Inference method and computing device

Abstract

The invention discloses a reasoning method and computing equipment, relates to the technical field of artificial intelligence, and can provide personalized intelligent services for users. Specifically, in response to a first reasoning task input by a user, a reasoning result of the first reasoning task is determined, then personalized processing is performed on the reasoning result based on a user portrait to obtain a first personalized result, and the first personalized result is output. Wherein the matching degree of the first personalized result and the user portrait is greater than the matching degree of the reasoning result and the user portrait.

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence technology, and in particular to a reasoning method and computing device. Background Art

[0002] With the continuous development of big model technology, intelligent services based on big models, such as speech synthesis and image processing, are gaining increasing attention. These intelligent services leverage big model technology to provide users with a convenient and efficient experience. However, when using big model technology to provide intelligent services, the output often deviates from user expectations. Even for the same intelligent service, these individual differences lead to individual needs and expectations for each user.

[0003] Therefore, how to provide users with personalized intelligent services has become an urgent problem that needs to be solved. Summary of the Invention

[0004] The embodiments of the present application provide a reasoning method and a computing device that can provide users with personalized intelligent services.

[0005] To achieve the above objectives, the embodiments of the present application adopt the following technical solutions:

[0006] In a first aspect, a reasoning method is provided, which is applied to a computing device. The method includes: responding to a first reasoning task input by a user, determining a reasoning result of the first reasoning task, and then personalizing the reasoning result based on a user profile to obtain a first personalized result, and outputting the first personalized result.

[0007] Among them, the degree of matching between the first personalized result and the user portrait is greater than the degree of matching between the inference result and the user portrait.

[0008] Through the above technical solution, the inference results can be personalized to obtain personalized results that match the user profile (such as the first personalized result mentioned above). In this way, user needs can be better met and personalized intelligent services can be provided to users.

[0009] In an optional embodiment, a first personalized model may be deployed in the computing device, and the first personalized model is used to personalize the inference result based on the user profile. Based on this, the above method may further include: adjusting the first personalized model based on the personalization level of the first personalized result.

[0010] The personalization degree is used to indicate the matching degree between the first personalized result and the preference information.

[0011] In the above technical solution, after personalizing the inference results using the personalized model, the personalized model can be adjusted based on the degree of match between the personalized results and the user profile to optimize the performance of the personalized model and increase its degree of personalization. This ensures that the personalized results, after being personalized using the personalized model, match the user profile more closely, better meeting user needs.

[0012] In an optional embodiment, the method may further include obtaining a second personalized result corresponding to the second reasoning task and determining a second similarity between the first personalized result and the second personalized result. Accordingly, adjusting the first personalized model based on the degree of personalization of the first personalized result may specifically include adjusting the first personalized model based on the second similarity.

[0013] The second reasoning task refers to a historical reasoning task for which the first similarity with the first reasoning task is greater than the first similarity threshold and the first satisfaction is greater than the satisfaction threshold. The first satisfaction refers to the user's satisfaction with the second personalized result. The second similarity is positively correlated with the degree of personalization of the first personalized result.

[0014] The above technical solution provides a method for adjusting the personalized model, that is, the personalized model can be adjusted based on the similarity (such as the second similarity mentioned above) between the personalized answer (such as the first personalized answer) corresponding to the current reasoning task (such as the first reasoning task mentioned above) and the personalized answer (such as the second personalized answer) corresponding to the historical reasoning task (such as the second reasoning task mentioned above) similar to the current reasoning task, so as to effectively improve the feasibility of this solution.

[0015] Furthermore, the above approach allows fine-tuning of the first personalized model during its use, eliminating the need to spend significant time training it during its idle time. This not only saves significant computing and time resources, but also allows the first personalized model to further align with the user's profile during use, enabling better personalized services to be provided to users subsequently.

[0016] In an optional embodiment, the above-mentioned adjustment of the first personalized model based on the second similarity may specifically include: when the second similarity is greater than a second similarity threshold, fusing the first personalized model with the second personalized model; when the second similarity is less than a third similarity threshold, using newly added training samples to train the first personalized model.

[0017] The personalized processing capability of the second personalized model is lower than the personalized processing capability of the first personalized model. The third similarity threshold is lower than the second similarity threshold.

[0018] When the similarity between the personalized answer corresponding to the current reasoning task and the personalized answer corresponding to the historical reasoning task similar to the current reasoning task is too high (e.g., greater than the second similarity threshold), it indicates that the currently adopted personalized model (e.g., the first personalized model) has overfitting. After a personalized model (e.g., the second personalized model) with lower personalized processing capability than the currently adopted personalized model is fused with the currently adopted personalized model, the personalized capability of the currently adopted personalized model can be reduced to avoid subsequent overfitting.

[0019] When the similarity between the personalized answer corresponding to the current reasoning task and the personalized answer corresponding to the historical reasoning task similar to the current reasoning task is too low (such as less than the third similarity threshold), it indicates that the currently used personalized model (such as the first personalized model) is insufficiently trained. Using newly added training samples to train the currently used personalized model can effectively improve the personalization ability of the currently used personalized model.

[0020] In an optional implementation, the manner of fusing the first personalized model with the second personalized model may include any one of horizontal superposition, direct addition, weighted addition, and vertical superposition.

[0021] The above technical solution expands the method of fusing the first personalized model with the second personalized model, thereby effectively improving the compatibility of this application.

[0022] In an optional embodiment, the method may further include obtaining a second personalized result corresponding to the second reasoning task and determining a degree of difference between the first personalized result and the second personalized result. Accordingly, adjusting the first personalized model based on the degree of personalization of the first personalized result may specifically include adjusting the first personalized model based on the degree of difference.

[0023] The difference is negatively correlated with the personalization degree of the first personalized result, and the second reasoning task refers to a reasoning task in the historical reasoning task, in which the first similarity between the task and the first reasoning task is greater than the first similarity threshold.

