Task processing method, dialogue task processing method, task planning model training method, information processing method based on task planning model, and model training platform

Abstract

Provided in the embodiments of the present disclosure are a task processing method, a dialogue task processing method, a task planning model training method, an information processing method based on a task planning model, and a model training platform. The task processing method comprises: acquiring task data of a target task; inputting the task data and object description information of candidate processing objects into a task planning model, and determining a plurality of processing objects corresponding to the target task, wherein the task planning model is used for planning the plurality of processing objects of the target task; using the plurality of processing objects to process the task data, so as to obtain processing results respectively output by the plurality of processing objects; and on the basis of the processing results respectively output by the plurality of processing objects, determining a task processing result of the target task. By means of using a task planning model to determine a plurality of processing objects for processing a target task, seamless integration between the model and the processing objects having different processing capabilities is realized, thereby expanding the task processing capability and making task processing more efficient and comprehensive.

Description

Task processing method, dialogue task processing method, task planning model training method, information processing method based on task planning model and model training platform

[0001] This disclosure claims priority to the Chinese patent application filed with the China Patent Office on March 21, 2024, with application number 202410333298.1 and application name “Task processing method, dialogue task processing method, task planning model training method, information processing method based on task planning model and model training platform”, the entire content of which is incorporated by reference into this disclosure. Technical Field

[0002] The embodiments of the present disclosure relate to the field of computer technology, and in particular to a task processing method, a dialogue task processing method, a task planning model training method, an information processing method based on a task planning model, and a model training platform. Background Art

[0003] With the advancement of computer technology, building an intelligent system capable of multimodal conversational interaction, similar to that of humans, has long been a long-term goal in achieving artificial general intelligence. While traditional intelligent systems have achieved high levels of performance in text-based conversations, they still struggle with multimodal interactions like voice, images, and video.

[0004] Currently, a user-entered task can be broken down into several subtasks. The user then selects the corresponding model from the model library, processes the subtask using the corresponding model, and outputs the results. However, this approach requires selecting the corresponding model from the model library, resulting in extremely low task processing efficiency. Therefore, a more efficient task processing solution is urgently needed. Summary of the Invention

[0005] In view of this, embodiments of the present disclosure provide a task processing method. One or more embodiments of the present disclosure also relate to a method for processing a dialogue task, a method for training a task planning model, an information processing method based on a task planning model, a model training platform, a task processing device, a dialogue task processing device, a task planning model training device, an information processing device based on a task planning model, a computing device, a computer-readable storage medium, and a computer program product to address technical deficiencies in the prior art.

[0006] According to a first aspect of an embodiment of the present disclosure, there is provided a task processing method, comprising:

[0007] Get the task data of the target task;

[0008] Inputting the task data and the object description information of the candidate processing objects into the task planning model to determine a plurality of processing objects corresponding to the target task, wherein the task planning model is used to plan the plurality of processing objects of the target task;

[0009] Using multiple processing objects to process task data, and obtaining processing results output by the multiple processing objects respectively;

[0010] The task processing result of the target task is determined according to the processing results respectively output by the multiple processing objects.

[0011] According to a second aspect of an embodiment of the present disclosure, a method for processing a dialog task is provided, including:

[0012] Obtaining task data of a target dialogue task, wherein the task data includes multiple dialogue roles;

[0013] Inputting the task data and the object description information of the candidate processing objects into the task planning model to determine multiple processing objects corresponding to the target dialogue task, wherein the task planning model is used to plan the multiple processing objects of the target dialogue task, and the processing objects correspond to the dialogue roles one by one;

[0014] Using multiple processing objects to process task data, and obtaining processing results output by the multiple processing objects respectively;

[0015] The task processing result of the target dialogue task is determined according to the processing results respectively output by the multiple processing objects.

[0016] According to a third aspect of an embodiment of the present disclosure, a task planning model training method is provided, comprising:

[0017] Acquire a plurality of sample task data, wherein the sample task data carries sample description information and object planning results of a sample processing object;

[0018] Inputting multiple sample task data and sample description information into the task planning model to obtain object prediction results corresponding to the multiple sample task data;

[0019] According to the object planning results and object prediction results, the parameters of the task planning model are adjusted to obtain the trained task planning model.

[0020] According to a fourth aspect of an embodiment of the present disclosure, there is provided an information processing method based on a task planning model, comprising:

[0021] receiving a task generation request, wherein the task generation request includes request information;

[0022] Based on the request information, a task planning model is obtained, wherein the task planning model is trained based on object prediction results corresponding to a plurality of sample task data and object planning results carried by the plurality of sample task data, the sample task data carries sample description information of the sample processing object, and the object prediction result is obtained by inputting the plurality of sample task data and the sample description information into the task planning model to be trained;

[0023] Based on the task planning model, task information is generated, wherein the task information is used to execute the target task.

[0024] According to a fifth aspect of an embodiment of the present disclosure, there is provided a task processing device, including:

[0025] A first acquisition module is configured to acquire task data of a target task;

[0026] a first input module configured to input task data and object description information of candidate processing objects into a task planning model to determine a plurality of processing objects corresponding to a target task, wherein the task planning model is used to plan the plurality of processing objects of the target task;

[0027] A first processing module is configured to process the task data using a plurality of processing objects to obtain processing results outputted by the plurality of processing objects respectively;

[0028] The first determining module is configured to determine a task processing result of the target task according to processing results respectively output by a plurality of processing objects.

[0029] According to a sixth aspect of an embodiment of the present disclosure, there is provided a dialog task processing apparatus, comprising:

[0030] A second acquisition module is configured to acquire task data of a target dialogue task, wherein the task data includes a plurality of dialogue roles;

[0031] a second input module configured to input the task data and object description information of the candidate processing objects into the task planning model to determine a plurality of processing objects corresponding to the target dialogue task, wherein the task planning model is used to plan the plurality of processing objects of the target dialogue task, and the processing objects correspond one-to-one to the dialogue roles;

[0032] A second processing module is configured to process the task data using multiple processing objects to obtain processing results output by the multiple processing objects respectively;

[0033] The second determining module is configured to determine the task processing result of the target dialogue task according to the processing results respectively output by the multiple processing objects.

[0034] According to a seventh aspect of an embodiment of the present disclosure, a task planning model training device is provided, comprising:

[0035] A third acquisition module is configured to acquire a plurality of sample task data, wherein the sample task data carries sample description information and object planning results of the sample processing object;

[0036] A third input module is configured to input a plurality of sample task data and sample description information into the task planning model to obtain object prediction results corresponding to the plurality of sample task data respectively;

[0037] The adjustment module is configured to adjust the parameters of the task planning model according to the object planning results and the object prediction results to obtain a trained task planning model.

[0038] According to an eighth aspect of an embodiment of the present disclosure, there is provided an information processing device based on a task planning model, comprising:

[0039] A receiving module is configured to receive a task generation request, wherein the task generation request includes request information;

[0040] a fourth acquisition module configured to acquire a task planning model based on the request information, wherein the task planning model is trained based on object prediction results corresponding to the plurality of sample task data and object planning results carried by the plurality of sample task data, the sample task data carrying sample description information of the sample processing object, and the object prediction results being obtained by inputting the plurality of sample task data and the sample description information into the task planning model to be trained;

[0041] The generation module is configured to generate task information based on the task planning model, wherein the task information is used to execute the target task.

[0042] According to a ninth aspect of an embodiment of the present disclosure, there is provided a model training platform, comprising a request interface unit and a model unit;

[0043] A request interface unit, configured to receive a model training request, wherein the model training request includes request information;

[0044] The model unit is used to obtain a task planning model; based on the request information, the task planning model is trained to obtain a trained task planning model, wherein the request information includes multiple sample task data, and the sample task data carries sample description information and object planning results of the sample processing object.

[0045] According to a tenth aspect of an embodiment of the present disclosure, there is provided a computing device, including:

[0046] memory and processor;

[0047] The memory is used to store computer programs / instructions, and the processor is used to execute the computer programs / instructions. When the computer programs / instructions are executed by the processor, the steps of the method provided in the first aspect, the second aspect, the third aspect, or the fourth aspect are implemented.

[0048] According to the eleventh aspect of the embodiments of the present disclosure, a computer-readable storage medium is provided, which stores a computer program / instruction, and when the computer program / instruction is executed by a processor, the steps of the method provided in the first aspect, the second aspect, the third aspect or the fourth aspect are implemented.

[0049] According to the twelfth aspect of the embodiments of the present disclosure, a computer program product is provided, comprising a computer program / instruction, which, when executed by a processor, implements the steps of the method provided in the first aspect, the second aspect, the third aspect, or the fourth aspect.

[0050] One embodiment of the present disclosure provides a task processing method, comprising: obtaining task data for a target task; inputting the task data and object description information of candidate processing objects into a task planning model to determine multiple processing objects corresponding to the target task, wherein the task planning model is used to plan the multiple processing objects for the target task; processing the task data using the multiple processing objects to obtain processing results output by the multiple processing objects; and determining a task processing result for the target task based on the processing results output by the multiple processing objects. By using the task planning model to determine the multiple processing objects for processing the target task, seamless integration between the model and processing objects with different processing capabilities is achieved, expanding task processing capabilities and making task processing more efficient and comprehensive. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] FIG1 is a flow chart of a question-answering method;

[0052] FIG2 is an architecture diagram of a task processing system provided by one embodiment of the present disclosure;

[0053] FIG3 is a flowchart of a task processing method provided by one embodiment of the present disclosure;

[0054] FIG4 is a flowchart of another task processing method provided by an embodiment of the present disclosure;

[0055] FIG5 is a flowchart of a task processing method according to an embodiment of the present disclosure;

[0056] FIG6 is a flowchart of a method for processing a dialogue task provided by an embodiment of the present disclosure;

[0057] FIG7 is a schematic diagram of a dialogue task processing interface provided by an embodiment of the present disclosure;

[0058] FIG8 is a flowchart of a task planning model training method provided by one embodiment of the present disclosure;

[0059] FIG9 is a flowchart of an information processing method based on a task planning model provided by one embodiment of the present disclosure;

[0060] FIG10 is a schematic diagram of a model training platform provided by one embodiment of the present disclosure;

[0061] FIG11 is a schematic structural diagram of a task processing device provided by one embodiment of the present disclosure;

[0062] FIG12 is a structural diagram of a dialog task processing device provided by one embodiment of the present disclosure;

[0063] FIG13 is a schematic structural diagram of a task planning model training device provided by one embodiment of the present disclosure;

[0064] FIG14 is a schematic diagram of the structure of an information processing device based on a task planning model provided by one embodiment of the present disclosure;

[0065] FIG15 is a structural block diagram of a computing device provided by an embodiment of the present disclosure. DETAILED DESCRIPTION

[0066] The following description sets forth many specific details to facilitate a full understanding of the present disclosure. However, the present disclosure can be implemented in many other ways than those described herein, and those skilled in the art can make similar generalizations without violating the scope of the present disclosure. Therefore, the present disclosure is not limited to the specific implementations disclosed below.

[0067] The terms used in one or more embodiments of the present disclosure are for the purpose of describing specific embodiments only and are not intended to limit one or more embodiments of the present disclosure. The singular forms "a", "the", and "the" used in one or more embodiments of the present disclosure and the appended claims are also intended to include plural forms, unless the context clearly indicates otherwise. It should also be understood that the term "and / or" used in one or more embodiments of the present disclosure refers to and includes any or all possible combinations of one or more associated listed items.

[0068] It should be understood that although the terms first, second, etc. may be used to describe various information in one or more embodiments of the present disclosure, such information should not be limited to these terms. These terms are only used to distinguish information of the same type from each other. For example, without departing from the scope of one or more embodiments of the present disclosure, the first may also be referred to as the second, and similarly, the second may also be referred to as the first. Depending on the context, the word "if" as used herein may be interpreted as "at the time of" or "when" or "in response to determining".

[0069] In addition, it should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in one or more embodiments of the present disclosure are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with the relevant laws, regulations and standards of the relevant regions, and provide corresponding operation entrances for users to choose to authorize or refuse.

