User Task Processing Method, Device, Equipment, and Medium Based on a Multi-Expert Model

Abstract

The present invention discloses a user task processing method, apparatus, device, and medium based on a multi-expert model, including: obtaining the total task of a target user according to multimodal data; decomposing the total task to obtain a number of subtasks; constructing under the multi-expert model; for each subtask, performing steps S141-S144, including: step S141, controlling a gating network unit to determine the confidence levels corresponding to the processing of the subtask by each vertical expert network; step S142, determining a number of target vertical expert networks; step S143, adjusting the target vertical expert networks; step S144, based on the number of target vertical expert networks, obtaining a multi-dimensional processing result of the subtask; according to a number of multi-dimensional processing results, obtaining a processing result of the total task of the target user. The present invention belongs to the field of task processing. The present invention can improve the accuracy of result hitting.

Description

Technical Field

[0001] The present invention relates to the field of task processing, and particularly to a user task processing method, device, equipment, and medium based on a multi-expert model. Background Art

[0002] When dealing with complex tasks, large models usually adopt deep learning architectures (such as Transformer) for hierarchical feature extraction and adapt to specific task requirements through pre-training and fine-tuning. At the same time, data augmentation, regularization techniques (such as Dropout), and optimization algorithms (such as Adam) are used to improve the generalization ability of the model, and distributed training and hardware acceleration (such as GPU, TPU) are used to improve the training efficiency.

[0003] However, a single model may not achieve optimal performance when dealing with diverse tasks. Therefore, how to improve the accuracy of the processing result hits is an urgent problem to be solved. Summary of the Invention

[0004] The present invention provides a user task processing method, device, equipment, and medium based on a multi-expert model, solves the technical problem that a single model in the prior art may not achieve optimal performance when dealing with diverse tasks, and realizes the technical effect of improving the accuracy of the processing result hits.

[0005] In a first aspect, the present invention provides a user task processing method based on a multi-expert model, and the method includes:

[0006] Obtain multi-modal data of a target user, and obtain the total task of the target user according to the multi-modal data;

[0007] Decompose the total task to obtain a number of subtasks;

[0008] Construct a multi-expert model, where the multi-expert model includes a gating network unit and a vertical expert network cluster. The vertical expert network cluster includes a number of vertical expert networks, each vertical expert network corresponds to a preset field, and all vertical expert networks are open networks;

[0009] For each subtask, execute steps S141 - S144, including:

[0010] Step S141, control the gating network unit to determine the confidence levels of each vertical expert network for processing the subtask;

[0011] Step S142, determine a number of target vertical expert networks according to the confidence levels of each vertical expert network;

[0012] Step S143: Adjust the target vertical expert network according to the dynamic processing information of the target vertical expert network or the relevance between target vertical expert networks, where the dynamic information includes the running state and the processing state;

[0013] Step S144: Obtain the multi-dimensional processing results of the subtask based on a number of target vertical expert networks;

[0014] Obtain the processing results of the total task of the target user according to a number of multi-dimensional processing results.

[0015] Further, control the gating network unit to determine the confidence levels corresponding to each vertical expert network for processing the subtask, including:

[0016] Perform feature preprocessing and dimension expansion on the initial features of the subtask to obtain expanded and fused features;

[0017] Map the expanded and fused features to a dimensional space with the same number of dimensions as the vertical expert networks through a dimensionality reduction layer, and generate a confidence distribution based on the Softmax function. The confidence distribution includes the confidence levels of each vertical expert network;

[0018] Adjust the confidence distribution based on the historical processing capabilities of the vertical expert networks, where the historical processing capabilities include historical accuracy and the passing rate of processing time limits.

[0019] Further, adjust the confidence distribution based on the historical processing capabilities of the vertical expert networks, including:

[0020]

[0021] where, is the confidence level of the th vertical expert network after adjustment, is the confidence level of the th vertical expert network before adjustment, is the historical accuracy, is the passing rate of processing time limits.

[0022] Further, determine a number of target vertical expert networks according to the confidence levels of each vertical expert network, including:

[0023] Determine the confidence density according to the confidence distribution;

[0024] Set a reference value for the number of expert networks, and determine the number G of target vertical expert networks according to the confidence density of the vertical expert network cluster and the reference value for the number of expert networks;

[0025] Select the top G vertical expert networks from the confidence distribution as the target vertical expert networks.