[0024] The above technical solution provides a method for adjusting the personalized model, that is, the personalized model can be adjusted based on the difference between the personalized answer (such as the first personalized answer) corresponding to the current reasoning task (such as the first reasoning task) and the personalized answer (such as the second personalized answer) corresponding to the historical reasoning task similar to the current reasoning task (such as the second reasoning task), so as to effectively improve the feasibility of this solution.

[0025] In an optional embodiment, the above-mentioned adjustment of the first personalized model based on the difference may specifically include: when the difference is greater than a first difference threshold, using newly added training samples to train the first personalized model; when the difference is less than a second difference threshold, fusing the first personalized model with the second personalized model.

[0026] The personalized processing capability of the second personalized model is lower than the personalized processing capability of the first personalized model, and the second difference threshold is lower than the first difference threshold.

[0027] When the difference between the personalized answer corresponding to the current reasoning task and the personalized answer corresponding to the historical reasoning task similar to the current reasoning task is too low (such as less than the second difference threshold), it indicates that the currently adopted personalized model (such as the first personalized model) has overfitting. After a personalized model (such as the second personalized model) with lower personalized processing capability than the currently adopted personalized model is fused with the currently adopted personalized model, the personalized capability of the currently adopted personalized model can be reduced to avoid subsequent overfitting.

[0028] When the difference between the personalized answer corresponding to the current reasoning task and the personalized answer corresponding to the historical reasoning task similar to the current reasoning task is large (e.g., greater than a first difference threshold), it indicates that the currently used personalized model (e.g., the first personalized model) is insufficiently trained. Using newly added training samples to train the currently used personalized model can effectively improve the personalization ability of the currently used personalized model.

[0029] In an optional embodiment, the above-mentioned determination of the difference between the first personalized result and the second personalized result may specifically include: obtaining a first satisfaction level and a second satisfaction level, and determining the difference between the first personalized result and the second personalized result based on the second satisfaction level and the first satisfaction level.

[0030] The second satisfaction level refers to the user's satisfaction with the first personalized result, and the first satisfaction level refers to the user's satisfaction with the second personalized result.

[0031] The above technical solution describes the process of determining the difference between the first personalized result and the second personalized result, that is, the difference between the first personalized result and the second personalized result can be determined by the user's satisfaction with the first personalized result and the user's satisfaction with the second personalized result. In this way, the feasibility of this solution can be effectively improved.

[0032] In an optional embodiment, a corresponding relationship between similarity and satisfaction may be stored in the computing device. The obtaining of the second satisfaction may specifically include: determining a second similarity between the first personalized result and the second personalized result, and using the satisfaction corresponding to the second similarity as the second satisfaction from the corresponding relationship.

[0033] Through the above technical solution, the computing device can determine the second satisfaction level through the corresponding relationship stored in itself, thereby effectively improving the efficiency of the computing device in determining the second satisfaction level.

[0034] In a second aspect, an inference device is provided, comprising: functional units for executing any of the methods provided in the first aspect, wherein the actions performed by each functional unit are implemented via hardware or via hardware executing corresponding software implementations. For example, the inference device may include: an acquisition unit, a processing unit, and an output unit. The acquisition unit is configured to determine an inference result of a first inference task input by a user. The processing unit is configured to personalize the inference result based on a user profile to obtain a first personalized result. The output unit is configured to output the first personalized result.

[0035] In a third aspect, a computing device is provided, comprising: a processor and a memory, wherein the processor is connected to the memory, the memory is used to store computer-executable instructions, and the processor executes the computer-executable instructions stored in the memory, thereby implementing any one of the methods provided in the first aspect.

[0036] In a fourth aspect, a chip is provided, comprising: a processor and an interface circuit; the interface circuit is configured to receive code instructions and transmit them to the processor; and the processor is configured to run the code instructions to execute any one of the methods provided in the first aspect.

[0037] In a fifth aspect, a computer-readable storage medium is provided, which stores computer execution instructions. When the computer execution instructions are run on a computer, the computer executes any one of the methods provided in the first aspect.

[0038] In a sixth aspect, a computer program product is provided, comprising computer execution instructions, which, when executed on a computer, enable the computer to execute any one of the methods provided in the first aspect.

[0039] Among them, the technical effects brought about by any implementation method in the second to sixth aspects can refer to the technical effects brought about by different implementation methods in the first aspect, and will not be repeated here. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] Figure 1 A schematic diagram of the architecture of a communication system provided in an embodiment of the present application;

[0041] Figure 2 A schematic diagram of the structure of a computing device provided in an embodiment of the present application;

[0042] Figure 3 A flowchart of an inference method provided in an embodiment of the present application;

[0043] Figure 4 A schematic diagram of the structure of another computing device provided in an embodiment of the present application;

[0044] Figure 5 A schematic diagram of the structure of another computing device provided in an embodiment of the present application;

[0045] Figure 6 A schematic diagram of the structure of another computing device provided in an embodiment of the present application;

[0046] Figure 7 A schematic diagram of the structure of an inference device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0047] The technical solutions in the embodiments of the present application will be described below in conjunction with the drawings in the embodiments of the present application.

[0048] In the description of this application, unless otherwise specified, " / " indicates that the objects associated before and after are in an "or" relationship, for example, A / B can represent A or B; "and / or" in this application is merely a description of the association relationship of associated objects, indicating that three relationships may exist, for example, A and / or B can represent: A exists alone, A and B exist at the same time, and B exists alone, where A and B can be singular or plural.

[0049] Furthermore, in the description of this application, unless otherwise specified, "plurality" means two or more than two. "At least one of the following" or similar expressions refers to any combination of these items, including any combination of single or plural items. For example, at least one of a, b, or c can mean: a, b, c, ab, ac, bc, or abc, where a, b, and c can be single or plural.