[0070] In one or more embodiments of the present disclosure, a large model refers to a deep learning model with large-scale model parameters, which typically contains hundreds of millions, tens of billions, hundreds of billions, trillions, or even more than ten trillion model parameters. A large model can also be called a cornerstone model / foundation model. The large model is pre-trained using large-scale unlabeled corpus to produce a pre-trained model with more than 100 million parameters. This model can adapt to a wide range of downstream tasks and has good generalization capabilities, such as a large-scale language model (LLM), a multi-modal pre-training model, etc.

[0071] When large models are used in practice, only a small number of samples are needed to fine-tune the pre-trained model and it can be applied to different tasks. Large models can be widely used in natural language processing (NLP), computer vision and other fields. Specifically, they can be applied to computer vision tasks such as visual question answering (VQA), image description (IC), image generation, as well as natural language processing tasks such as text-based sentiment classification, text summary generation, and machine translation. The main application scenarios of large models include digital assistants, intelligent robots, search, online education, office software, e-commerce, intelligent design, etc.

[0072] First, the terms involved in one or more embodiments of the present disclosure are explained.

[0073] Agent system: refers to an artificial intelligence entity that relies on a large model as the core, combines tool calls and knowledge bases to complete more complex tasks, and also refers to the task processing system in the embodiments of the present disclosure.

[0074] Large Language Model (LLM): Large language models are neural network models with billions or even trillions of parameters. These models can be used for natural language processing tasks such as language generation, question answering, and text classification. These models are typically trained using large amounts of text data and are able to perform well on a variety of different tasks.

[0075] Generative AI (GAI): GAI refers to systems based on AI technology that can generate data such as natural language text, images, audio, or video. Unlike traditional tasks like classification and regression, GAI models are creative and imaginative, generating new, unseen content based on input conditions. Common examples of GAI include natural language generation models and image generation models.

[0076] Tool API (Application Programming Interface) integration: Tool API integration involves packaging AI models into application programming interfaces (APIs) and providing them to developers or users. Through the APIs, developers or users can leverage AI model functionality, such as language translation, image recognition, and speech synthesis, thereby enriching applications with richer functionality.

[0077] General artificial intelligence: General artificial intelligence refers to an artificial intelligence system that has the same broad intelligence capabilities as humans, can continuously learn and adapt to various environments and tasks, and has the potential to surpass current human intelligence.

[0078] Convolutional Neural Networks (CNN) model: A multi-layer deep learning model with forward propagation and backpropagation, and a convolution filter that processes feature data.

[0079] Recurrent Neural Network (RNN) model: A recursive deep learning model that processes vector representations recursively and connects intermediate layers in a chain-like manner.

[0080] Long Short Term Memory (LSTM) model: A deep learning model that has the ability to memorize both long-term and short-term information and has a convolutional filter that processes feature data.

[0081] Deep Self-Attention Model (Transformer Model): A deep learning architecture based on the attention mechanism for processing sequential data such as natural language.

[0082] Bidirectional Encoder Representations from Transformers (BERT) is a special Transformer model trained using a bidirectional Transformer encoder and large-scale unlabeled text data.

[0083] With the advancement of computer technology, building a general artificial intelligence system capable of multimodal conversational interactions like humans has long been a long-term goal for achieving artificial general intelligence. While traditional intelligent systems have achieved high levels of performance in text-based conversations, they still struggle with multimodal interactions such as voice, images, and video. The Model-as-a-Service (Model Scope) sharing platform provides developers with a rich, easy-to-use, and low-cost one-stop model service, making model application easier. This model service includes a rich selection of pre-trained state-of-the-art models, diverse and open datasets, model inference capabilities with one line of code, rapid construction of specialized industry models with ten lines of code, a ready-to-use online development platform, a flexible model framework and deployment methods, and extensive educational content and technical resources. This enables knowledge Q&A, detailed understanding of requirements, and the recommendation of the most suitable models. It also offers innovative experiences, such as simple and complex tasks, and multi-application programming interface (API) planning.

[0084] See Figure 1, which shows a flowchart of a question-answering method. The question-answering method obtains the description of the model in the model library and integrates it into the prompt to establish a connection with the large model. After that, the large model can act as the brain to determine the answer to the user's question. As shown in Figure 1, the workflow of the question-answering method includes four stages: task planning, model selection, task execution, and response generation. The task planning stage can decompose the user input into several subtasks and determine the type and difficulty of each subtask. The model selection stage can search for a suitable model from the model library to perform the subtask. The task execution stage can convert the input into the corresponding format of the selected model and use the selected model to predict the user's input. The response generation stage can convert the model processing results into the format expected by the user.

[0085] However, the above solution requires jumping to the model library to select the model corresponding to each subtask, and its work execution process is a tree structure, resulting in extremely low task processing efficiency.

[0086] In order to solve the above problems, the embodiment of the present disclosure proposes a flexible and customizable multi-agent task processing system with a task planning model as the core, based on a model-as-a-service sharing platform. In this generative artificial intelligence system, the capabilities of the model can be enhanced on the basis of the task planning model with a large model as the base model, and voice, picture, video and other functions can be integrated through model collaboration, model API, plug-in API and other methods. The task processing system is a customizable and comprehensive framework that can provide a customizable engine design, covering functions such as data collection, tool retrieval, tool registration, storage management, customized model training and practical application, which can be used to quickly implement applications in actual scenarios. In the task processing system, a large model can be used as a core component to support model training on multiple large models. The task processing system also includes a diverse and comprehensive API to achieve seamless integration with the model API and common functional API in a unified manner.

[0087] Specifically, an embodiment of the present disclosure proposes a task processing method, which obtains task data of a target task; inputs the task data and object description information of candidate processing objects into a task planning model to determine multiple processing objects corresponding to the target task, wherein the task planning model is used to plan multiple processing objects of the target task; uses multiple processing objects to process the task data to obtain processing results output by multiple processing objects respectively; and determines the task processing result of the target task based on the processing results output by multiple processing objects respectively.

[0088] In the present disclosure, a task processing method is provided. The present disclosure also involves a dialogue task processing method, a task planning model training method, an information processing method based on a task planning model, a model training platform, a task processing device, a dialogue task processing device, a task planning model training device, an information processing device based on a task planning model, a computing device, a computer-readable storage medium and a computer program product, which are described in detail one by one in the following embodiments.

[0089] 2 , which shows an architecture diagram of a task processing system provided by an embodiment of the present disclosure, the task processing system may include a client 100 and a server 200 ;

[0090] The client 100 is used to send task data of the target task to the server 200;

[0091] The server 200 is configured to input task data and object description information of candidate processing objects into a task planning model, determine multiple processing objects corresponding to a target task, wherein the task planning model is configured to plan multiple processing objects for the target task; process the task data using the multiple processing objects to obtain processing results outputted by the multiple processing objects; determine a task processing result for the target task based on the processing results outputted by the multiple processing objects; and send the task processing result to the client 100;

[0092] The client 100 is also used to receive the task processing result sent by the server 200.

[0093] By applying the solution of the embodiment of the present disclosure, a task planning model is used to determine multiple processing objects for processing target tasks, thereby achieving seamless integration between the model and processing objects with different processing capabilities, expanding the task processing capabilities, and making task processing more efficient and comprehensive.

[0094] In an optional embodiment of the present disclosure, the task processing system may include multiple clients 100 and a server 200. Multiple clients 100 may establish communication connections through the server 200. In a task processing scenario, the server 200 is used to provide task processing services between the multiple clients 100. The multiple clients 100 may act as senders or receivers, respectively, and communicate through the server 200.

[0095] Users can interact with the server 200 through the client 100 to receive data sent by other clients 100, or send data to other clients 100. In the task processing scenario, users can publish data streams to the server 200 through the client 100. The server 200 generates task processing results based on the data stream and pushes the task processing results to other clients with which communication has been established.

[0096] The client 100 and the server 200 are connected via a network. The network provides a medium for the communication link between the client 100 and the server 200. The network can include various connection types, such as wired or wireless communication links or fiber optic cables. The data transmitted by the client 100 may need to be encoded, transcoded, compressed, or other processing before being released to the server 200.

[0097] The client 100 can be any one of a browser, an APP (Application), a web application such as an H5 (HyperText Markup Language 5, version 5) application, a light application (also known as a mini-program, a lightweight application), or a cloud application. The client 100 can be based on the software development kit (SDK) of the corresponding service provided by the server 200, such as developed based on a real-time communication (RTC) SDK. The client 100 can be deployed in an electronic device and needs to rely on the device to run or certain APPs in the device to run. For example, the electronic device can have a display screen and support information browsing, such as a personal mobile terminal such as a mobile phone, a tablet computer, a personal computer, etc. Various other types of applications can also be configured in the electronic device, such as human-computer dialogue applications, model training applications, text processing applications, web browser applications, shopping applications, search applications, instant messaging tools, email clients, social platform software, etc.

[0098] The server 200 may include servers that provide various services, such as servers that provide communication services to multiple clients, servers that provide background training to support models used on clients, and servers that process data sent by clients. It should be noted that the server 200 can be implemented as a distributed server cluster consisting of multiple servers, or as a single server. The server can also be a server in a distributed system, or a server that is integrated with a blockchain. The server can also be a cloud server for basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, content delivery networks (CDNs), and big data and artificial intelligence platforms, or an intelligent cloud computing server or intelligent cloud host with artificial intelligence technology.

[0099] It is worth noting that the task processing methods provided in the embodiments of the present disclosure are generally executed by the server. However, in other embodiments of the present disclosure, the client may also have similar functions to the server and thus execute the task processing methods provided in the embodiments of the present disclosure. In other embodiments, the task processing methods provided in the embodiments of the present disclosure may also be executed jointly by the client and the server.

[0100] 3 , which shows a flowchart of a task processing method provided by an embodiment of the present disclosure, specifically comprising the following steps:

[0101] Step 302: Obtain task data of the target task.

[0102] In one or more embodiments of the present disclosure, task data of a target task may be acquired, and the task data may be processed based on the target task to obtain a task processing result of the target task.

[0103] Specifically, the target tasks can be tasks in different scenarios, such as question-answering tasks in intelligent question-answering scenarios, story generation tasks in story generation scenarios, video generation tasks in short film and video production scenarios, and conversation tasks in chat room debate scenarios. Task data can be data in different modalities, such as text data in text modality, image data in image modality, speech data in audio modality, video data in video modality, and so on.

[0104] In practical applications, there are multiple ways to obtain task data for a target task, and the method to be used depends on the actual situation. The embodiments of the present disclosure do not impose any restrictions on this. In one possible implementation of the present disclosure, task data for a target task sent by a front-end user can be received. In another possible implementation of the present disclosure, task data for a target task can be read from other data acquisition devices or databases.

[0105] In an optional embodiment of the present disclosure, before obtaining the task data of the target task, the following steps may be further included:

[0106] Sending a processing prompt message to the front-end user, wherein the processing prompt message is used to guide the front-end user to send task data for processing the target task;

[0107] Receive task data of the target task sent by the front-end user based on the processing prompt information.

[0108] It should be noted that when processing a target task begins, a processing prompt message can be sent to the front-end user. This prompt message allows the front-end user to understand the processing flow or requirements of the target task and enter task data that is more relevant to their actual needs. For example, the processing prompt message could read, "Hello, I'm your intelligent service assistant. I'm ready to handle various tasks for you. You can directly enter the task data for the target task you want to process."

[0109] Furthermore, after sending the processing prompt information to the front-end user, the task data of the target task sent by the front-end user based on the processing prompt information can be received, and the task data and the object description information of the candidate processing objects can be input into the task planning model to determine the multiple processing objects corresponding to the target task.

[0110] Using the solution of the disclosed embodiment, a processing prompt is sent to a front-end user; task data for a target task sent by the front-end user based on the processing prompt is received. The processing prompt guides the front-end user to enter the task data for the target task, thereby enhancing human-computer interaction and ensuring that the task data for the target task accurately reflects the user's actual needs.