[0026] Further, it includes:

[0027]

[0028] Among them, is the confidence density, is the number of vertical expert networks;

[0029] It also includes:

[0030]

[0031] Among them, is the number of target vertical expert networks and is rounded up, is the reference value of the number of expert networks.

[0032] Further, according to the dynamic processing information of the target vertical expert network or the relevance between the target vertical expert networks, the target vertical expert network is adjusted, including:

[0033] Sequentially judge whether the target vertical expert network is in a faulty state or an occupied state according to the operating state of the target vertical expert network;

[0034] If it is in such a state, the target vertical expert network is removed and filled in order with the vertical expert networks that have not been selected; if it is not in such a state, no removal is performed;

[0035] Determine the first expert network among several target vertical expert networks;

[0036] Based on the preset knowledge graph, determine the degree of relevance between the first expert network and the remaining target vertical expert networks, and sort the remaining target vertical expert networks according to the degree of relevance.

[0037] Further, based on several target vertical expert networks, obtain the multi-dimensional processing results of this subtask, including:

[0038] Based on several target vertical expert networks, process this subtask to obtain the first single-dimensional processing result of each target vertical expert network;

[0039] According to the sorting of the target vertical expert networks, determine the result weights corresponding to each target vertical expert network;

[0040] According to the result weights corresponding to each target vertical expert network and the corresponding first single-dimensional processing results, obtain the second single-dimensional processing result;

[0041] According to several second single-dimensional processing results, obtain the multi-dimensional processing results of this subtask.

[0042] Second aspect, the present invention provides a user task processing device based on a multi-expert model, the device comprising:

[0043] An acquisition module, configured to acquire multi-modal data of a target user and obtain a total task of the target user according to the multi-modal data;

[0044] A decomposition module, configured to decompose the total task to obtain a plurality of subtasks;

[0045] A construction module, configured to construct a multi-expert model, wherein the multi-expert model includes a gating network unit and a vertical expert network cluster, the vertical expert network cluster includes a plurality of vertical expert networks, each vertical expert network corresponds to a preset field and the vertical expert networks are all open networks;

[0046] A multi-dimensional processing module, configured to execute steps S141-S144 for each subtask, including: step S141, controlling the gating network unit to determine the confidence levels of each vertical expert network for processing the subtask; step S142, determining a plurality of target vertical expert networks according to the confidence levels of each vertical expert network; step S143, adjusting the target vertical expert networks according to the dynamic processing information of the target vertical expert networks or the relevance between the target vertical expert networks, wherein the dynamic information includes an operating state and a processing state; step S144, obtaining a multi-dimensional processing result of the subtask based on the plurality of target vertical expert networks;

[0047] A result module, configured to obtain a processing result of the total task of the target user according to the plurality of multi-dimensional processing results.

[0048] Third aspect, the present invention provides an electronic device, comprising:

[0049] A processor;

[0050] A memory for storing instructions executable by the processor;

[0051] Wherein, the processor is configured to execute to implement the user task processing method based on a multi-expert model provided in the first aspect.

[0052] Fourth aspect, the present invention provides a non-transitory computer-readable storage medium, when the instructions in the storage medium are executed by the processor of the electronic device, enabling the electronic device to execute and implement the user task processing method based on a multi-expert model provided in the first aspect.

[0053] One or more technical solutions provided in the present invention have at least the following technical effects or advantages:

[0054] The present invention combines the historical performance (accuracy and timeliness) of experts with the relevance of the current task by comprehensively considering the historical accuracy rate and the passing rate of processing time limits, avoiding the deviation of a single indicator, enhancing the reliability of decision-making, achieving an accurate match between task requirements and expert capabilities through time limit and accuracy indicators, and ultimately taking into account the response speed and resource utilization while ensuring high precision.

[0055] The present invention can optimize the configuration quantity of the target vertical expert network through the confidence density, improve the accuracy of result generation, and avoid the waste of computing resources.