[0050] In addition, in order to facilitate the clear description of the technical solutions of the embodiments of the present application, in the embodiments of the present application, words such as "first" and "second" are used to distinguish between identical or similar items with substantially the same functions and effects. Those skilled in the art will understand that words such as "first" and "second" do not limit the quantity and execution order, and words such as "first" and "second" do not necessarily limit differences. At the same time, in the embodiments of the present application, words such as "exemplary" or "for example" are used to indicate examples, illustrations or explanations. Any embodiment or design described as "exemplary" or "for example" in the embodiments of the present application should not be interpreted as being more preferred or more advantageous than other embodiments or design schemes. Specifically, the use of words such as "exemplary" or "for example" is intended to present related concepts in a concrete way for easy understanding.

[0051] First, the application scenarios of the embodiments of the present application are exemplarily introduced.

[0052] Embodiments of the present application provide a reasoning method applied to a computing device. The computing device may, in response to a first reasoning task input by a user, determine a reasoning result of the first reasoning task, then personalize the reasoning result based on a user profile to obtain a first personalized result, and output the first personalized result. The first personalized result may match the user profile to a greater degree than the reasoning result matches the user profile.

[0053] Through the above technical solution, the inference results can be personalized to obtain personalized results that match the user profile (such as the first personalized result mentioned above). In this way, user needs can be better met and personalized intelligent services can be provided to users.

[0054] The following is an exemplary introduction to the system architecture of the embodiment of the present application.

[0055] Figure 1 This is a schematic diagram of the architecture of a communication system provided in an embodiment of the present application. Figure 1 As shown, the communication system may include a terminal device 101 and a computing device 102. The terminal device 101 and the computing device 102 are in communication connection.

[0056] The terminal device 101 may also be referred to as user equipment (UE) or terminal equipment (TE). Exemplarily, the terminal device may include a personal digital assistant (PDA), an ultra-mobile personal computer (UMPC), a laptop computer, a netbook, a desktop computer, an all-in-one computer, a mobile phone, a tablet computer (pad), an in-vehicle device, or a wearable device.

[0057] Computing device 102 may be a network device. Network devices may include servers, etc. A server may be a single physical server, or two or more physical servers sharing different responsibilities and collaborating to implement various server functions, or a virtual server (also referred to as a virtual machine) running on a physical server. For example, the server may be a blade server, a high-density server, a rack server, or a tower server.

[0058] In an embodiment of the present application, when a user intends to use artificial intelligence (AI) for intelligent question-and-answer services, the user may input a corresponding reasoning task (also referred to as a question, such as the first reasoning task in the present application) into a terminal device 101 (such as a mobile phone) held by the user. After receiving the reasoning task input by the user, the terminal device 101 may send the reasoning task to the computing device 102. After receiving the reasoning task, the computing device may respond to the reasoning task, determine the reasoning result corresponding to the reasoning task, and personalize the reasoning result based on the user profile to obtain a personalized result that matches the user profile (also referred to as an answer, such as the first personalized result in the present application).

[0059] Afterwards, the computing device 102 may send the personalized result to the terminal device 101. After receiving the personalized result, the terminal device 101 may output the personalized result, for example, by displaying the personalized result on its own display screen or playing it through a speaker. The manner in which the terminal device 101 outputs the personalized result is not specifically limited herein.

[0060] Among them, the degree of matching between personalized results and user portraits is greater than the degree of matching between inference results and user portraits.

[0061] It should be noted that the embodiment of the present application does not limit the device form of the computing device 102. The system architecture of the computing device 102 provided in the embodiment of the present application is described below using a server as an example.

[0062] Figure 2 Schematic diagram of a computing device 102 provided in an embodiment of the present application. Figure 2 As shown, the computing device 102 includes a processor 202 and a memory 204. The processor 202 is connected to the memory 204 via a double data rate (DDR) bus 203. Here, the DDR bus 203 can also be replaced with other types of buses, and the embodiment of the application does not limit the bus type. In addition, the computing device 102 also includes various I / O devices, and the processor 202 can access these I / O devices 207 via a high-speed peripheral component interconnect express (PCIe) bus 205.

[0063] The processor 202 is the computing core and control core of the computing device 102. The processor 202 may include one or more processor cores 201. The processor 202 may be a very large-scale integrated circuit. An operating system and other software programs are installed in the processor 202, so that the processor 202 can access the memory 204 and various PCIe devices. It is understood that in the embodiment of the present invention, the core 201 in the processor 202 may be, for example, a central processing unit (CPU) or other application-specific integrated circuit (ASIC). The processor 202 may also be other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. In actual applications, the computing device 102 may also include multiple processors.

[0064] A memory controller is a bus circuit controller within computing device 102 that controls memory 204 and manages and schedules data transfers from memory 204 to core 201. The memory controller enables data exchange between memory 204 and core 201. The memory controller can be a separate chip connected to core 201 via the system bus.

[0065] Those skilled in the art will appreciate that the memory controller may be integrated into processor 202, embedded in the north bridge, or a separate memory controller chip. The embodiments of the present invention do not limit the specific location or form of the memory controller. In practical applications, the memory controller may control the necessary logic to write data to or read data from memory 204. Memory controller 204 may be a memory controller in a processor system such as a general-purpose processor, a dedicated accelerator, a GPU, an FPGA, or an embedded processor.

[0066] Memory 204 is the main memory of computing device 102. Memory 204 is typically used to store various running software in the operating system, input and output data, and information exchanged with external memory. To improve the access speed of processor 202, memory 204 needs to have a fast access speed. In traditional computer system architectures, dynamic random access memory (DRAM) is typically used as memory 204. Processor 202 can access memory 204 at high speed through a memory controller, performing read and write operations on any storage unit in memory 204. In addition to DRAM, memory 204 can also be other random access memories, such as static random access memory (SRAM). Memory 204 can also be read-only memory (ROM). For example, ROM can be programmable read-only memory (PROM) or erasable programmable read-only memory (EPROM). This embodiment does not limit the quantity or type of memory 204. In addition, the memory 204 can be configured to have a power-saving function. The power-saving function means that the data stored in the memory will not be lost when the system loses power and then powers on again. The memory 204 with the power-saving function is called a non-volatile memory.