[0111] Step 304: Input the task data and the object description information of the candidate processing objects into the task planning model to determine multiple processing objects corresponding to the target task, wherein the task planning model is used to plan the multiple processing objects of the target task.

[0112] In one or more embodiments of the present disclosure, after obtaining the task data of the target task, the task data and object description information of the candidate processing objects may be further input into the task planning model to determine multiple processing objects corresponding to the target task.

[0113] Specifically, the task planning model can be a pre-trained large model, or it can be a deep learning model trained based on multiple sample task data using the large model as the foundation, wherein the sample task data carries sample description information and object planning results of the sample processing object. Task planning models include but are not limited to CNN models, RNN models, LSTM models, and BERT models. The processing object is used to process task data, and the processing object can be at least one of a task processing plug-in and a task processing model. Task processing models include Wensheng graph models, Wensheng audio models, graph dialogue models, translation models, and the like. The task processing model can be provided to front-end users using the Tool API integration method. Task processing plug-ins can be understood as task processing tools, such as weather tools, map tools, and the like. Object description information is used to describe the attribute information of the processing object. If the processing object is a task processing plug-in, the object description information includes but is not limited to the plug-in API name, plug-in parameters, and plug-in request function. If the processing object is a task processing model, the object description information includes but is not limited to the model identifier, model storage address, and model parameter information.

[0114] In an optional embodiment of the present disclosure, candidate processing objects may be determined before inputting task data and object description information of candidate processing objects into the task planning model. In practical applications, there are multiple ways to determine candidate processing objects, and the specific method to be selected depends on the actual situation. This embodiment of the present disclosure does not impose any restrictions on this. In one possible implementation of the present disclosure, all original processing objects in the object library can be determined as candidate processing objects.

[0115] In another possible implementation of the present disclosure, since the object library may include a large number of original processing objects, in order to reduce the amount of data calculation of the task planning model, candidate processing objects may be determined based on the task data and the original object description information of the original processing objects. That is, before inputting the task data and the object description information of the candidate processing objects into the task planning model and determining the multiple processing objects corresponding to the target task, the following steps may be further included:

[0116] Obtaining original object description information of multiple original processing objects;

[0117] According to the task data and the original object description information, candidate processing objects are retrieved from multiple original processing objects.

[0118] It should be noted that there are multiple ways to obtain the original object description information of multiple original processing objects, and the specific method to be selected depends on the actual situation. The embodiments of this disclosure do not impose any restrictions on this. In one possible implementation of this disclosure, the original object description information of multiple original processing objects can be received from a front-end user. In another possible implementation of this disclosure, the original object description information of multiple original processing objects can be read from other data acquisition devices or databases.

[0119] In actual applications, there are many ways to retrieve candidate processing objects from multiple original processing objects based on task data and original object description information. The specific selection is made according to the actual situation, and the embodiments of the present disclosure do not impose any restrictions on this. In one possible implementation method of the present disclosure, the similarity between the original object description information and the task description information of each original processing object can be calculated, so as to determine the degree of match between each original processing object and the target task based on the similarity (the higher the similarity, the higher the degree of match), and select a preset number of original processing objects with a higher degree of match as candidate processing objects, wherein the method of calculating the similarity includes but is not limited to cosine similarity and edit distance. In another possible implementation method of the specification, the task data and the original object description information can be respectively extracted for keywords, and the degree of match between each original processing object and the target task is determined based on the keywords in the task data and the keywords in the original object description information (the more the number of identical keywords, the higher the degree of match), and select a preset number of original processing objects with a higher degree of match as candidate processing objects.

[0120] Using the solution of the disclosed embodiment, original object description information for multiple original processing objects is obtained; based on the task data and the original object description information, candidate processing objects are retrieved from the multiple original processing objects. By retrieving candidate processing objects from the multiple original processing objects, the data computational complexity of the task planning model is reduced, and the efficiency of determining processing objects is improved.

[0121] In an optional embodiment of the present disclosure, before inputting the task data and the object description information of the candidate processing objects into the task planning model and determining the multiple processing objects corresponding to the target task, the following steps may be further included:

[0122] Obtain reference data;

[0123] According to the task data, target reference data is retrieved from the reference data;

[0124] Inputting the task data and the object description information of the candidate processing objects into the task planning model to determine the multiple processing objects corresponding to the target task may include the following steps:

[0125] Model prompt information, task data, object description information of candidate processing objects and target reference data are input into the task planning model to obtain multiple processing objects corresponding to the target task.

[0126] Specifically, reference data is used to enhance the knowledge of the task planning model. Reference data includes, but is not limited to, questions and answers related to the target task, prior knowledge, and historical processing data. Target reference data refers to data within the reference data that is associated with the target task. Model prompts guide the task planning model in determining processing objects, such as "You are a professional task planner. Please identify multiple processing objects corresponding to the target task from candidate processing objects based on the task data and object description information."

[0127] It should be noted that there are multiple ways to obtain reference data, and the specific method to be used depends on the actual situation. The embodiments of this disclosure do not impose any restrictions on this. In one possible implementation of this disclosure, reference data sent by a front-end user can be received. In another possible implementation of this disclosure, reference data can be read from other data acquisition devices or databases.

[0128] In actual applications, the implementation method of "retrieving target reference data from reference data based on task data" is the same as the above-mentioned implementation method of "retrieving candidate processing objects from multiple original processing objects based on task data and original object description information", and will not be repeated in this disclosed embodiment.

[0129] Furthermore, when inputting the model prompt information, task data, object description information of candidate processing objects and target reference data into the task planning model, the model prompt information, task data, object description information of candidate processing objects and target reference data can be directly input into the task planning model, or the model prompt information, task data, object description information of candidate processing objects and target reference data can be assembled to obtain the model data to be processed, and the model data to be processed can be input into the task planning model to obtain multiple processing objects corresponding to the target task.

[0130] Applying the solution of the disclosed embodiment, reference data is obtained; target reference data is retrieved from the reference data based on the task data; model prompt information, task data, object description information of candidate processing objects, and target reference data are input into the task planning model to obtain multiple processing objects corresponding to the target task. By using model prompt information to guide the task planning model and using target reference data to enhance the knowledge of the task planning model's processing process, the accuracy of the processing objects is improved.

[0131] Step 306: Utilize multiple processing objects to process the task data, and obtain processing results outputted by the multiple processing objects respectively.

[0132] In one or more embodiments of the present disclosure, task data of a target task is obtained; the task data and object description information of candidate processing objects are input into a task planning model, and after multiple processing objects corresponding to the target task are determined, the task data can be further processed using the multiple processing objects to obtain processing results output by the multiple processing objects respectively.

[0133] In an optional embodiment of the present disclosure, the above-mentioned processing of task data using multiple processing objects to obtain processing results outputted by the multiple processing objects may include the following steps:

[0134] Calling multiple task processing plug-ins to process task data and obtaining processing results respectively fed back by the multiple task processing plug-ins, wherein the processing objects include the task processing plug-ins; and / or,

[0135] Task data is input into a plurality of task processing models to obtain processing results respectively output by the plurality of task processing models, wherein the processing object includes the task processing model.

[0136] Specifically, the task processing plug-in can be a processing plug-in integrated into the task processing system, or it can be a third-party processing plug-in. The task processing model is used to process task data. The task processing model can be a pre-trained large model, or it can be a deep learning model trained based on a large model and the sample processing results corresponding to each sample task data. Task processing models include but are not limited to CNN models, RNN models, LSTM models, and BERT models. When calling multiple task processing plug-ins to process task data, the plug-in function can be called through the API of the task processing plug-in, and after the task processing plug-in completes the processing, the processing results fed back by the task processing plug-in are received.

[0137] By applying the solution of the embodiment of the present disclosure, multiple task processing plug-ins are called to process task data, and processing results respectively fed back by the multiple task processing plug-ins are obtained, wherein the processing object includes the task processing plug-in; and / or, task data is input into multiple task processing models, and processing results respectively output by the multiple task processing models are obtained, wherein the processing object includes the task processing model. By using the task processing plug-in to process task data, or using the task processing model to process task data, or using the task processing plug-in in combination with the task processing model to process the task processing model, seamless integration between the model and processing plug-ins with different processing capabilities is achieved, the task processing capability is expanded, and task processing is made more efficient and comprehensive.

[0138] Referring to FIG4 , FIG4 shows a flowchart of another task processing method provided by an embodiment of the present disclosure, which specifically includes:

[0139] Obtain task data of the target task, plan the target task using a task planning model (planner), determine multiple processing objects (callers) corresponding to the target task, wherein the multiple processing objects include a task processing model and / or a task processing tool, which is responsible for calling and executing the model API or tool API, and using the multiple processing objects to process the task data, obtain the processing results output by the multiple processing objects respectively, and feed the processing results output by the multiple processing objects respectively back to the task planning model, and the task planning model determines whether the target task is completed: if the target task is not completed, continue to call the multiple processing objects in a loop to perform task processing, that is, return to the step of processing the task data using the multiple processing objects and obtaining the processing results output by the multiple processing objects respectively; if the target task is completed, input the processing results output by the multiple processing objects respectively into the result generation model (summarizer) to obtain the task processing result of the target task, wherein the result generation model is used to summarize the processing results output by the multiple processing objects respectively.

[0140] By applying the solution of the embodiments of the present disclosure, multiple models and / or tools can be coordinated to allow each model and / or tool to have a clear division of labor, which can greatly improve the task processing effect.

[0141] In an optional embodiment of the present disclosure, the above-mentioned processing of task data using multiple processing objects to obtain processing results outputted by the multiple processing objects may include the following steps:

[0142] determining a current processing object from a plurality of processing objects;

[0143] Use the current processing object to process the task data and obtain the current processing result output by the current processing object;

[0144] The task planning model is used to analyze the current processing results. When it is determined that the target task has not been completed, the step of determining the current processing object from multiple processing objects is returned to execute until the target task is completed, and the processing results output by multiple processing objects are obtained.

[0145] Specifically, the current processing object refers to a processing object that has not processed the task data. Assuming that the multiple processing objects include a first processing object and a second processing object, and the first processing object is a completed processing object, the current processing object is the second processing object.

[0146] It should be noted that the implementation method of "using the current processing object to process the task data and obtain the current processing result output by the current processing object" is the same as the above-mentioned "calling multiple task processing plug-ins to process the task data and obtain the processing results respectively fed back by multiple task processing plug-ins, wherein the processing object includes the task processing plug-in; and / or, inputting the task data into multiple task processing models, obtaining the processing results respectively output by multiple task processing models, wherein the processing object includes the task processing model", and this disclosure will not elaborate on it.

[0147] In practical applications, after processing task data using the current processing object and obtaining the current processing result output by the current processing object, the task planning model can be used to analyze the current processing result. If the current processing result can solve the target task, it is determined that the target task has been processed. The task planning model then proceeds to determine the task processing result of the target task based on the processing results output by the multiple processing objects. If the current processing result cannot solve the target task, it is determined that the target task has not been processed. In this case, the process returns to the step of determining the current processing object from the multiple processing objects until the target task is processed and the processing results output by the multiple processing objects are obtained.

[0148] Using the solution of the disclosed embodiment, a current processing object is determined from among multiple processing objects; task data is processed using the current processing object to obtain the current processing result output by the current processing object; the current processing result is analyzed using a task planning model. If it is determined that the target task has not been processed, the process returns to the step of determining the current processing object from among multiple processing objects until the target task is processed and the processing results output by each of the multiple processing objects are obtained. By determining the current processing object that has not processed the task data, the same processing object is avoided from being reused for repeated processing, thus reducing resource waste.

[0149] In a possible implementation of the present disclosure, before the above-mentioned processing of the task data using the current processing object to obtain the current processing result output by the current processing object, the following steps may also be included:

[0150] Splitting the target task into multiple target subtasks, and determining subtask data of the multiple target subtasks based on the task data, wherein the target subtasks correspond to the processing objects one by one;

[0151] Using the current processing object to process the task data to obtain the current processing result output by the current processing object may include the following steps:

[0152] The current subtask data is processed using the current processing object to obtain the current processing result output by the current processing object, wherein the current subtask data is the subtask data of the target subtask corresponding to the current processing object.