[0056] The present invention first pre-orders a number of target vertical expert networks through a preset knowledge graph and confidence distribution, and assigns weights to each result according to the sorting result, making the processing results corresponding to subtasks more in line with the actual situation and improving the answer hit rate.

[0057] The present invention integrates and splits a problem through a total (total task)-sub (subtask)-total (total result) structure, and combines the results of the split processing as the total result, which can improve the hit rate of the result. Brief Description of the Drawings

[0058] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the following drawings are some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0059] Figure 1 It is a schematic flowchart of the user task processing method based on a multi-expert model provided by the present invention;

[0060] Figure 2 It is a schematic structural diagram of the user task processing device based on a multi-expert model provided by the present invention. Detailed Embodiments

[0061] The embodiments of the present invention solve the technical problem that a single model may not achieve optimal performance when processing diverse tasks in the prior art by providing a user task processing method based on a multi-expert model.

[0062] The technical solution of the present invention to solve the above technical problem is generally as follows:

[0063] User task processing method based on a multi-expert model. The method includes: obtaining multi-modal data of a target user, and obtaining the total task of the target user based on the multi-modal data; decomposing the total task to obtain several sub-tasks; constructing a multi-expert model, where the multi-expert model includes a gating network unit and a vertical expert network cluster, and the vertical expert network cluster includes several vertical expert networks, each vertical expert network corresponds to a preset field and all vertical expert networks are open networks; for each sub-task, execute steps S141 - S144, including: step S141, controlling the gating network unit to determine the confidence levels of each vertical expert network for processing the sub-task; step S142, determining several target vertical expert networks according to the confidence levels of each vertical expert network; step S143, adjusting the target vertical expert networks according to the dynamic processing information of the target vertical expert networks or the relevance between the target vertical expert networks, where the dynamic information includes the running state and the processing state; step S144, obtaining the multi-dimensional processing result of the sub-task based on several target vertical expert networks; obtaining the processing result of the total task of the target user according to several multi-dimensional processing results.

[0064] To better understand the above technical solution, the above technical solution will be described in detail below in conjunction with the accompanying drawings of the specification and specific implementation manners.

[0065] First, it should be noted that the term "and / or" appearing in this article is merely a description of the association relationship of associated objects, indicating that there can be three relationships. For example, A and / or B can represent: A exists alone, A and B exist simultaneously, and B exists alone. In addition, the character " / " in this article generally represents an "or" relationship between the front and back associated objects.

[0066] The present invention provides a Figure 1 user task processing method based on a multi-expert model as shown, including steps S11 - S15:

[0067] Step S11, obtain multi-modal data of a target user, and obtain the total task of the target user based on the multi-modal data.

[0068] The target user refers to the user for whom the processing result is to be obtained. The multi-modal data refers to data from different sources or in different forms, and specifically may include: text data, image data, audio data, video data, behavior data, or sensor data, etc.

[0069] After obtaining the multi-modal data, preprocessing can be performed on the multi-modal data, including: cleaning, alignment, annotation, and normalization, etc. These are conventional data processing means and will not be elaborated here.

[0070] Subsequently, in the feature extraction stage, multi-modal data is merged and feature interactions are performed to obtain fused features (for example, concatenating text word vectors and image CNN features and then inputting them into LSTM). For the fused features, a pre-trained model (BERT / XLNet) is used for intent classification (for example, inputting the fused features → fully connected layer → Softmax classification), identifying the key information of the task (such as date, location, amount), and based on the task key information, obtaining the task description, which is the overall task referred to in the present invention (for example, ordering a gluten-free Margherita pizza on August 24, 2024 and delivering it to the specified address A).

[0071] Step S12, decompose the overall task to obtain a number of subtasks.

[0072] After obtaining the overall task in step S11, the description of the overall task can be analyzed based on natural language processing (NLP) techniques to extract key information.

[0073] Taking the above example: "Ordering a gluten-free Margherita pizza on August 24, 2024 and delivering it to the specified address A".

[0074] The key information includes:

[0075] Time: August 24, 2024

[0076] Pizza type: Margherita.

[0077] Special requirement: Gluten-free.

[0078] Delivery address: A.

[0079] Payment: Payment needs to be completed.