[0067] Input / output (I / O) devices 207 refer to hardware that enables data transmission and can also be understood as devices that interface with an I / O interface. Common I / O devices include network cards, printers, keyboards, mice, and the like. All external storage devices, such as hard disks, floppy disks, and optical disks, can also serve as I / O devices. Processor 202 can access each IO device 207 via PCIe bus 205. It should be noted that PCIe bus 205 is only an example and can be replaced by other buses, such as a unified bus (UB) bus.

[0068] The baseboard management controller (BMC) 206 can perform firmware upgrades, manage the device's operating status, and troubleshoot problems even when the computing device 102 is powered off. The processor 202 can access the BMC 206 via the PCIe bus 205. The BMC 206 can also be connected to at least one sensor. The sensor can acquire status data from the computing device 102, including temperature, current, and voltage data. The type of status data is not specifically limited in this application. The BMC 206 communicates with the processor 202 via the PCIe bus or other bus types, for example, transmitting acquired status data to the processor 202 for processing. The BMC 206 can also maintain program code in memory, including upgrading or restoring it. The BMC 206 can also control the power supply circuit or clock circuit within the computing device 102. In summary, the BMC 206 can manage the computing device 102 using the above methods. However, the BMC 206 is an optional device. In some implementations, the processor 202 can communicate directly with the sensor to directly manage and maintain the computing device 102 .

[0069] It should be noted that the system architecture and application scenarios described in the embodiments of the present application are intended to more clearly illustrate the technical solutions of the embodiments of the present application, and do not constitute a limitation on the technical solutions provided in the embodiments of the present application. Ordinary technicians in this field can know that with the evolution of the system architecture and the emergence of new business scenarios, the technical solutions provided in the embodiments of the present application are also applicable to similar technical problems.

[0070] The following Figure 1 Taking the computing device shown as an example, the reasoning method provided in the embodiment of the present application is introduced in detail. Figure 3 A flowchart of a reasoning method provided in an embodiment of the present application is shown in FIG. Figure 3 As shown, the method includes S301-S303.

[0071] S301 : In response to a first reasoning task input by a user, determining a reasoning result of the first reasoning task.

[0072] Among them, the reasoning tasks (such as the first reasoning task and the second reasoning task in the embodiment of the present application) are used to indicate the needs of the user. The embodiment of the present application does not specifically limit the content of the reasoning task. In one example, the reasoning task can be text information input by the user, such as tomorrow's weather forecast, or text information such as how to use a rice cooker. In another example, the reasoning task can include image information input by the user. In another example, the reasoning task can be audio information input by the user, etc.

[0073] Specifically, taking the user's terminal device as a mobile phone as an example, an AI question-and-answer platform can be deployed in the mobile phone, and the AI question-and-answer platform is used to provide intelligent question-and-answer services. On this basis, when the user intends to use AI for intelligent question-and-answer services, he can start the AI question-and-answer platform on his mobile phone and log in to his account on the AI question-and-answer platform. After logging in, the user can enter the corresponding reasoning task (hereinafter referred to as the first reasoning task) on the AI question-and-answer platform. After the mobile phone receives the first reasoning task, it can send the first reasoning task to the computing device. The computing device receives the first reasoning task and can respond to the first reasoning task to determine the reasoning result corresponding to the first reasoning task.

[0074] In the embodiments of the present application, there is no specific limitation on the AI question-and-answer platform. In one example, the AI question-and-answer platform can be AI software installed in a terminal device that can be used to provide intelligent question-and-answer services. At this time, the process for the above-mentioned user to start the AI question-and-answer platform on his or her mobile phone is as follows: an icon of the AI software can be displayed on the mobile phone desktop, and the user can click on the icon of the AI software to start the AI software. In another example, the AI question-and-answer platform can be an AI website for providing intelligent question-and-answer services. At this time, the process for the above-mentioned user to start the AI question-and-answer platform on his or her mobile phone is as follows: the user can enter the URL of the AI website in the mobile phone browser to log in to the AI website.

[0075] For example, let's assume a user needs to know tomorrow's weather conditions. For example, if the AI Q&A platform is AI software, the user launches the AI software on their phone and enters a corresponding first reasoning task in the AI software's dialog box: tomorrow's weather conditions. After receiving the first reasoning task, the phone can send the first reasoning task to the computing device, which then determines tomorrow's weather conditions (i.e., the reasoning result).

[0076] In order to effectively improve the efficiency of determining the reasoning result of the first reasoning task, in an optional embodiment, a reasoning model can be deployed in the computing device, and the reasoning model can be used to reason the first reasoning task input by the user to obtain the reasoning result of the first reasoning task.

[0077] The embodiments of this application do not specifically limit the inference model. For example, the inference model can be a large language model (LLM), a multimodal question-answering model, an embedding model, etc., built based on a transformer network, a recursive neural network (RNN), or a convolutional neural network (CNN).

[0078] Specifically, after receiving the first reasoning task, the computing device may input the first reasoning task into the reasoning model, and the embedding layer of the reasoning model may encode the first reasoning task to obtain the reasoning result of the first reasoning task.

[0079] The specific reasoning model performs reasoning on the first reasoning task and the process of obtaining the reasoning result can be referred to related technologies and will not be described in detail here.

[0080] For example, if the reasoning model is an LLM and the first reasoning task is "What will the weather be tomorrow?", the computing device can input "What will the weather be tomorrow?" into the LLM. The LLM has language generation, logical reasoning, and resource integration capabilities. Accordingly, it can reason on the reasoning task "What will the weather be tomorrow?" input by the user and obtain tomorrow's weather conditions.