[0153] It should be noted that there are many ways to split the target task into multiple target subtasks and determine the subtask data of multiple target subtasks based on the task data. The specific selection depends on the actual situation. The embodiment of the present disclosure does not impose any restrictions on this. In one possible implementation of the present disclosure, the work breakdown structure (WBS) can be used to gradually subdivide the target task from top to bottom according to the logical relationship to form a multi-level tree structure, so as to determine multiple target subtasks and subtask data of multiple target subtasks based on the tree structure. In another possible implementation of the present disclosure, the target task and the task data of the target task can be input into the task splitting model to obtain the subtask data of multiple target subtasks and multiple target subtasks, wherein the task splitting model is trained based on multiple sample tasks (carrying sample task data) and sample subtasks (carrying sample subtask data) corresponding to multiple sample subtasks.

[0154] For example, assuming that the target task is "reading the task text data aloud with a female voice, and matching the task text data with a video", the target task is split into two target subtasks, subtask data 1 of target subtask 1 is "reading the task text data aloud with a female voice", and subtask data 2 of target subtask 2 is "matching the task text data with a video". Assuming that the current processing object is a speech generation model, the task text data is input into the speech generation model to obtain the female voice audio output by the speech generation model.

[0155] Applying the solution of the disclosed embodiment, a target task is split into multiple target subtasks. Subtask data for each target subtask is determined based on the task data, where each target subtask corresponds to a processing object. The current subtask data is then processed using the current processing object to obtain the current processing result output by the current processing object. The current subtask data is the subtask data for the target subtask corresponding to the current processing object. By splitting tasks, the processing capacity of the processing object is accurately utilized to process the target task in parallel, improving the flexibility of task processing.

[0156] In another possible implementation of the present disclosure, before the above-mentioned processing of the task data by the current processing object to obtain the current processing result output by the current processing object, the following steps may be further included:

[0157] Determining a completed processing object from a plurality of processing objects, and obtaining a completed processing result output by the completed processing object, wherein the completed processing object is a processing object that has completed processing of the task data;

[0158] Using the current processing object to process the task data to obtain the current processing result output by the current processing object may include the following steps:

[0159] The completed processing result is processed using the current processing object to obtain the current processing result output by the current processing object.

[0160] It should be noted that when using multiple processing objects to process task data, the task data can be processed using the processing objects separately, and the completed processing results output by the completed processing objects can also be used as the input of the unprocessed objects. The specific selection should be made according to the actual situation, and this disclosure does not impose any restrictions on this.

[0161] For example, assuming that the task data of the target task is "write a short story, read it aloud with a female voice, and add a video at the same time", the target task is divided into 3 target subtasks. Subtask data 1 of target subtask 1 is "generate a story", subtask data 2 of target subtask 2 is "generate voice and read it out with a female voice, and show it to the user", and subtask data 3 of target subtask 3 is "generate a video for the generated story content". During the processing, a story is first generated, and then the corresponding voice generation model is called to process the generated story, obtain voice and read it out with a female voice, and show it to the user. Finally, the video generation model is called to generate a video for the generated story content. In this process, the user does not need to configure the tools that may need to be called for the current request, which greatly improves the ease of use.

[0162] Using the solution of the disclosed embodiments, a completed processing object is identified from multiple processing objects, and the completed processing result output by the completed processing object is obtained. The completed processing object is a processing object that has completed processing task data. The completed processing result is then processed using the current processing object to obtain the current processing result output by the current processing object. By using the completed processing result output by the completed processing object as the input of the current processing object, serial processing of the target task is achieved, improving the flexibility of task processing.

[0163] Step 308: Determine the task processing result of the target task according to the processing results output by the multiple processing objects.

[0164] In one or more embodiments of the present disclosure, task data of a target task is obtained; the task data and object description information of candidate processing objects are input into a task planning model to determine multiple processing objects corresponding to the target task; the task data is processed using multiple processing objects to obtain processing results respectively output by the multiple processing objects, and then, further, the task processing result of the target task can be determined based on the processing results respectively output by the multiple processing objects.

[0165] Specifically, the task processing result of the target task is related to the target task. If the target task is a question-answering task, the task processing result is the answer result; if the target task is a story generation task, the task processing result is the target story.

[0166] By applying the solution of the embodiment of the present disclosure, a task planning model is used to determine multiple processing objects for processing target tasks, thereby achieving seamless integration between the model and processing objects with different processing capabilities, expanding the task processing capabilities, and making task processing more efficient and comprehensive.

[0167] In practical applications, there are multiple ways to determine the task processing result of the target task based on the processing results output by multiple processing objects. The specific method is selected according to the actual situation, and the embodiments of this disclosure do not impose any restrictions on this. In one possible implementation of this disclosure, the processing results output by multiple processing objects can be directly spliced ​​to obtain the task processing result.

[0168] In another possible implementation of the present disclosure, a result generation model may be used to process the processing results respectively output by multiple processing objects to obtain a task processing result of a target task. That is, determining the task processing result of a target task based on the processing results respectively output by multiple processing objects may include the following steps:

[0169] The processing results respectively output by the multiple processing objects are input into the result generation model to obtain the task processing result of the target task, wherein the result generation model is used to summarize the processing results respectively output by the multiple processing objects.

[0170] Specifically, the result generation model can not only aggregate the processing results output by multiple processing objects but also perform deduplication and other processing. The result generation model can be a pre-trained large model or a deep learning model trained on multiple sample result sets and the sample processing results corresponding to each sample result set, using the large model as a foundation. The sample structure set includes multiple sample results. Result generation models include but are not limited to CNN models, RNN models, LSTM models, and BERT models.

[0171] By applying the method of the disclosed embodiment, the processing results output by multiple processing objects are input into the result generation model to obtain the task processing result of the target task. The result generation model aggregates the processing results output by multiple processing objects, thereby improving the quality of the task processing result.

[0172] In an optional embodiment of the present disclosure, after determining the task processing result of the target task based on the processing results respectively output by the multiple processing objects, the following steps may be further included:

[0173] Send task processing results to front-end users;

[0174] Receive result feedback information sent by the front-end user, wherein the result feedback information is information that provides feedback on the task processing result based on the task information of the target task;

[0175] Based on the feedback information from the results, build a model to optimize the data;

[0176] Use model optimization data to adjust the parameters of the mission planning model.

[0177] Specifically, result feedback information can include feedback on the content, quality, and completion of task processing results, reflecting the front-end user's true feelings and expectations about the task processing results. Result feedback information includes, but is not limited to, result quality evaluation information, corrected and accurate task processing results, and model optimization areas. Model optimization data refers to accurately optimized sample data used to optimize the task planning model.

[0178] In practical applications, there are multiple ways to construct model optimization data based on the result feedback information, and the specific method to be selected depends on the actual situation. The embodiments of the present disclosure do not impose any restrictions on this. In one possible implementation of the present disclosure, the model optimization data can be directly obtained from other data acquisition devices or databases based on the result feedback information. In another possible implementation of the present disclosure, optimization prompt information can be generated based on the result feedback information, and model optimization data sent by the front-end user based on the optimization prompt information can be received.

[0179] It should be noted that, taking the construction of model optimization data directly based on the result feedback information as an example, if the result feedback information is an accurate and corrected task processing result, the model optimization data can be constructed based on the task data of the target task and the corrected accurate task processing result. If the result feedback information is the optimization field of the model, such as the XXX field, the sample task data of the XXX field can be obtained, and the sample task data of the XXX field can be determined as the model optimization data. Among them, the process of adjusting the parameters of the task planning model using the model optimization data is the same as the training process of the task planning model, and the embodiments of the present disclosure will not be repeated. At the same time, the parameters of other models involved in the task processing process (such as the task processing model and the result generation model) can also be adjusted.

[0180] Using the solution of the disclosed embodiments, task processing results are sent to front-end users; feedback information is received from the front-end users, where the feedback information provides feedback on the task processing results based on the target task information; model optimization data is constructed based on the feedback information; and the model optimization data is used to adjust the parameters of the task planning model. By collecting and utilizing the feedback information, the performance of the task planning model is continuously optimized to more accurately meet the actual needs of the front-end users and improve the quality and accuracy of the final task processing results.

[0181] In an optional embodiment of the present disclosure, constructing the model optimization data based on the result feedback information may include the following steps:

[0182] Generate optimization prompt information based on the result feedback information, wherein the optimization prompt information is used to guide the front-end user to send model optimization data for optimizing the task planning model;

[0183] Send optimization prompt information to front-end users, and receive model optimization data sent by front-end users based on the optimization prompt information.

[0184] It should be noted that there are multiple ways to generate optimization prompt information based on the result feedback information, and the specific selection should be made according to the actual situation. The embodiments of the present disclosure do not impose any limitations on this.

[0185] In one possible implementation of the present disclosure, preset prompt information can be directly obtained, and the result feedback information can be added to the preset prompt information to obtain optimized prompt information. For example, the preset prompt information is "I am very sorry for providing you with inaccurate information. Please point out the specific inaccuracies or provide the correct answers to relevant questions. I will correct and optimize my answers as soon as possible to better serve you." If the result feedback information is "the result is inaccurate," the optimized prompt information is "Regarding your feedback on the inaccurate results, I am very sorry for providing you with inaccurate information. Please point out the specific inaccuracies or provide the correct answers to relevant questions. I will correct and optimize my answers as soon as possible to better serve you."

[0186] In another possible implementation of the present disclosure, the result feedback information may be type-identified to determine the information type of the result feedback information, and the information type may be further matched with the prompt type of each prompt information in the prompt information library, and the prompt information having the same prompt type as the information type may be determined as the optimized prompt information.

[0187] Applying the solution of the disclosed embodiment, optimization prompt information is generated based on the result feedback information. The optimization prompt information is used to guide the front-end user to send model optimization data for optimizing the task planning model. The optimization prompt information is sent to the front-end user, and the model optimization data sent by the front-end user based on the optimization prompt information is received. Obtaining model optimization data through interactive guidance improves interactivity with users and increases user satisfaction.

[0188] In an optional embodiment of the present disclosure, after determining the task processing result of the target task based on the processing results respectively output by the multiple processing objects, the following steps may be further included:

[0189] Mark the key information in the task processing results to obtain the updated task processing results;

[0190] Send the updated task processing results to the front-end user.

[0191] Specifically, key information is related to the modality of the task processing result. If the task processing result is text, the key information is key text information such as keywords, key sentences, and keywords. If the task processing result is an image, the key information is visual key points (such as eyes, eyebrows, corners of the mouth, etc. in the image). Taking the key information as key text information as an example, key text information refers to facts, data, concepts, sentences or words that can highlight the theme of the text, convey the core idea, support the author's point of view, or describe the essence of the event. Key information helps to quickly understand the main idea of ​​the text, grasp the structure of the article, answer related questions, and make effective decisions or extract information.

[0192] In actual applications, the key information in the task processing results is marked, and before obtaining the updated task processing results, the key information of the task processing results can be identified to determine the key information in the task processing results. There are many ways to identify the key information of the task processing results, and the specific selection is made according to the actual situation. The embodiments of the present disclosure do not impose any restrictions on this. In one possible implementation of the present disclosure, the task processing results can be matched with preset key information, and the information that appears in the task processing results and is the same as the preset key information is determined as key information. In another possible implementation of the present disclosure, the task processing results can be input into a key information identification model to obtain key information, wherein the key information identification model is trained based on multiple training data and key information labels of each training data.

[0193] It should be noted that when marking key information in the task processing results, the key information can be bolded, highlighted, tilted, bordered, etc. The marking method of the key information is selected according to the actual situation, and the embodiment of the present disclosure does not impose any restrictions on this.