[0080] According to the extracted key information, the overall task is decomposed into a number of subtasks. Each subtask can correspond to one or more fields. For example, if image recognition is required, it is classified into the image recognition field; if text reasoning is required, it is classified into the text reasoning recognition field; if speech processing is required, it is the speech recognition field. The preset fields can be set according to the actual situation and are not limited here.

[0081] Step S13, construct a multi-expert model. Among them, the multi-expert model includes a gating network unit and a vertical expert network cluster. The vertical expert network cluster includes a number of vertical expert networks, each vertical expert network corresponds to a preset field and all vertical expert networks are open networks.

[0082] In practical applications, the vertical expert network cluster is usually a feed-forward network, and the gating network determines which input data is sent to which vertical expert network.

[0083] The vertical expert network cluster referred to in the present invention includes several vertical expert networks, specifically, it can be domain-specific experts, data modality-specific experts, algorithm-specific experts, problem complexity-specific experts, user group-specific experts, geographical area-specific experts, etc.;

[0084] Among the data modality-specific experts, there can be experts in specific fields. For example, a corresponding vertical expert network can be constructed for text, image, and audio respectively.

[0085] Each expert network is an independent neural network. The method in the embodiments of this specification can be applied to a server equipped with a multi-expert model or a business processing platform.

[0086] In practical applications, a standardized interaction interface can be designed to facilitate interaction with the vertical expert network. For example, using the RESTful API interface to define a unified request and response format, which is convenient for integration between different systems.

[0087] Step S14. For each subtask, execute steps S141 - S144, including:

[0088] Step S141. Control the gating network unit to determine the confidence levels corresponding to each vertical expert network for processing the subtask.

[0089] Specifically, it includes:

[0090] Perform feature preprocessing and dimension expansion on the initial features of the subtask to obtain expanded and fused features; map the expanded and fused features to a dimensional space with the same number of dimensions as the vertical expert networks through a dimensionality reduction layer, and generate a confidence distribution based on the Softmax function. The confidence distribution includes the confidence levels of each vertical expert network; adjust the confidence distribution based on the historical processing capabilities of the vertical expert networks, where the historical processing capabilities include historical accuracy and processing time limit compliance rate.

[0091] In a multi-expert network system, the role of the gating network is to assign a confidence level to each vertical expert network, thereby determining which vertical expert networks should be called and what weights should be given to them when processing subtasks.

[0092] Feature preprocessing can include normalization, etc.

[0093] In order to better capture the feature information of the subtask, it is necessary to perform dimension expansion on the preprocessed features. This can be achieved through some non-linear or linear transformations, such as using a multi-layer perceptron (MLP) or a convolutional neural network (CNN) (reference can also be made to patent document CN118520903A). Dimension expansion can increase the expressive ability of features, enabling the model to learn more complex feature relationships.

[0094] The expanded fusion features are mapped to a dimensional space with the same number of vertical expert networks through a dimensionality reduction layer, and the confidence distribution is generated based on the Softmax function.

[0095] The dimension of the expanded fusion feature may be high, which is not conducive to subsequent calculation and analysis. Therefore, it is necessary to map it to the same dimensional space as the number of vertical expert networks through a dimensionality reduction layer. The dimensionality reduction layer can be implemented using a fully connected layer, which converts the high-dimensional feature vector into a vector equal to the number of vertical expert networks.

[0096] Based on the historical processing capabilities of the vertical expert network, the confidence distribution is adjusted, including:

[0097]

[0098] in, After adjustment The confidence of a vertical expert network, Before adjustment The confidence of a vertical expert network, is the historical accuracy, It is the processing time limit achievement rate.

[0099] The historical accuracy rate refers to the accuracy rate of the vertical expert network in processing subtasks in this field in the past. For example, if 90 out of 100 tasks are accurate, the accuracy rate is 90%. The processing time limit compliance rate refers to the timeliness rate of completing tasks within the processing time limit when the vertical expert network processed subtasks in this field in the past. For example, if 80 out of 100 tasks are completed within the time limit, the accuracy rate is 80%.