[0081] S302: Based on the user portrait, personalize the inference result to obtain a first personalized result.

[0082] S303: Output a first personalized result.

[0083] User profiles are used to reflect user preferences and personal information. User preferences may include, but are not limited to, preferred communication methods and topics of interest, while personal information may include, but is not limited to, occupation and address.

[0084] The degree of match between the first personalized result and the user portrait is greater than the degree of match between the inference result and the user portrait.

[0085] In the embodiments of the present application, there is no specific limitation on the method for characterizing the degree of match between the first personalized result and the user portrait, and the degree of match between the inference result and the user portrait. In one example, the degree of match between the first personalized result and the user portrait can be characterized by the user's satisfaction with the first personalized result (hereinafter referred to as the second satisfaction), that is: the higher the second satisfaction, the higher the degree of match between the first personalized result and the user portrait. Correspondingly, the degree of match between the inference result and the user portrait can also be characterized by the user's satisfaction with the inference result, that is: the higher the user's satisfaction with the inference result, the higher the degree of match between the inference result and the user portrait. At this time, the degree of match between the first personalized result and the user portrait is greater than the degree of match between the inference result and the user portrait can be replaced by: the user's satisfaction with the first personalized result is greater than the user's satisfaction with the inference result.

[0086] The embodiment of the present application does not specifically limit the representation method of satisfaction (such as the second satisfaction level). In one example, satisfaction can be represented in the form of text, such as: dissatisfied, relatively dissatisfied, relatively satisfied, satisfied, and very satisfied. In another example, satisfaction can be represented in the form of a score, such as 1 point for dissatisfied, 2 points for relatively dissatisfied, 3 points for relatively satisfied, 4 points for satisfied, and 5 points for very satisfied.

[0087] In addition, it should be noted that the user's satisfaction with the first personalized result and the user's satisfaction with the inference result can be user feedback or prediction by the computing device based on the user portrait. The embodiments of this application do not make specific limitations on this.

[0088] In the case where the user's satisfaction with the first personalized result and the user's satisfaction with the inference result can be user feedback, the embodiment of the present application does not specifically limit the way in which the user feedbacks satisfaction. Taking the user feedback on the satisfaction with the first personalized result as an example, in one example, an input box can be displayed on the AI question-and-answer platform. After the user sees the first personalized result on the AI question-and-answer platform, the user can enter his or her satisfaction with the first personalized result in the input box, such as satisfaction in text form such as satisfied, dissatisfied, or satisfaction in the form of scores such as 4 points or 1 point. In another example, a satisfaction control (such as a like control, a dislike control, or a score control, etc.) can be displayed on the AI question-and-answer platform. After the user sees the first personalized result on the AI question-and-answer platform, the user can operate different controls to feedback his or her satisfaction with the first personalized result. For example, assuming that the user is relatively satisfied with the first personalized result, the user can operate the like control. For another example, assuming that the user is dissatisfied with the first personalized result, the user can operate the dislike control.

[0089] Specifically, the computing device may store a correspondence between a user identifier and a user profile. Accordingly, when the terminal device sends the first inference task to the computing device, it may also send the user identifier to the computing device. Based on the user identifier, the computing device may determine the user profile corresponding to the user identifier from the correspondence stored in the computing device. After obtaining the user profile, the computing device may personalize the inference result based on the user profile to obtain a personalized inference result (i.e., a first personalized result) that matches the user profile. The computing device may then send the first personalized result to the terminal device, which may then display the first personalized result.

[0090] The embodiment of the present application does not specifically limit the user identifier. For example, the user identifier can be the user's login account, or the user's mobile phone number, name, nickname, etc.

[0091] For example, if the user profile obtained by the computing device indicates that the user prefers short answers, and the inference result obtained above is "Tomorrow's weather will be sunny and then cloudy. It is recommended that the user wear a coat and keep warm," the computing device can simplify the inference result based on the user profile to obtain a first personalized result containing "Sunny and then cloudy." The computing device can send "Sunny and then cloudy" to the terminal device, and the terminal device can display "Sunny and then cloudy."

[0092] Through the above technical solution, the inference results can be personalized to obtain the first personalized result that matches the user profile. In this way, user needs can be better met and personalized intelligent services can be provided to users.

[0093] In order to speed up the efficiency of the computing device in personalizing the inference results, in an optional embodiment, a personalized model (hereinafter referred to as the first personalized model) can be deployed in the computing device, and the first personalized model can be used to personalize the inference results based on the user portrait.

[0094] As can be seen from the above description, a reasoning model can also be deployed in the computing device. The embodiment of the present application does not specifically limit the deployment method of the reasoning model and the first personalized model deployed in the computing device. In one example, the reasoning model and the first personalized model can be deployed in series, such as Figure 4 In another example, the inference model and the first personalized model can be deployed in an embedded manner, such as Figure 5 As shown, the first personalized model can be embedded in the hidden layer of the inference model. In this case, the first personalized model can also be called a hidden layer model.

[0095] The hidden layer refers to the layer in the inference model other than the input layer and the output layer. The main function of the hidden layer is to extract the features of the inference task input by the user and perform nonlinear transformations so that the inference model can understand the user's intention.

[0096] On the basis of deploying the first personalized model in the computing device, in order to optimize the performance of the personalized model and improve the personalization level of the personalized model, the computing device may further adjust the first personalized model based on the personalization level of the first personalized result.

[0097] The personalization degree is used to indicate the degree of matching between the first personalization result and the user portrait.