[0194] By applying the solution of the embodiment of the present disclosure, key information in the task processing results is marked to obtain updated task processing results; the updated task processing results are sent to front-end users so that they can view the results, thereby improving user experience.

[0195] In an optional embodiment of the present disclosure, after determining the task processing result of the target task based on the processing results respectively output by the multiple processing objects, the following steps may be further included:

[0196] Send task processing results to front-end users;

[0197] Receive editing information sent by the front-end user, where the editing information is used to edit the task processing results;

[0198] The task processing result is edited according to the editing information to obtain the edited task processing result.

[0199] Specifically, editing information is used to describe the front-end user's editing requirements for the task processing results. Editing requirements include, but are not limited to, translation requirements, synonym replacement requirements, and interpretation requirements. Furthermore, editing information can also be editing requirements for key information in the task processing results.

[0200] For example, assuming the key information is "diligent and conscientious" and the editing information is "Please explain the meaning of the key information", according to the editing information, add the explanation "Diligent and conscientious is used to describe a person's attitude towards work or tasks, always maintaining vigilance and awe. It is often used to praise people's meticulous and conscientious spirit at work" near the key information in the target result.

[0201] Using the solution of the disclosed embodiment, the task processing result is sent to the front-end user; editing information sent by the front-end user is received, wherein the editing information is used to edit the task processing result; and the task processing result is edited based on the editing information to obtain the edited task processing result. By editing the task processing result based on the editing information sent by the front-end user, human-computer interaction is enhanced, and the adaptability and flexibility of task processing are improved.

[0202] In an optional embodiment of the present disclosure, a training method for a task planning model is described. That is, before inputting the task data and the object description information of the candidate processing objects into the task planning model and determining the multiple processing objects corresponding to the target task, the following steps may be further included:

[0203] Acquire a plurality of sample task data, wherein the sample task data carries sample description information and object planning results of a sample processing object;

[0204] Inputting multiple sample task data and sample description information into the task planning model to obtain object prediction results corresponding to the multiple sample task data;

[0205] According to the object planning results and object prediction results, the parameters of the task planning model are adjusted to obtain the trained task planning model.

[0206] Specifically, the task planning model is trained using supervised training. This means that sample task data carries real-world processing labels (object planning results). These object planning results serve as the generation targets for the task planning model, guiding its training. Using sample task data and sample descriptions of sample processing objects, the model predicts the processing objects and generates object prediction results. The task planning model can be a pre-trained large model or an untrained natural language neural network model.

[0207] Sample task data includes, but is not limited to, model API data, API-oriented question-and-answer data, and API-independent data. API-oriented question-and-answer data can be used to provide users with appropriate recommendations on how to use various APIs to solve complex tasks. API-independent data can enhance the ability to follow general instructions and increase user engagement, emphasizing the inherent functionality of the system rather than relying on API calls.

[0208] It should be noted that there are multiple ways to obtain multiple sample task data, and the specific method to be selected depends on the actual situation. The embodiments of this disclosure do not impose any restrictions on this. In one possible implementation of this disclosure, a large amount of sample task data can be read from other data acquisition devices or databases. In another possible implementation of this disclosure, multiple sample task data sent by a user can be received.

[0209] In actual applications, when adjusting the parameters of the task planning model according to the object planning results and the object prediction results, the loss value can be calculated according to the object planning results and the object prediction results, and the model parameters of the task planning model can be adjusted according to the loss value until the training process meets the preset stop conditions, thereby obtaining a trained task planning model. Among them, there are many functions for calculating the loss value, such as the cross entropy loss function, the L1 norm loss function, the maximum loss function, the mean square error loss function, the logarithmic loss function, etc., which are selected according to the actual situation, and the embodiments of the present disclosure do not impose any restrictions on this. The preset stopping conditions include but are not limited to the loss value being less than or equal to the preset threshold and the number of iterations reaching the preset number of iterations, wherein the preset threshold and the preset number of iterations are selected according to the actual situation, and the embodiments of the present disclosure do not impose any restrictions on this.

[0210] In one possible implementation of the present disclosure, after calculating the loss value, the loss value is compared with a preset threshold. Specifically, if the loss value is greater than the preset threshold, it indicates that the difference between the object planning result and the object prediction result is large, and the task planning model has poor prediction ability for the processing object. In this case, the model parameters of the task planning model can be adjusted, and the step of inputting multiple sample task data and sample description information into the task planning model to obtain the object prediction results corresponding to the multiple sample task data is returned. The task planning model is then trained until the loss value is less than or equal to the preset threshold, indicating that the difference between the object planning result and the object prediction result is small, and the preset stop condition is met, thereby obtaining a task planning model that has completed training.

[0211] In another possible implementation of the present disclosure, in addition to comparing the loss value with a preset threshold, the number of iterations may also be used to determine whether the current task planning model has been trained. Specifically, if the loss value is greater than the preset threshold, the model parameters of the task planning model are adjusted, and the process returns to the step of inputting multiple sample task data and sample description information into the task planning model to obtain object prediction results corresponding to the multiple sample task data. The task planning model is then trained until the preset number of iterations is reached, at which point iterations are stopped to obtain a fully trained task planning model.

[0212] Using the solution of the disclosed embodiments, a task planning model is trained based on the object planning results and object prediction results. If a preset stopping condition is not met, the task planning model continues to be trained until the preset stopping condition is met, completing the training and obtaining the task planning model. By continuously adjusting the model parameters of the task planning model, the resulting task planning model can be made more accurate.

[0213] In an optional embodiment of the present disclosure, the first weight of the sample task data and the second weight of the sample description information are different; the inputting of the plurality of sample task data and the sample description information into the task planning model to obtain the object prediction results corresponding to the plurality of sample task data may include the following steps:

[0214] A plurality of sample task data, sample description information, a first weight and a second weight are input into a task planning model to obtain object prediction results corresponding to the plurality of sample task data respectively.

[0215] It should be noted that during the training of the task planning model, it is also possible to weight the loss of some tokens to improve the task planning model's ability to process object calls. In other words, different weights can be assigned to sample task data and sample description information. Furthermore, different weights can be assigned to model prompt information and target reference data to achieve loss weighting.

[0216] For example, for model prompt information, the loss weight can be set to 0 so that the model does not calculate the loss of the prompt. For data not related to the API, the loss weight can be set to 1. Finally, for important text of the API call, such as the API name, parameters, uniform resource locator, etc., the loss weight can be set to 2, which can improve the generation accuracy of the API call. See Table 1 below, which is an accuracy comparison table, which shows that the accuracy of the task planning model proposed in the embodiment of the present disclosure is higher than that of the traditional model:

[0217] Table 1 Accuracy comparison table

[0218] By applying the solution of the embodiment of the present disclosure, multiple sample task data, sample description information, first weight and second weight are input into the task planning model to obtain object prediction results corresponding to the multiple sample task data, thereby improving the accuracy of the task planning model.

[0219] Referring to FIG5 , FIG5 shows a flowchart of a task processing method provided by an embodiment of the present disclosure. The task processing process can be divided into two parts: model training and task execution:

[0220] Model training: During model training, data is first collected, including sample task data including model API data, API-oriented question-and-answer data, and API-independent data. The sample task data is then used for model training. Finally, the trained task planning model is evaluated using both automatic and manual evaluation methods. Automatic evaluation can calculate the Action Exact Match score, ROUGE-L (Recall-Oriented Understudy for Gisting Evaluation-Longest Common Subsequence) score, F1 score, and other metrics.

[0221] Task execution: After model training is completed, the trained task planning model is deployed and used to execute tasks. The execution process includes processing object retrieval (determining multiple processing objects), memory control (determining target reference data and generating model input data), task splitting (splitting the target task into multiple target subtasks), plug-in use (calling multiple processing plug-ins to process the target task), API execution (task processing model API processes task data) and response generation (determining the task processing results of the target task based on the processing results output by multiple processing objects).

[0222] It should be noted that the task processing method can be applied to the task processing system, which is a general and customizable intelligent agent framework that can support single-agent (including a single large model) and multi-agent (including multiple large models, such as open chat and multi-model collaboration) applications. The task processing system is based on the task planning model as its core and includes a memory control unit, a tool usage unit, an output unit, and so on. The task planning model is mainly responsible for processing user input and deciding to call the appropriate tool; the memory control unit mainly includes knowledge retrieval and prompt information management; the tool usage unit includes a tool library, tool retrieval, and tool customization; the output unit is used to parse the response of the task planning model into the tool to be called and the corresponding parameters.

[0223] In practical applications, the task processing system proposed in the embodiments of the present disclosure can integrate a distributed framework to better improve the efficiency of multi-agent collaboration and can also support user-defined configurations. For example, for user privacy-related configurations such as user_token, which should not be made public, the task processing system can recommend that users use configuration files (such as the dotenv package and .env files) to set these configurations.

[0224] Specifically, the task processing system provides a template file (.env.template). Users can copy and change the file name to manage their personal configurations, and load these configurations through the load function (load_dotenv(find_dotenv())). Alternatively, users can directly configure tokens by setting environment variables. In addition, the task processing system also provides a model configuration file template (cfg_model_template.json) and a tool configuration file template (cfg_tool_template.json). The task processing system has already filled in the default configuration, which users can use directly or copy and change the file name for in-depth customization.

[0225] In practical applications, the task processing method proposed in the embodiments of the present disclosure can be applied to different target tasks, such as single-step tool call tasks (the agent system selects the correct processing object and generates a request), multi-step tool call tasks (the agent system plans, schedules, executes and responds to multiple processing objects), multi-round dialogue tasks (the agent system mines parameters passed to the processing object from historical dialogues), knowledge question-answering tasks based on retrieval tools, and so on.

[0226] The task processing method provided by the embodiment of the present disclosure can be applied to different task scenarios, such as intelligent question-and-answer scenarios, story generation scenarios, movie and short video production scenarios, chat room debate scenarios, etc. The following, in conjunction with Figure 6, takes the application of the task processing method provided by the present disclosure in a chat room debate as an example to further illustrate the task processing method. Figure 6 shows a flowchart of a dialogue task processing method provided by an embodiment of the present disclosure, which specifically includes the following steps:

[0227] Step 602: Obtain task data of the target dialogue task, wherein the task data includes multiple dialogue roles.

[0228] Step 604: Input the task data and the object description information of the candidate processing objects into the task planning model to determine multiple processing objects corresponding to the target dialogue task, wherein the task planning model is used to plan multiple processing objects of the target dialogue task, and the processing objects correspond one-to-one to the dialogue roles.

[0229] Step 606: Utilize multiple processing objects to process the task data, and obtain processing results outputted by the multiple processing objects respectively.

[0230] Step 608: Determine the task processing result of the target dialogue task according to the processing results outputted by the multiple processing objects.

[0231] It should be noted that the implementation of steps 602 to 608 is the same as the implementation of steps 302 to 308, and will not be described in detail in this embodiment of the present disclosure.

[0232] In practical applications, in a chatroom debate scenario, multiple debate topics, along with corresponding role information and perspectives, can be provided. Different processing objects can be assigned different dialogue roles. The front-end user can also be assigned a dialogue role, enabling debates between the user and the model. Multiple dialogue roles engage in heated discussions around the debate topic. The goal of the debate is persuasion, meaning that through various debates, the dialogue player convinces the other party. The challenge is considered a success when the other party agrees with their perspective.

[0233] By applying the solution of the embodiment of the present disclosure, a task planning model is used to determine multiple processing objects for processing the target dialogue task, thereby achieving seamless integration between the model and processing objects with different processing capabilities, expanding the dialogue task processing capabilities, and making dialogue task processing more efficient and comprehensive.

[0234] In an optional embodiment of the present disclosure, after determining the task processing result of the target dialogue task based on the processing results respectively output by the multiple processing objects, the following steps may also be included:

[0235] Send task processing results to front-end users;

[0236] Receive result feedback information sent by the front-end user, wherein the result feedback information is information that provides feedback on the task processing result based on the task information of the target dialogue task;

[0237] The result feedback information is sent to the model training platform, where the model training platform is used to adjust the parameters of the task planning model using the result feedback information.

