the processing result of the current object to be processed, until all object information to be processed is processed, and the model processing result matching each object information to be processed is obtained. Optionally, the determining module performs information generation processing on the information to be processed to obtain at least one target data information, including: Confirm the second prompt message; Based on the second prompt information, the information to be processed is processed by an information recognition model to obtain at least one target data information. Optionally, the second prompt information includes decomposed prompt information. The determining module, based on the second prompt information and using an information recognition model, performs information generation processing on the information to be processed to obtain at least one target data information, including: Based on the decomposition prompt information, the target object information for at least one category feature is identified from the information to be processed using the information recognition model; Furthermore, the information to be processed is generated into at least one target data information according to the target object information through the information recognition model; Optionally, before the determining module processes the target data information according to the category feature information to obtain the target processing result, the determining module further includes: Based on the timing recognition prompt information in the second prompt information, the processing timing information between each of the target data information is identified through the information recognition model.

12. An electronic device, characterized in that, The electronic device includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the method as described in any one of claims 1-10.

13. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, implements the steps of the method as described in any one of claims 1-10.

Images