[0098] Specifically, the computing device may determine, from historical reasoning tasks, a personalized result (hereinafter referred to as the second personalized result) corresponding to a reasoning task (hereinafter referred to as the second reasoning task) whose similarity to the first reasoning task (hereinafter referred to as the first similarity) is greater than a similarity threshold (hereinafter referred to as the first similarity threshold) and whose first satisfaction level is greater than the satisfaction threshold. The computing device may then determine the degree of personalization of the first personalized result based on the first personalized result corresponding to the first reasoning task and the second personalized result corresponding to the second reasoning task, and adjust the first personalized model based on the degree of personalization of the first personalized result.

[0099] The first satisfaction level refers to the user's satisfaction with the second personalized result.

[0100] In one example, the computing device may include Figure 4 or Figure 5 The judgment model shown (also referred to as a judgment module, a decision module or a decision model) is used to determine the personalization degree of the first personalized result and adjust the first personalized model based on the personalization degree of the first personalized result.

[0101] The above-mentioned historical reasoning tasks refer to the reasoning tasks entered by the user before entering the first reasoning task. Different users correspond to different historical reasoning tasks. For example, if the user entering the first reasoning task is user A, then the historical reasoning tasks refer to the reasoning tasks entered by user A before entering the first reasoning task. If the user entering the first reasoning task is user B, then the historical reasoning tasks refer to the reasoning tasks entered by user B before entering the first reasoning task.

[0102] In one example, the computing device may obtain the historical reasoning task in the following manner: the computing device may store the historical reasoning task corresponding to the user identifier. After obtaining the user identifier in the above manner, the computing device may obtain the corresponding historical reasoning task based on the user identifier.

[0103] The embodiment of the present application does not specifically limit the method for determining similarity (such as the first similarity in the embodiment of the present application). For example, the computing device may determine the similarity through Levenshtein distance, longest common subsequence (LCS), or cosine similarity.

[0104] For example, it is assumed that the computing device determines the similarity between task A and task B (i.e., the first similarity) by cosine similarity, wherein task A is the first reasoning task and task B is the historical reasoning task. The computing device can perform word segmentation processing on task A and task B respectively to obtain task A after word segmentation processing and task B after word segmentation processing. The computing device can merge the words contained in task A after word segmentation processing and the words contained in task B after word segmentation processing to obtain a vocabulary. Afterwards, the computing device can generate a vector corresponding to task A based on the frequency of occurrence of each word in the vocabulary in task A, and generate a vector corresponding to task B based on the frequency of occurrence of each word in the vocabulary in task B. The computing device can determine the similarity between task A and task B based on the vector corresponding to task A and the vector corresponding to task B.

[0105] It can be understood that when the similarity between the first personalized result and the second personalized result (hereinafter referred to as the second similarity) is higher, or when the difference between the first personalized result and the second personalized result is lower, the degree of personalization of the first personalized result is higher. When the second similarity is lower, or when the difference between the first personalized result and the second personalized result is higher, the degree of personalization of the first personalized result is lower. Therefore, the second similarity is positively correlated with the degree of personalization of the first personalized result, and the difference between the first personalized result and the second personalized result is negatively correlated with the degree of personalization of the first personalized result.

[0106] The embodiments of the present application do not specifically limit the method for representing the difference between the first personalized result and the second personalized result. For example, the difference between the first personalized result and the second personalized result can be represented by a loss value or a percentage.

[0107] The embodiments of this application do not specifically limit the method for representing the degree of personalization of the first personalized result. For example, the degree of personalization of the first personalized result can be represented by the similarity between the first personalized result and the second personalized result (hereinafter referred to as the second similarity). For another example, the degree of personalization of the first personalized result can be represented by the difference between the first personalized result and the second personalized result.

[0108] The following describes the process of adjusting the first personalized model in two cases: A, characterizing the degree of personalization of the first personalized result by the second similarity; and B, characterizing the degree of personalization of the first personalized result by the difference.

[0109] A. Characterize the personalization degree of the first personalized result through the second similarity.

[0110] In this embodiment, after the computing device determines the second similarity, the above-mentioned adjustment of the first personalized model based on the personalization degree of the first personalized result can be replaced by adjusting the first personalized model based on the second similarity.

[0111] Specifically, to prevent the first personalized model from overfitting or undertraining, the computing device may store two similarity thresholds, namely a second similarity threshold and a third similarity threshold, with the second similarity threshold being greater than the third similarity threshold. The second similarity threshold is used to determine whether the first personalized model is overfitting, and the third similarity threshold is used to determine whether the first personalized model is undertrained.

[0112] Based on this, after obtaining the second similarity, the computing device may compare the second similarity with a second similarity threshold and a third similarity threshold, respectively. If the second similarity is greater than the second similarity threshold, it indicates that the first personalized model is overfitting. The computing device may fuse the first personalized model with a personalized model (hereinafter referred to as the second personalized model) having a lower personalized processing capability than the first personalized model, thereby reducing the personalized capability of the first personalized model.

[0113] When the second similarity is less than the third similarity threshold, it indicates that the first personalized model has an insufficient training problem. The computing device can add training samples and use the newly added training samples to train the first personalized model to improve the personalization ability of the first personalized model.

[0114] When the second similarity is greater than the third similarity threshold and less than the second similarity threshold, it indicates that the personalization capability of the first personalized model is in an ideal state, and the computing device may not adjust the first personalized model.

[0115] The embodiments of the present application do not specifically limit the situation where the second similarity is equal to the third similarity threshold, or equal to the second similarity threshold. For example, when the second similarity is equal to the third similarity threshold, the computing device may use the newly added training samples to train the first personalized model, or the computing device may not adjust the first personalized model. For another example, when the second similarity is equal to the second similarity threshold, the computing device may fuse the first personalized model with the second personalized model, or the computing device may not adjust the first personalized model.