[0238] Specifically, the model training platform can be deployed on a terminal device or on a cloud device, and the specific settings are based on actual conditions. The embodiments of the present disclosure do not impose any restrictions on this. The model training platform provides large-scale data processing and high-performance computing resources for training, optimizing and deploying various machine learning models, especially deep learning models. On the model training platform, users can upload data, select or customize algorithm models, and efficiently train and verify models through distributed computing capabilities. The model training platform can receive task generation requests from terminal devices, obtain corresponding task planning models based on the request information, and generate task information based on the task planning model. The model training platform can also quickly respond to the needs of different tasks, call appropriate task planning models for task processing, and ultimately generate high-quality task processing results.

[0239] In actual applications, after receiving the result feedback information sent by the front-end user, the result feedback information can be sent to the model training platform. The model training platform can then adjust the parameters of the task planning model based on the result feedback information. The method for the model training platform to adjust the parameters of the task planning model based on the result feedback information can refer to the above-mentioned implementation method of "constructing model optimization data based on the result feedback information; using the model optimization data to adjust the parameters of the task planning model", and will not be further described in this embodiment of the disclosure.

[0240] By applying the solution of the embodiment of the present disclosure, the parameters of the task planning model are adjusted based on the result feedback information through the model training platform, which can reduce the model training cost while ensuring the accuracy and training efficiency of the task planning model, and provide users with convenient and efficient model training services.

[0241] Refer to Figure 7, which shows an interface diagram of a dialogue task processing interface provided by an embodiment of the present disclosure. As shown in Figure 7, the task data of the target dialogue task includes the role information of three roles (role A, role B, role C). The role information of role A is: a fan of Team X, who insists on the view that a's historical status is second to none; the role information of role B is: a fan of Team Y, who insists on the view that a has been defeated by Team Y many times, has many runner-ups in his career, and his historical status is at most in the top five; the role information of role C is: a supporter of c, who insists on the view that there is still a big gap between a and c. Different processing objects are used as the three roles to conduct dialogues, as shown in Figure 3, and the dialogue content includes:

[0242] Role A: a has won the championship four times and is ranked first in total points. His historical honors are enough to rank second and hope to be first.

[0243] Character B: No, they were still defeated by Team Y many times, second in history, with seven runner-up titles.

[0244] Role C: C has never lost in the finals, and C has never been the first in the team.

[0245] Role A: Seven runner-up titles are enough to demonstrate his historical dominance. He is a perennial finals player. How many players have never even touched the finals floor?

[0246] Character B: Okay, I'm convinced, I agree with you.

[0247] Character C: Okay, I'm convinced, I agree with you.

[0248] 8 , which shows a flowchart of a task planning model training method provided by one embodiment of the present disclosure, specifically comprising the following steps:

[0249] Step 802: Acquire multiple sample task data, wherein the sample task data carries sample description information and object planning results of sample processing objects.

[0250] Step 804: Input the plurality of sample task data and sample description information into the task planning model to obtain object prediction results corresponding to the plurality of sample task data.

[0251] Step 806: According to the object planning results and the object prediction results, the parameters of the task planning model are adjusted to obtain a trained task planning model.

[0252] It should be noted that the implementation method of steps 802 to 806 can refer to the training method of the above-mentioned task planning model, and will not be described in detail in this embodiment of the present disclosure.

[0253] In actual applications, after obtaining the trained task planning model, the model parameters of the trained task planning model can be sent to the terminal device. The front-end user can build the task planning model locally based on the model parameters and use the task planning model to perform task processing.

[0254] Using the solution of the disclosed embodiments, a task planning model is trained based on the object planning results and object prediction results. If a preset stopping condition is not met, the task planning model continues to be trained until the preset stopping condition is met, completing the training and obtaining the task planning model. By continuously adjusting the model parameters of the task planning model, the resulting task planning model can be made more accurate.

[0255] 9 , which shows a flowchart of an information processing method based on a task planning model provided by an embodiment of the present disclosure, specifically comprising the following steps:

[0256] Step 902: Receive a task generation request, wherein the task generation request includes request information.

[0257] Specifically, the information processing method based on the task planning model can be applied to terminal devices and can also be applied to model training platforms. The task generation request is used to request the generation of task information for the target task. The task generation request usually includes the task type, the expected output format, and the request information. For example, when the user selects the "Smart Question and Answer" function on the front-end interface of the model training platform and uploads the task data, a task generation request can be constructed. The task generation request includes information such as task data, task type (i.e., smart question and answer), and data modality of the task data. Request information refers to the parameters or descriptive information related to the target task carried in the task generation request. The request information includes but is not limited to the task scenario identifier of the target task, the task model identifier, or the model training data of the target task.

[0258] Step 904: Based on the request information, obtain a task planning model, wherein the task planning model is trained based on object prediction results corresponding to multiple sample task data and object planning results carried by multiple sample task data, the sample task data carries sample description information of the sample processing object, and the object prediction result is obtained by inputting multiple sample task data and sample description information into the task planning model to be trained.

[0259] In an optional embodiment of the present disclosure, obtaining the task planning model based on the request information may include the following steps:

[0260] Based on the task scenario identifier, determine a target scenario template from a plurality of preset scenario templates, and based on the target scenario template, search for a task planning model from a model library, wherein the model library stores a plurality of models, and the request information includes the task scenario identifier of the target task; or

[0261] Based on the task model identifier, the task planning model is searched from the model library, wherein the request information includes the task model identifier of the target task.

[0262] Specifically, the task scenario identifier refers to a unique or specific label used to distinguish different task application scenarios. In the embodiment of the present disclosure, the task scenario identifier is part of the request information. Through the task scenario identifier, a target scenario template that matches the request information can be selected from a series of preset scenario templates to generate task information. For example, if the task scenario identifier is "intelligent question and answer", it means that the front-end user wants to generate an answer for the uploaded task data, then the intelligent question and answer scenario template can be selected from multiple preset scenario templates according to the task scenario identifier.

[0263] Preset scenario templates are predefined standard configuration scenario templates for different task application scenarios. Each template contains model information and task processing flow information that matches the task application scenario. Using a series of preset scenario templates, you can quickly respond to task generation requests for different scenarios. Different preset scenario templates correspond to different task types, model information, and processing flows.

[0264] The target scenario template is the scenario template that matches the task scenario identifier. When parsing a task generation request, the corresponding target scenario template can be located based on the task scenario identifier. Based on the model information included in the target scenario template, the corresponding task planning model and other related configuration information can be selected from the model library.

[0265] The model library is a centralized repository for deep learning models that have been trained and optimized to solve different tasks. The models stored in the model library include, but are not limited to, task processing models and task planning models. Furthermore, the models in the model library can be divided into different versions based on their applicable scenarios. For example, the model library may contain multiple versions of task processing models, such as speech generation models for speech tasks and image generation models for image tasks. Task model identifiers are unique or specific labels used to distinguish models applicable to different tasks.

[0266] Using the solution of the disclosed embodiments, a target scenario template is determined from multiple preset scenario templates based on a task scenario identifier. Based on the target scenario template, a task planning model is searched from a model library, where the model library stores multiple models and the request information includes the task scenario identifier of the target task. Alternatively, a task planning model is searched from the model library based on the task model identifier, where the request information includes the task model identifier of the target task. By utilizing predefined task scenario templates, task model identifiers, and model library resources, the process of acquiring task planning models becomes more flexible, efficient, and standardized.

[0267] In another optional embodiment of the present disclosure, in addition to selecting a pre-trained task planning model from a model library, a task planning model can be specifically trained based on sample task data in the request information. That is, the request information includes multiple sample task data of the target task. The above-mentioned acquisition of the task planning model based on the request information may include the following steps:

[0268] Based on multiple sample task data, the task planning model corresponding to the target task is trained to obtain a trained task planning model.

[0269] It should be noted that the implementation method of "training the task planning model corresponding to the target task based on multiple sample task data to obtain a trained task planning model" is the same as the training method of the above-mentioned task planning model, and will not be repeated in this disclosed embodiment.

[0270] By applying the solution of the embodiment of the present disclosure, a task planning model corresponding to the target task is trained based on multiple sample task data to obtain a trained task planning model, thereby ensuring that the task planning model better meets user needs and the accuracy of the task planning model.

[0271] Step 906: Generate task information based on the task planning model, wherein the task information is used to execute the target task.

[0272] Specifically, the task information includes the model configuration and processing flow required to execute the target task. Based on the task information, the terminal device or other server-side components can correctly use the task planning model to process the target task.

[0273] It should be noted that when generating task information based on a task planning model, the task planning model can be directly packaged to obtain the task information. Alternatively, the model information of the task planning model can be obtained and the task information can be constructed based on the model information. This model information includes model parameter configuration, input data processing methods, expected output specifications, possible intermediate steps involved, and other auxiliary information.

[0274] By applying the solution of the embodiment of the present disclosure, by generating task information of the target task, it is possible to reduce system deployment and operation and maintenance costs while ensuring the quality and efficiency of target task processing, and provide users with convenient and efficient task processing services.

[0275] Referring to FIG10 , FIG10 shows a schematic diagram of a model training platform provided by an embodiment of the present disclosure, the model training platform includes a request interface unit 1002 and a model unit 1004;

[0276] The request interface unit 1002 is configured to receive a model training request, wherein the model training request includes request information;

[0277] Model unit 1004 is used to obtain a task planning model; based on the request information, the task planning model is trained to obtain a trained task planning model, wherein the request information includes multiple sample task data, and the sample task data carries sample description information and object planning results of the sample processing object.

[0278] It should be noted that the processing method of the request interface unit and the model unit is the same as the training method of the above-mentioned task planning model, and will not be described in detail in the embodiment of the present disclosure.

[0279] In an optional embodiment of the present disclosure, the model training platform further includes a model library;

[0280] The model unit is also used to store the trained mission planning model in the model library.

[0281] It should be noted that the model unit trains the task planning model based on the request information. After obtaining the trained task planning model, the trained task planning model can also be stored in the model library, so that the trained task planning model can be directly obtained from the model library later without repeated training. Among them, the method of obtaining the trained task planning model from the model library can refer to the above-mentioned "based on the task scenario identifier, determine the target scenario template from multiple preset scenario templates, and based on the target scenario template, search for the task planning model from the model library, wherein the model library stores multiple models and the request information includes the task scenario identifier of the target task; or, based on the task model identifier, search for the task planning model from the model library, wherein the request information includes the task model identifier of the target task", and the embodiments of this disclosure will not be repeated.

[0282] Corresponding to the above method embodiment, the present disclosure also provides an embodiment of a task processing device. FIG11 shows a schematic diagram of the structure of a task processing device provided by an embodiment of the present disclosure. As shown in FIG11 , the device includes:

[0283] A first acquisition module 1102 is configured to acquire task data of a target task;

[0284] A first input module 1104 is configured to input task data and object description information of candidate processing objects into a task planning model to determine a plurality of processing objects corresponding to a target task, wherein the task planning model is used to plan the plurality of processing objects of the target task;

[0285] The first processing module 1106 is configured to process the task data using multiple processing objects to obtain processing results output by the multiple processing objects respectively;

[0286] The first determining module 1108 is configured to determine the task processing result of the target task according to the processing results respectively output by the multiple processing objects.

[0287] Optionally, the first processing module 1106 is further configured to call multiple task processing plug-ins to process the task data, obtain processing results respectively fed back by the multiple task processing plug-ins, wherein the processing object includes the task processing plug-in; and / or input the task data into multiple task processing models, obtain processing results respectively output by multiple task processing models, wherein the processing object includes the task processing model.

[0288] Optionally, the first processing module 1106 is further configured to determine the current processing object from multiple processing objects; use the current processing object to process the task data to obtain the current processing result output by the current processing object; use the task planning model to analyze the current processing result, and when it is determined that the target task has not been processed, return to execute the step of determining the current processing object from multiple processing objects until the target task is processed and the processing results output by multiple processing objects are obtained.