[0100] The present invention combines the historical performance (accuracy and timeliness) of experts with the relevance to the current task by comprehensively considering the historical accuracy rate and the processing time limit compliance rate, avoids the deviation of a single indicator, improves the reliability of decision-making, and achieves precise matching of task requirements and expert capabilities through time limit and accuracy indicators. Ultimately, it takes into account both response speed and resource utilization while ensuring high accuracy.

[0101] Step S142: determining a number of target vertical expert networks according to the confidence of each vertical expert network.

[0102] Specifically, the method includes: determining the confidence density according to the confidence distribution; setting a reference value for the number of expert networks, and determining the number G of target vertical expert networks according to the confidence density of the vertical expert network cluster and the reference value for the number of expert networks; and selecting the first G vertical expert networks from the confidence distribution as the target vertical expert networks.

[0103] include:

[0104]

[0105] Among them, is the confidence density, is the number of vertical expert networks;

[0106] It also includes:

[0107]

[0108] Among them, is the number of target vertical expert networks and is rounded up to G, is the reference value of the number of expert networks. It can be understood that when processing a certain subtask, multiple vertical expert networks may be required to cooperate, and the confidence levels of each vertical expert network in processing this subtask are different. The reference value K of the number of expert networks can be determined according to the actual situation, or it can be the number of all vertical expert networks in the present invention.

[0109] For example, when processing a certain subtask, the confidence level distributions of each vertical expert network are 98%, 90%, 50%, and 0%. Then the confidence level differences are 8%, 40%, and 50% in sequence, and the confidence density can be obtained therefrom.

[0110] If the confidence density is larger, it means that the difference between adjacent confidence levels is large, the confidence level distribution is steep, and a few experts are significantly better than other experts. Then the number of target vertical expert networks should be adjusted downward, which can reduce the waste of computing resources while avoiding generating results with low relevance; if the confidence density is larger, it means that the difference between adjacent confidence levels is small, the confidence level distribution is average, the results generated by each expert are highly relevant and relatively average, and more experts can be used as much as possible to improve the accuracy of result generation.

[0111] The present invention can optimize the configuration number of target vertical expert networks through the confidence density, improve the accuracy of result generation, and avoid waste of computing resources.

[0112] Step S143, adjust the target vertical expert network according to the dynamic processing information of the target vertical expert network or the correlation between the target vertical expert networks, where the dynamic information includes the running state and the processing state.

[0113] Specifically, it includes: successively judging whether the target vertical expert network is in a fault state or an occupied state according to the running state of the target vertical expert network; if it is, then removing the target vertical expert network and filling it in order with the vertical expert networks that have not been selected; if it is not in either state, then no removal is performed; determining the first expert network among a number of target vertical expert networks; determining the degree of relevance between the first expert network and the remaining target vertical expert networks based on a preset knowledge graph, and sorting the remaining target vertical expert networks according to the degree of relevance.

[0114] If the target vertical expert network is in a fault state or an occupied state, then remove the target vertical expert network and fill it in according to the order in the confidence distribution. For example, the confidence distributions of each vertical expert network are 98%, 90%, 50%, 0%. If the vertical expert network corresponding to 90% is in a fault state or an occupied state, then remove it and fill it with the vertical expert network corresponding to 50%, and use the vertical expert network with the highest confidence as the first expert network.

[0115] The preset knowledge graph can include knowledge in several fields. In the preset knowledge graph, the association between the field where the first expert network is located and the fields where the remaining target vertical expert networks are located can be found. Specifically, the more times the association between the field where the first expert network is located and the remaining target vertical expert networks, the higher the degree of relevance.

[0116] For example, if the first expert network is in the field of image processing and a certain target vertical expert network is in speech recognition, then "speech" and "image" can be used as keywords. Each time these two words appear in the same relationship, it is recorded as one association. Sort the target vertical expert networks according to the number of associations. It should be noted that all the target vertical expert networks after sorting need to participate in the processing of the subtask.

[0117] Step S144, obtaining a multi-dimensional processing result of the subtask based on a number of target vertical expert networks.

[0118] Obtaining a multi-dimensional processing result of the subtask based on a number of target vertical expert networks includes: processing the subtask based on a number of target vertical expert networks to obtain a first single-dimensional processing result of each target vertical expert network; determining the result weight corresponding to each target vertical expert network according to the sorting of the target vertical expert networks; obtaining a second single-dimensional processing result according to the result weight corresponding to each target vertical expert network and the corresponding first single-dimensional processing result; obtaining a multi-dimensional processing result of the subtask according to a number of second single-dimensional processing results.