[0116] The personalized processing capability of a personalized model (such as the first personalized model or the second personalized model) is positively correlated with user satisfaction with the personalized results output by the personalized model (such as the first personalized result). That is, the higher the user's satisfaction with the personalized results output by the personalized model, the higher the personalized processing capability of the personalized model. The lower the user's satisfaction with the personalized results output by the personalized model, the lower the personalized processing capability of the personalized model. Therefore, the statement that the personalized processing capability of the second personalized model is lower than that of the first personalized model can be replaced by the statement that user satisfaction with the personalized results output by the second personalized model is lower than user satisfaction with the personalized results output by the first personalized model.

[0117] For example, consider a case where the first reasoning task is Question A, the first personalized result is Answer B, the second reasoning task is Question 1, the second personalized result is Answer 2, the second similarity threshold is 80%, and the third similarity threshold is 30%. After determining the second similarity between the first and second personalized results, the computing device may fuse the first and second personalized models if the second similarity is greater than 80%. If the second similarity is less than 30%, additional training samples may be added and used to train the first personalized model.

[0118] In an optional embodiment, the computing device may further include Figure 4 or Figure 5 The fusion model shown may include a second personalized model, and the fusion model is used to fuse the first personalized model with the second personalized model.

[0119] In the embodiment of the present application, there is no specific limitation on the method of fusing the first personalized model and the second personalized model. For example, the first personalized model and the second personalized model can be fused by any of horizontal superposition, direct addition, weighted addition, and vertical superposition.

[0120] Horizontal stacking means placing the same type of layers (such as convolutional layers, fully connected layers, and hidden layers) of the first and second personalized models side by side, allowing them to process the inference results output by the inference models in parallel to obtain the corresponding personalized results. The personalized results output by the first and second personalized models can then be merged to obtain the final personalized result.

[0121] Direct addition means directly adding the personalized result output by the first personalized model and the personalized result output by the second personalized model to obtain a final personalized result.

[0122] Weighted addition means that the first and second personalized models are assigned different weights. After obtaining the personalized results output by the first and second personalized models, the personalized results output by the first and second personalized models can be weighted added together to obtain the final personalized result.

[0123] Vertical stacking means that the first and second personalized models are connected in series. In one example, the personalized results output by the first personalized model can be used as the input to the second personalized model. The second personalized model processes the personalized results output by the first personalized model to obtain the final personalized results. In another example, the personalized results output by the second personalized model can be used as the input to the first personalized model. The first personalized model processes the personalized results output by the second personalized model to obtain the final personalized results.

[0124] B. Characterize the personalization degree of the first personalized result through the difference degree.

[0125] In this embodiment, after the computing device determines the difference between the first personalized result and the second personalized result, the above-mentioned adjustment of the first personalized model based on the personalization degree of the first personalized result can be replaced by: adjusting the first personalized model based on the difference.

[0126] The embodiments of the present application do not specifically limit the manner in which the computing device determines the degree of difference between the first personalized result and the second personalized result. For example, the degree of difference between the first personalized result and the second personalized result may be determined using a cross-entropy function, or using a longest common subsequence (LCS) algorithm.

[0127] Specifically, to prevent the first personalized model from overfitting or undertraining, the computing device may store two difference thresholds: a first difference threshold and a second difference threshold, wherein the second difference threshold is smaller than the first difference threshold. The second difference threshold is used to determine whether the first personalized model is overfitting, and the first difference threshold is used to determine whether the first personalized model is undertrained.

[0128] On this basis, after obtaining the difference between the first personalized result and the second personalized result, the computing device may compare the difference with a first difference threshold and a second difference threshold, respectively. If the difference is less than the second difference threshold, it indicates that the first personalized model is overfitting. The computing device may then fuse the first personalized model with the second personalized model to reduce the personalization capability of the first personalized model.

[0129] When the difference is greater than the first difference threshold, it indicates that the first personalized model has an insufficient training problem. The computing device can add training samples and use the newly added training samples to train the first personalized model to improve the personalization ability of the first personalized model.

[0130] When the second similarity is greater than the second difference threshold and less than the first difference threshold, it indicates that the personalization capability of the first personalized model is in an ideal state, and the computing device may not adjust the first personalized model.

[0131] The embodiments of the present application do not specifically limit the situation where the difference between the first personalized result and the second personalized result is equal to the first difference threshold, or equal to the second difference threshold. For example, if the difference between the first personalized result and the second personalized result is equal to the second difference threshold, the computing device may use the newly added training samples to train the first personalized model, or the computing device may not adjust the first personalized model. For another example, if the difference between the first personalized result and the second personalized result is equal to the first difference threshold, the computing device may fuse the first personalized model with the second personalized model, or the computing device may not adjust the first personalized model.

[0132] For example, consider a first reasoning task of Question A, a first personalized result of Answer B, a second reasoning task of Question 1, a second personalized result of Answer 2, a first difference threshold of 80%, and a second difference threshold of 30%. After determining the difference between the first and second personalized results, the computing device can add training samples if the difference is greater than 80% and use the newly added training samples to train the first personalized model. If the second similarity is less than 30%, the first and second personalized models are fused.

[0133] The above technical solution shows that when the first personalized model is detected to be overfitting, a personalized model with lower personalized processing capabilities (i.e., a second personalized model) can be fused with the first personalized model to reduce the personalization capabilities of the first personalized model and avoid subsequent overfitting. When the first personalized model is detected to be undertrained, the currently used personalized model can be trained with newly added training samples to effectively improve the personalization capabilities of the currently used personalized model.

[0134] Furthermore, the above approach allows adjustments to be made to the first personalized model during its use, eliminating the need to waste significant time training the model during its idle time. This not only saves significant computing and time resources, but also allows the first personalized model to further align with the user's profile during use, enabling better personalized services to be provided to the user subsequently.

[0135] In an optional embodiment, the computing device determines the degree of difference between the first personalized result and the second personalized result, which may specifically include: obtaining the user's satisfaction with the first personalized result (hereinafter referred to as the second satisfaction), and the user's satisfaction with the second personalized result (hereinafter referred to as the first satisfaction). The computing device may determine the degree of difference between the first personalized result and the second personalized result based on the second satisfaction and the first satisfaction.