[0289] Optionally, the device also includes: a splitting module, configured to split the target task into multiple target subtasks, and determine the subtask data of the multiple target subtasks based on the task data, wherein the target subtasks correspond one-to-one to the processing objects; a first processing module 1106, further configured to use the current processing object to process the current subtask data to obtain the current processing result output by the current processing object, wherein the current subtask data is the subtask data of the target subtask corresponding to the current processing object.

[0290] Optionally, the device also includes: a third determination module, configured to determine a completed processing object from multiple processing objects, and obtain a completed processing result output by the completed processing object, wherein the completed processing object is a processing object that has completed processing of task data; the first processing module 1106, further configured to use the current processing object to process the completed processing result to obtain the current processing result output by the current processing object.

[0291] Optionally, the device further includes: a first retrieval module configured to obtain original object description information of multiple original processing objects; and retrieve candidate processing objects from the multiple original processing objects based on the task data and the original object description information.

[0292] Optionally, the device also includes: a second retrieval module, configured to obtain reference data; based on the task data, retrieving target reference data from the reference data; a first processing module 1106, further configured to input model prompt information, task data, object description information of candidate processing objects and target reference data into the task planning model to obtain multiple processing objects corresponding to the target task.

[0293] Optionally, the first determination module 1108 is further configured to input the processing results respectively output by the multiple processing objects into a result generation model to obtain the task processing result of the target task, wherein the result generation model is used to summarize the processing results respectively output by the multiple processing objects.

[0294] Optionally, the device also includes: a construction module, configured to send the task processing results to a front-end user; receive result feedback information sent by the front-end user, wherein the result feedback information is information that provides feedback on the task processing results based on the task information of the target task; construct model optimization data based on the result feedback information; and use the model optimization data to adjust the parameters of the task planning model.

[0295] Optionally, the construction module is further configured to generate optimization prompt information based on the result feedback information, wherein the optimization prompt information is used to guide the front-end user to send model optimization data for optimizing the task planning model; send the optimization prompt information to the front-end user, and receive the model optimization data sent by the front-end user based on the optimization prompt information.

[0296] Optionally, the device further includes: a marking module configured to mark key information in the task processing result to obtain an updated task processing result; and send the updated task processing result to a front-end user.

[0297] Optionally, the device also includes: an editing module, configured to send the task processing results to a front-end user; receive editing information sent by the front-end user, wherein the editing information is used to edit the task processing results; edit the task processing results according to the editing information to obtain the edited task processing results.

[0298] Optionally, the device also includes: a first sending module, configured to send processing prompt information to the front-end user, wherein the processing prompt information is used to guide the front-end user to send task data for target task processing; and receive the task data of the target task sent by the front-end user based on the processing prompt information.

[0299] Optionally, the device also includes: a fourth input module, configured to obtain multiple sample task data, wherein the sample task data carries sample description information and object planning results of the sample processing object; input the multiple sample task data and sample description information into the task planning model to obtain object prediction results corresponding to the multiple sample task data; adjust the parameters of the task planning model according to the object planning results and the object prediction results to obtain a trained task planning model.

[0300] Optionally, the first weight of the sample task data and the second weight of the sample description information are different; the fourth input module is further configured to input multiple sample task data, sample description information, the first weight and the second weight into the task planning model to obtain object prediction results corresponding to the multiple sample task data.

[0301] By applying the solution of the embodiment of the present disclosure, a task planning model is used to determine multiple processing objects for processing target tasks, thereby achieving seamless integration between the model and processing objects with different processing capabilities, expanding the task processing capabilities, and making task processing more efficient and comprehensive.

[0302] The above is a schematic scheme of a task processing device of this embodiment. It should be noted that the technical scheme of the task processing device and the technical scheme of the task processing method described above are of the same concept. For details not described in detail in the technical scheme of the task processing device, please refer to the description of the technical scheme of the task processing method described above.

[0303] Corresponding to the above-mentioned method embodiment, the present disclosure also provides an embodiment of a dialog task processing device. FIG12 shows a schematic diagram of the structure of a dialog task processing device provided by an embodiment of the present disclosure. As shown in FIG12 , the device includes:

[0304] A second acquisition module 1202 is configured to acquire task data of a target dialogue task, wherein the task data includes a plurality of dialogue roles;

[0305] The second input module 1204 is configured to input the task data and the object description information of the candidate processing objects into the task planning model to determine multiple processing objects corresponding to the target dialogue task, wherein the task planning model is used to plan multiple processing objects for the target dialogue task, and the processing objects correspond to the dialogue roles one by one;

[0306] The second processing module 1206 is configured to process the task data using multiple processing objects to obtain processing results output by the multiple processing objects respectively;

[0307] The second determining module 1208 is configured to determine the task processing result of the target dialogue task according to the processing results respectively output by the multiple processing objects.

[0308] Optionally, the device also includes: a second sending module, configured to send the task processing result to the front-end user; receive result feedback information sent by the front-end user, wherein the result feedback information is information that provides feedback on the task processing result based on the task information of the target dialogue task; and send the result feedback information to the model training platform, wherein the model training platform is used to use the result feedback information to adjust the parameters of the task planning model.

[0309] By applying the solution of the embodiment of the present disclosure, a task planning model is used to determine multiple processing objects for processing the target dialogue task, thereby achieving seamless integration between the model and processing objects with different processing capabilities, expanding the dialogue task processing capabilities, and making dialogue task processing more efficient and comprehensive.

[0310] The above is a schematic diagram of a dialog task processing device according to this embodiment. It should be noted that the technical solution of this dialog task processing device and the technical solution of the dialog task processing method described above are based on the same concept. For details not described in detail in the technical solution of the dialog task processing device, please refer to the description of the technical solution of the dialog task processing method described above.

[0311] Corresponding to the above method embodiment, the present disclosure also provides an embodiment of a task planning model training device. FIG13 shows a schematic diagram of the structure of a task planning model training device provided by an embodiment of the present disclosure. As shown in FIG13, the device includes:

[0312] The third acquisition module 1302 is configured to acquire a plurality of sample task data, wherein the sample task data carries sample description information and object planning results of the sample processing object;

[0313] The third input module 1304 is configured to input a plurality of sample task data and sample description information into the task planning model to obtain object prediction results corresponding to the plurality of sample task data respectively;

[0314] The adjustment module 1306 is configured to adjust the parameters of the task planning model according to the object planning result and the object prediction result to obtain a trained task planning model.

[0315] Using the solution of the disclosed embodiments, a task planning model is trained based on the object planning results and object prediction results. If a preset stopping condition is not met, the task planning model continues to be trained until the preset stopping condition is met, completing the training and obtaining the task planning model. By continuously adjusting the model parameters of the task planning model, the resulting task planning model can be made more accurate.

[0316] The above is a schematic diagram of a task planning model training device according to this embodiment. It should be noted that the technical solution of the task planning model training device and the technical solution of the task planning model training method described above are based on the same concept. For details not described in detail in the technical solution of the task planning model training device, please refer to the description of the technical solution of the task planning model training method described above.

[0317] Corresponding to the above-mentioned method embodiment, the present disclosure also provides an embodiment of an information processing device based on a task planning model. FIG14 shows a schematic diagram of the structure of an information processing device based on a task planning model provided by one embodiment of the present disclosure. As shown in FIG14 , the device includes:

[0318] The receiving module 1402 is configured to receive a task generation request, wherein the task generation request includes request information;

[0319] A fourth acquisition module 1404 is configured to acquire a task planning model based on the request information, wherein the task planning model is trained based on object prediction results corresponding to the plurality of sample task data and object planning results carried by the plurality of sample task data, the sample task data carrying sample description information of the sample processing object, and the object prediction results are obtained by inputting the plurality of sample task data and the sample description information into the task planning model to be trained;

[0320] The generation module 1406 is configured to generate task information based on the task planning model, wherein the task information is used to perform the target task.

[0321] Optionally, the fourth acquisition module 1404 is further configured to determine a target scenario template from multiple preset scenario templates based on the task scenario identifier, and search for a task planning model from a model library based on the target scenario template, wherein the model library stores multiple models and the request information includes the task scenario identifier of the target task; or, search for a task planning model from the model library based on the task model identifier, wherein the request information includes the task model identifier of the target task.

[0322] Optionally, the request information includes multiple sample task data of the target task; the fourth acquisition module 1404 is further configured to train the task planning model corresponding to the target task based on the multiple sample task data to obtain a trained task planning model.

[0323] By applying the solution of the embodiment of the present disclosure, by generating task information of the target task, it is possible to reduce system deployment and operation and maintenance costs while ensuring the quality and efficiency of target task processing, and provide users with convenient and efficient task processing services.

[0324] The above is a schematic diagram of an information processing device based on a task planning model according to this embodiment. It should be noted that the technical solution of the information processing device based on the task planning model and the technical solution of the information processing method based on the task planning model are based on the same concept. For details not described in detail in the technical solution of the information processing device based on the task planning model, please refer to the description of the technical solution of the information processing method based on the task planning model.

[0325] Figure 15 shows a block diagram of a computing device according to an embodiment of the present disclosure. Components of the computing device 1500 include, but are not limited to, a memory 1510 and a processor 1520. The processor 1520 is connected to the memory 1510 via a bus 1530, and a database 1550 is used to store data.

[0326] The computing device 1500 also includes an access device 1540 that enables the computing device 1500 to communicate via one or more networks 1560. Examples of such networks include a public switched telephone network (PSTN), a local area network (LAN), a wide area network (WAN), a personal area network (PAN), or a combination of communication networks such as the Internet. The access device 1540 may include one or more of any type of network interface (e.g., a network interface card (NIC)) whether wired or wireless, such as an IEEE 802.11 wireless local area network (WLAN) wireless interface, a World Interoperability for Microwave Access (Wi-MAX) interface, an Ethernet interface, a universal serial bus (USB) interface, a cellular network interface, a Bluetooth interface, a near field communication (NFC) interface, and the like.

[0327] In one embodiment of the present disclosure, the aforementioned components of computing device 1500 and other components not shown in FIG15 may also be connected to each other, for example, via a bus. It should be understood that the block diagram of the computing device structure shown in FIG15 is for illustrative purposes only and does not limit the scope of the present disclosure. Those skilled in the art may add or replace other components as needed.

[0328] Computing device 1500 may be any type of stationary or mobile computing device, including a mobile computer or mobile computing device (e.g., a tablet computer, a personal digital assistant, a laptop computer, a notebook computer, a netbook computer, etc.), a mobile phone (e.g., a smartphone), a wearable computing device (e.g., a smartwatch, smart glasses, etc.), or other types of mobile devices, or a stationary computing device such as a desktop computer or a personal computer (PC). Computing device 1500 may also be a mobile or stationary server.

[0329] Among them, the processor 1520 is used to execute computer programs / instructions, which, when executed by the processor, implement the steps of the above-mentioned task processing method or dialogue task processing method or task planning model training method or information processing method based on the task planning model.

[0330] The above is a schematic diagram of a computing device according to this embodiment. It should be noted that the technical solution of this computing device is based on the same concept as the technical solutions of the aforementioned task processing method, dialogue task processing method, task planning model training method, and task planning model-based information processing method. For details not described in detail in the technical solution of the computing device, please refer to the description of the technical solution of the aforementioned task processing method, dialogue task processing method, task planning model training method, or task planning model-based information processing method.

[0331] An embodiment of the present disclosure also provides a computer-readable storage medium storing a computer program / instruction, which, when executed by a processor, implements the steps of the above-mentioned task processing method or dialogue task processing method or task planning model training method or information processing method based on the task planning model.

[0332] The above is a schematic scheme of a computer-readable storage medium of this embodiment. It should be noted that the technical scheme of this storage medium and the technical schemes of the aforementioned task processing method, dialogue task processing method, task planning model training method, and task planning model-based information processing method are based on the same concept. For details not described in detail in the technical scheme of the storage medium, please refer to the description of the technical scheme of the aforementioned task processing method, dialogue task processing method, task planning model training method, or task planning model-based information processing method.