[0119] The sub - tasks are processed by the target vertical expert network to obtain the corresponding one - dimensional processing first results. The result weights can be assigned according to the sorting. For example, the present invention provides fixed assignments of 70%, 20%, and 10%, that is, only a fixed number of assignments are taken, and the first three corresponding one - dimensional processing first results are selected. Of course, assignments can also be made according to other rules, as long as the result weights are positively correlated with the sorting.

[0120] The one - dimensional processing first results are converted into structured data (vector form) and weighted according to the corresponding result weights. Then, for different data types (such as text, image), feature - level or decision - level fusion can be used to obtain the multi - dimensional processing results of the sub - tasks.

[0121] The present invention first pre - sorts several target vertical expert networks through a preset knowledge graph and confidence distribution, and assigns result weights to each result according to the sorting results, so that the processing results corresponding to the sub - tasks are more in line with the actual situation and the answer hit rate is improved.

[0122] Step S15: Obtain the processing result of the total task of the target user according to several multi - dimensional processing results.

[0123] The multi - dimensional processing results corresponding to each sub - task are converted into vector form, and feature - level or decision - level fusion is used to obtain the processing result of the total task.

[0124] In summary, the present invention provides a user task processing method based on a multi-expert model. The method includes: obtaining multi-modal data of a target user and obtaining the total task of the target user according to the multi-modal data; decomposing the total task to obtain a number of sub-tasks; constructing a multi-expert model, where the multi-expert model includes a gating network unit and a vertical expert network cluster, and the vertical expert network cluster includes a number of vertical expert networks, each vertical expert network corresponds to a preset field and all vertical expert networks are open networks; for each sub-task, execute steps S141 - S144, including: step S141, controlling the gating network unit to determine the confidence levels of each vertical expert network for processing the sub-task; step S142, determining a number of target vertical expert networks according to the confidence levels of each vertical expert network; step S143, adjusting the target vertical expert networks according to the dynamic processing information of the target vertical expert networks or the relevance between the target vertical expert networks, where the dynamic information includes the running state and the processing state; step S144, obtaining a multi-dimensional processing result of the sub-task based on a number of target vertical expert networks; and obtaining a processing result of the total task of the target user according to a number of multi-dimensional processing results. The present invention combines the historical performance (accuracy and timeliness) of experts with the current task relevance by comprehensively considering the historical accuracy rate and the processing time limit compliance rate, avoiding single-index deviation, improving decision reliability, achieving a precise match between task requirements and expert capabilities through time limit and accuracy indicators, and ultimately taking into account the response speed and resource utilization while ensuring high precision. The present invention can optimize the configuration quantity of target vertical expert networks through confidence density, improve the accuracy of result generation, and avoid waste of computing resources. The present invention first pre-orders a number of target vertical expert networks through a preset knowledge graph and confidence distribution, and assigns weights to each result according to the sorting result, making the processing result corresponding to the sub-task more in line with the actual situation and improving the answer hit rate. The present invention integrates and splits the problem through a total (total task) - sub (sub-task) - total (total result) structure, and combines the results of the split processing as the total result, which can improve the hit rate of the result.

[0125] Based on the same inventive concept, the present invention provides a user task processing device based on a multi-expert model as shown in Figure 2 and the device includes:

[0126] An acquisition module 21, configured to obtain multi-modal data of a target user and obtain the total task of the target user according to the multi-modal data;

[0127] A decomposition module 22, configured to decompose the total task to obtain a number of sub-tasks;

[0128] A building module 23 for building a multi-expert model, where the multi-expert model includes a gating network unit and a vertical expert network cluster. The vertical expert network cluster includes a number of vertical expert networks, each vertical expert network corresponding to a preset field and all vertical expert networks being open networks;

[0129] A multi-dimensional processing module 24 for performing steps S141 - S144 for each sub-task, including: Step S141, controlling the gating network unit to determine the confidence levels of each vertical expert network for processing the sub-task; Step S142, determining a number of target vertical expert networks according to the confidence levels of each vertical expert network; Step S143, adjusting the target vertical expert networks according to the dynamic processing information of the target vertical expert networks or the relevance between the target vertical expert networks, where the dynamic information includes the running state and the processing state; Step S144, obtaining a multi-dimensional processing result of the sub-task based on a number of target vertical expert networks;

[0130] A result module 25 for obtaining a processing result of the total task of the target user according to a number of multi-dimensional processing results.

[0131] Based on the same inventive concept, the present invention also provides an electronic device as shown, including:

[0132] A processor;

[0133] A memory for storing instructions executable by the processor;

[0134] Wherein, the processor is configured to execute to implement the user task processing method based on the multi-expert model as provided above.

[0135] Based on the same inventive concept, the present invention also provides a non-transitory computer-readable storage medium. When the instructions in the storage medium are executed by the processor of the electronic device, the electronic device can execute to implement the user task processing method based on the multi-expert model as provided above.

[0136] Since the electronic device introduced in this embodiment is the electronic device used to implement the information processing method in the embodiment of the present invention, based on the information processing method introduced in the embodiment of the present invention, those skilled in the art can understand the specific implementation manners and various variations of the electronic device in this embodiment. Therefore, the specific implementation of how this electronic device implements the method in the embodiment of the present invention will not be described in detail here. As long as the electronic device used by those skilled in the art to implement the information processing method in the embodiment of the present invention falls within the scope of protection of the present invention.

[0137] Those skilled in the art will understand that the embodiments of the present invention can be provided as a method, a system, or a computer program product. Therefore, the present invention can take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present invention can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that contain computer-usable program code.

[0138] The present invention is described with reference to the flowcharts and / or block diagrams of methods, apparatuses (systems), and computer program products according to embodiments of the present invention. It should be understood that each flow and / or block in the flowchart and / or block diagram, and the combination of flows and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing devices to generate a machine, such that the instructions executed by the processor of the computer or other programmable data processing devices generate means for implementing the functions specified in one Figure 1 one flow or multiple flows and / or blocks Figure 1 one block or multiple blocks.

[0139] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to work in a specific manner, such that the instructions stored in the computer-readable memory generate a manufactured article including instruction means that implement the functions specified in one Figure 1 one flow or multiple flows and / or blocks Figure 1 one block or multiple blocks.

[0140] These computer program instructions can also be loaded onto a computer or other programmable data processing device, such that a series of operation steps are performed on the computer or other programmable device to generate a computer-implemented process, and thus the instructions executed on the computer or other programmable device provide steps for implementing the functions specified in one Figure 1 one flow or multiple flows and / or blocks Figure 1 one block or multiple blocks.

[0141] Although the preferred embodiments of the present invention have been described, those skilled in the art can make additional changes and modifications to these embodiments once they know the basic creative concepts. Therefore, the appended claims are intended to be construed as including the preferred embodiments and all changes and modifications that fall within the scope of the present invention.

[0142] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalent technologies, the present invention also intends to include these modifications and variations.

Claims

1. A non-transitory computer-readable storage medium, characterized in that: When the instructions in the storage medium are executed by a processor of an electronic device, the electronic device is enabled to execute a user task processing method based on a multi-expert model in the field of text, image or audio, including: Acquire multimodal data of a target user, and obtain a total task of the target user according to the multimodal data; Decomposing the overall task to obtain a number of subtasks; Constructing a multi-expert model, wherein the multi-expert model includes a gated network unit and a vertical expert network cluster, wherein the vertical expert network cluster includes a plurality of vertical expert networks, each vertical expert network corresponds to a preset field and all the vertical expert networks are open networks; For each subtask, steps S141-S144 are executed, including: Step S141, controlling the gated network unit to determine the confidence corresponding to each vertical expert network processing subtask; Step S142, determining a number of target vertical expert networks according to the confidence of each vertical expert network; Step S143, adjusting the target vertical expert network according to the dynamic processing information of the target vertical expert network or the correlation between the target vertical expert networks, wherein the dynamic information includes the running state and the processing state; Step S144, obtaining a multi-dimensional processing result of the subtask based on a plurality of target vertical expert networks; According to the plurality of multi-dimensional processing results, a processing result of the total task of the target user is obtained.