[0333] An embodiment of the present disclosure also provides a computer program product, including a computer program / instruction, which, when executed by a processor, implements the steps of the above-mentioned task processing method or dialogue task processing method or task planning model training method or information processing method based on the task planning model.

[0334] The above is a schematic scheme of a computer program product of this embodiment. It should be noted that the technical scheme of this computer program product and the technical schemes of the aforementioned task processing method, dialogue task processing method, task planning model training method, and task planning model-based information processing method are based on the same concept. For details not described in detail in the technical scheme of the computer program product, please refer to the description of the technical scheme of the aforementioned task processing method, dialogue task processing method, task planning model training method, or task planning model-based information processing method.

[0335] The foregoing description describes specific embodiments of the present disclosure. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recited in the claims can be performed in an order different from that described in the embodiments and still achieve the desired results. Furthermore, the processes depicted in the accompanying drawings do not necessarily require the specific order shown or the sequential order to achieve the desired results. In certain embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0336] The computer instructions include computer program code, which may be in source code form, object code form, executable file, or some intermediate form. The computer-readable medium may include: any entity or device capable of carrying the computer program code, recording medium, USB flash drive, mobile hard disk, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electric carrier signal, telecommunication signal, and software distribution medium. It should be noted that the content contained in the computer-readable medium may be appropriately increased or decreased according to the requirements of patent practice. For example, in some regions, according to patent practice, computer-readable media does not include electric carrier signals and telecommunication signals.

[0337] It should be noted that for the aforementioned method embodiments, for the sake of simplicity of description, they are all expressed as a series of action combinations, but those skilled in the art should be aware that the embodiments of the present disclosure are not limited by the order of the actions described, because according to the embodiments of the present disclosure, certain steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should also be aware that the embodiments described in the specification are all preferred embodiments, and the actions and modules involved are not necessarily required by the embodiments of the present disclosure.

[0338] In the above embodiments, the description of each embodiment has its own focus. For parts that are not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0339] The preferred embodiments of the present disclosure disclosed above are only used to help illustrate the present disclosure. The optional embodiments do not describe all details in detail, nor do they limit the invention to only the specific embodiments described. Obviously, many modifications and variations can be made based on the content of the embodiments of the present disclosure. The present disclosure selects and specifically describes these embodiments in order to better explain the principles and practical applications of the embodiments of the present disclosure, so that those skilled in the art can better understand and utilize the present disclosure. The present disclosure is limited only by the claims and their full scope and equivalents.

Claims

1. A task processing method, comprising: Get the task data of the target task; Inputting the task data and object description information of the candidate processing objects into a task planning model to determine a plurality of processing objects corresponding to the target task, wherein the task planning model is used to plan the plurality of processing objects of the target task; Processing the task data using the multiple processing objects to obtain processing results outputted by the multiple processing objects respectively; The task processing result of the target task is determined according to the processing results respectively output by the multiple processing objects.

2. The method according to claim 1, wherein the processing of the task data by the plurality of processing objects to obtain processing results outputted by the plurality of processing objects respectively comprises: Calling multiple task processing plug-ins to process the task data, and obtaining processing results respectively fed back by the multiple task processing plug-ins, wherein the processing object includes the task processing plug-ins; and / or, The task data is input into a plurality of task processing models to obtain processing results respectively output by the plurality of task processing models, wherein the processing object includes the task processing model.

3. The method according to claim 1, wherein the processing of the task data by the plurality of processing objects to obtain processing results outputted by the plurality of processing objects respectively comprises: determining a current processing object from the plurality of processing objects; Processing the task data using the current processing object to obtain a current processing result output by the current processing object; The task planning model is used to analyze the current processing result. When it is determined that the target task has not been completed, the step of determining the current processing object from the multiple processing objects is returned to execute until the target task is completed, and the processing results output by the multiple processing objects are obtained.

4. The method according to claim 3, before processing the task data using the current processing object to obtain the current processing result output by the current processing object, further comprising: Splitting the target task into a plurality of target subtasks, and determining subtask data of the plurality of target subtasks based on the task data, wherein the target subtasks correspond to the processing objects one by one; The processing of the task data by using the current processing object to obtain a current processing result output by the current processing object includes: The current subtask data is processed using the current processing object to obtain a current processing result output by the current processing object, wherein the current subtask data is the subtask data of the target subtask corresponding to the current processing object.

5. The method according to claim 3, before processing the task data using the current processing object to obtain the current processing result output by the current processing object, further comprising: Determining a completed processing object from the multiple processing objects, and obtaining a completed processing result output by the completed processing object, wherein the completed processing object is a processing object that has completed processing the task data; The processing of the task data by using the current processing object to obtain a current processing result output by the current processing object includes: The completed processing result is processed using the current processing object to obtain a current processing result output by the current processing object.

6. The method according to any one of claims 1 to 5, further comprising: before inputting the task data and the object description information of the candidate processing objects into the task planning model and determining the multiple processing objects corresponding to the target task; Obtaining original object description information of multiple original processing objects; The candidate processing object is retrieved from the plurality of original processing objects according to the task data and the original object description information.

7. The method according to any one of claims 1 to 5, further comprising: before inputting the task data and the object description information of the candidate processing objects into a task planning model and determining the multiple processing objects corresponding to the target task; Obtain reference data; Retrieving target reference data from the reference data according to the task data; The step of inputting the task data and the object description information of the candidate processing objects into a task planning model to determine a plurality of processing objects corresponding to the target task includes: The model prompt information, the task data, the object description information of the candidate processing objects and the target reference data are input into the task planning model to obtain a plurality of processing objects corresponding to the target task.

8. The method according to any one of claims 1 to 7, wherein determining the task processing result of the target task based on the processing results respectively output by the multiple processing objects comprises: The processing results respectively output by the plurality of processing objects are input into a result generation model to obtain a task processing result of the target task, wherein the result generation model is used to summarize the processing results respectively output by the plurality of processing objects.

9. The method according to any one of claims 1 to 8, further comprising: after determining the task processing result of the target task based on the processing results respectively output by the plurality of processing objects; Sending the task processing result to the front-end user; Receiving result feedback information sent by the front-end user, wherein the result feedback information is information providing feedback on the task processing result based on the task information of the target task; Building a model to optimize data based on the result feedback information; The model optimization data is used to adjust parameters of the mission planning model.

10. The method according to claim 9, wherein constructing model optimization data based on the result feedback information comprises: Generate optimization prompt information according to the result feedback information, wherein the optimization prompt information is used to guide the front-end user to send model optimization data for optimizing the task planning model; The optimization prompt information is sent to the front-end user, and the model optimization data sent by the front-end user based on the optimization prompt information is received.

11. The method according to any one of claims 1 to 10, further comprising: after determining the task processing result of the target task based on the processing results respectively output by the plurality of processing objects; Marking key information in the task processing result to obtain an updated task processing result; The updated task processing result is sent to the front-end user.

12. The method according to any one of claims 1 to 11, further comprising: after determining the task processing result of the target task based on the processing results respectively output by the plurality of processing objects; Sending the task processing result to the front-end user; receiving editing information sent by the front-end user, wherein the editing information is used to edit the task processing result; The task processing result is edited according to the editing information to obtain an edited task processing result.

13. The method according to any one of claims 1 to 12, before obtaining the task data of the target task, further comprising: Sending processing prompt information to the front-end user, wherein the processing prompt information is used to guide the front-end user to send task data for target task processing; Receive task data of the target task sent by the front-end user based on the processing prompt information.

14. The method according to any one of claims 1 to 13, further comprising: before inputting the task data and the object description information of the candidate processing objects into a task planning model and determining the multiple processing objects corresponding to the target task; Acquire a plurality of sample task data, wherein the sample task data carries sample description information and object planning results of sample processing objects; Inputting the plurality of sample task data and the sample description information into a task planning model to obtain object prediction results corresponding to the plurality of sample task data respectively; According to the object planning result and the object prediction result, the parameters of the task planning model are adjusted to obtain a trained task planning model.

15. The method according to claim 14, wherein the first weight of the sample task data and the second weight of the sample description information are different; The step of inputting the plurality of sample task data and the sample description information into a task planning model to obtain object prediction results corresponding to the plurality of sample task data respectively includes: The plurality of sample task data, the sample description information, the first weight, and the second weight are input into a task planning model to obtain object prediction results corresponding to the plurality of sample task data respectively.

16. A method for processing a dialogue task, comprising: Acquire task data of a target dialogue task, wherein the task data includes a plurality of dialogue roles; Inputting the task data and object description information of the candidate processing objects into a task planning model to determine a plurality of processing objects corresponding to the target dialogue task, wherein the task planning model is used to plan the plurality of processing objects of the target dialogue task, and the processing objects correspond one-to-one to the dialogue roles; Processing the task data using the multiple processing objects to obtain processing results outputted by the multiple processing objects respectively; The task processing result of the target dialogue task is determined according to the processing results respectively output by the multiple processing objects.

17. The method according to claim 16, after determining the task processing result of the target dialogue task based on the processing results respectively output by the multiple processing objects, further comprising: Sending the task processing result to the front-end user; Receiving result feedback information sent by the front-end user, wherein the result feedback information is information providing feedback on the task processing result based on the task information of the target dialogue task; The result feedback information is sent to a model training platform, wherein the model training platform is used to adjust parameters of the task planning model using the result feedback information.

18. A task planning model training method, comprising: Acquire a plurality of sample task data, wherein the sample task data carries sample description information and object planning results of sample processing objects; Inputting the plurality of sample task data and the sample description information into a task planning model to obtain object prediction results corresponding to the plurality of sample task data respectively; According to the object planning result and the object prediction result, the parameters of the task planning model are adjusted to obtain a trained task planning model.

19. An information processing method based on a task planning model, comprising: receiving a task generation request, wherein the task generation request includes request information; Based on the request information, a task planning model is obtained, wherein the task planning model is trained based on object prediction results corresponding to a plurality of sample task data and object planning results carried by the plurality of sample task data, the sample task data carries sample description information of a sample processing object, and the object prediction results are obtained by inputting the plurality of sample task data and the sample description information into the task planning model to be trained; Based on the task planning model, task information is generated, wherein the task information is used to perform a target task.

20. The method according to claim 19, wherein obtaining a mission planning model based on the request information comprises: Based on the task scenario identifier, determine a target scenario template from a plurality of preset scenario templates, and based on the target scenario template, search for a task planning model from a model library, wherein the model library stores a plurality of models, and the request information includes the task scenario identifier of the target task; or Based on the task model identifier, a task planning model is searched from the model library, wherein the request information includes the task model identifier of the target task.

21. The method according to claim 19, wherein the request information includes a plurality of sample task data of the target task; The obtaining of a task planning model based on the request information includes: Based on the multiple sample task data, the task planning model corresponding to the target task is trained to obtain a trained task planning model.

22. A model training platform, comprising a request interface unit and a model unit; The request interface unit is used to receive a model training request, wherein: The model training request includes request information; The model unit is used to obtain a task planning model; based on the request information, the task planning model is trained to obtain a trained task planning model, wherein the request information includes multiple sample task data, and the sample task data carries sample description information and object planning results of the sample processing object.

23. The model training platform according to claim 22, further comprising a model library; The model unit is further used to store the trained task planning model in the model library.

24. A computing device comprising: memory and processor; The memory is used to store computer programs / instructions, and the processor is used to execute the computer programs / instructions. When the computer program / instructions are executed by the processor, the steps of the method described in any one of claims 1 to 15 or any one of claims 16 to 17 or claim 18 or any one of claims 19 to 21 are implemented.

25. A computer-readable storage medium storing a computer program / instruction, which, when executed by a processor, implements the steps of the method described in any one of claims 1 to 15 or any one of claims 16 to 17 or claim 18 or any one of claims 19 to 21.

26. A computer program product comprising a computer program / instructions, which, when executed by a processor, implements the steps of the method according to any one of claims 1 to 15 or any one of claims 16 to 17 or claim 18 or any one of claims 19 to 21.

Images