2. The non-transitory computer-readable storage medium of claim 1, wherein: Controlling the gated network unit to determine the confidence corresponding to each vertical expert network processing subtask includes: Perform feature preprocessing and dimension expansion on the initial features of the subtask to obtain expanded fusion features; The expanded fusion features are mapped to the same dimensional space as the number of vertical expert networks through the dimensionality reduction layer, and the confidence distribution is generated based on the Softmax function. The confidence distribution includes the confidence of each vertical expert network. The confidence distribution is adjusted based on the historical processing capability of the vertical expert network, where the historical processing capability includes the historical accuracy rate and the processing time limit compliance rate.

3. A non-transitory computer-readable storage medium as claimed in claim 2, characterized in that: Based on the historical processing capabilities of the vertical expert network, the confidence distribution is adjusted, including: in, After adjustment The confidence of a vertical expert network, Before adjustment The confidence of a vertical expert network, is the historical accuracy, It is the processing time limit achievement rate.

4. The non-transitory computer-readable storage medium of claim 1, wherein: According to the confidence of each vertical expert network, several target vertical expert networks are determined, including: According to the confidence distribution, the confidence density is determined; Setting a reference value for the number of expert networks, and determining the number G of target vertical expert networks according to the confidence density of the vertical expert network cluster and the reference value for the number of expert networks; The top G vertical expert networks are selected from the confidence distribution as the target vertical expert networks.

5. The non-transitory computer-readable storage medium of claim 4, wherein: include: in, is the confidence density, is the number of vertical expert networks; Also includes: in, is the number of target vertical expert networks and G is rounded up, is the reference value of the number of expert networks.

6. The non-transitory computer-readable storage medium of claim 1, wherein: According to the dynamic processing information of the target vertical expert network or the correlation between the target vertical expert networks, the target vertical expert network is adjusted, including: According to the operation status of the target vertical expert network, determine whether the target vertical expert network is in a fault state or an occupied state; If so, the target vertical expert network will be eliminated and filled with the vertical expert networks that have not been selected in order; if neither of them is in the same situation, no elimination will be performed; Among several target vertical expert networks, determining a first expert network; Based on the preset knowledge graph, the degree of correlation between the first expert network and the remaining target vertical expert networks is determined, and the remaining target vertical expert networks are sorted according to the degree of correlation.

7. The non-transitory computer-readable storage medium of claim 6, wherein: Based on several target vertical expert networks, the multi-dimensional processing results of this subtask are obtained, including: Based on a number of target vertical expert networks, the subtask is processed to obtain a first single-dimensional processing result of each target vertical expert network; According to the ranking of the target vertical expert networks, the result weights corresponding to each target vertical expert network are determined; According to the result weights corresponding to each target vertical expert network and the corresponding first result of the single-dimensional processing, a second result of the single-dimensional processing is obtained; According to the second results of the single-dimensional processing, a multi-dimensional processing result of the subtask is obtained.

8. A user task processing device based on a multi-expert model, characterized in that: The device comprises: An acquisition module, used to acquire multimodal data of a target user, and obtain a total task of the target user according to the multimodal data, wherein the multimodal data includes text data, image data or audio data; A decomposition module, used to decompose the overall task into a plurality of subtasks; A construction module is used to construct a multi-expert model, wherein the multi-expert model includes a gated network unit and a vertical expert network cluster, the vertical expert network cluster includes a number of vertical expert networks, each vertical expert network corresponds to a preset field and the vertical expert networks are all open networks; The multi-dimensional processing module is used to execute steps S141-S144 for each subtask, including: step S141, controlling the gated network unit to determine the confidence corresponding to each vertical expert network processing subtask; step S142, determining a plurality of target vertical expert networks according to the confidence of each vertical expert network; step S143, adjusting the target vertical expert network according to the dynamic processing information of the target vertical expert network or the correlation between the target vertical expert networks, wherein the dynamic information includes the running state and the processing state; step S144, obtaining the multi-dimensional processing result of the subtask based on the plurality of target vertical expert networks; The result module is used to obtain the processing result of the total task of the target user according to a plurality of multi-dimensional processing results.

Images