[0136] Specifically, the computing device may store user satisfaction scores for personalized results corresponding to multiple historical reasoning tasks. Accordingly, the computing device may retrieve a first satisfaction score from the stored satisfaction scores and determine a second satisfaction score. The computing device may then compare the second satisfaction score with the first satisfaction score to determine the difference between the first personalized result and the second personalized result.

[0137] The embodiments of the present application do not specifically limit the manner in which the computing device determines the second satisfaction level. In one example, the second satisfaction level may be user feedback. Specifically, after the computing device outputs the first personalized result to the user, the user may provide content feedback based on the first personalized result to characterize his or her satisfaction with the first personalized result, i.e., the second satisfaction level. In another example, a correspondence between similarity and satisfaction level may be stored in the computing device. Accordingly, after the computing device determines the second similarity between the first personalized result and the second personalized result, it may use the satisfaction level corresponding to the second similarity from the correspondence stored in itself as the second satisfaction level. In another example, the second satisfaction level may be determined based on a combination of the satisfaction level fed back by the user and the satisfaction level determined by the computing device from the correspondence stored in itself.

[0138] The following describes the process of determining the second satisfaction level by the computing device, taking the second satisfaction level as an example, which is determined based on the satisfaction level fed back by the user and the satisfaction level determined by the computing device from the corresponding relationships stored in the computing device. After the computing device determines the satisfaction level corresponding to the second similarity (hereinafter referred to as satisfaction level 1) from the corresponding relationships stored in the computing device and receives the satisfaction level fed back by the user (hereinafter referred to as satisfaction level 2), it can normalize satisfaction level 1 and satisfaction level 2 to obtain normalized satisfaction level 1 and normalized satisfaction level 2. Thereafter, the computing device can obtain a comprehensive satisfaction level, i.e., the second satisfaction level, based on the weight corresponding to satisfaction level 1, the weight corresponding to satisfaction level 2, the normalized satisfaction level 1, and the normalized satisfaction level 2.

[0139] The embodiment of the present application does not specifically limit the normalization processing method. For example, the normalization processing can be performed using minimum-maximum normalization, decimal calibration normalization, or standard deviation normalization.

[0140] For example, assuming that both satisfaction 1 and satisfaction 2 are represented in the form of scores, and satisfaction 1 is 3 points, satisfaction 2 is 5 points, the weight corresponding to satisfaction 1 is 30%, the weight corresponding to satisfaction 2 is 70%, the user feedback satisfaction is "satisfied", the normalized satisfaction 1 is 0.3, and the normalized satisfaction 2 is 0.5, then the computing device can determine that the second satisfaction is 0.3*30%+0.5*70%=0.44.

[0141] In the case where the second satisfaction level may be user feedback, in an optional embodiment, the computing device may further include: Figure 6 The feedback module shown is used to receive a second satisfaction level fed back by the user.

[0142] In an optional embodiment, in order to better train the first personalized model and make the personalization capabilities of the first personalized model more consistent with the user portrait, the computing device can also obtain the user's satisfaction with the first personalized result (i.e., the second satisfaction), and use the second satisfaction, the first reasoning task and the first personalized result to train the first personalized model.

[0143] Specifically, after the computing device obtains the second satisfaction level, the second satisfaction level, the first reasoning task, and the first personalized result can be input into the first personalized model, and the first personalized model can be trained using the second satisfaction level, the first reasoning task, and the first personalized result.

[0144] In an optional embodiment, the computing device may include Figure 6 The reward module shown is configured to normalize the second satisfaction score and send the normalized satisfaction score, the first reasoning task, and the first personalized result to the first personalized model. After receiving the normalized satisfaction score, the first reasoning task, and the first personalized result, the first personalized model can use the normalized satisfaction score, the first reasoning task, and the first personalized result for training.

[0145] In an optional embodiment, before using the first personalized model, the computing device may further use data for characterizing the user portrait (hereinafter referred to as user portrait data) to train the initial model to obtain the first personalized model.

[0146] The embodiment of the present application does not specifically limit the user portrait data. For example, the user portrait data may include at least one of the user's chat records, work summaries, report materials, and other information.

[0147] It is understandable that different users may have different corresponding user portrait data. For example, the user portrait data of a painter may include the painter's image works, and the user portrait data of a writer may include the writer's literary works, etc.

[0148] Specifically, the user's terminal device may store user profile data for the user. The computing device may send user profile data acquisition requests to the terminal devices of different users. Upon receiving the user profile data acquisition request sent by the computing device, each terminal device may acquire the user profile data stored therein and send the user profile data to the computing device. After receiving the user profile data sent by the different terminal devices, the computing device may encode the user profile data using a preset encoding method and train an initial model based on the encoded user profile data to obtain a first personalized model.

[0149] In the embodiments of the present application, the preset encoding method is not specifically limited. For example, the preset encoding method may be multimodal encoding or word embedding encoding, etc. Take the preset encoding method as an example of multimodal encoding. The user portrait data may contain data of multiple modalities, such as text modality, audio modality, image modality, and video modality. The computing device may use multimodal encoding to map data of different modalities in the user portrait data into a unified feature space, that is, converting data of different modalities into vectors. Afterwards, the computing device may input the converted vector into the initial model for training to obtain a first personalized model.

[0150] It is understood that the computing device trains the initial model to obtain a first personalized model that is universal, meaning it can be applied to every user. Subsequently, for each user, the computing device can continuously fine-tune the first personalized model during interactions with that user based on the degree of match between the personalized results output by the first personalized model and the user profile of that user, thereby optimizing the performance of the personalized model and making it more tailored to the user's needs, thereby providing better services to the user.

[0151] To prevent the leakage of a user's private data when a computing device acquires the user's personalized data, in an optional embodiment, the process of the user's terminal device acquiring the user's personalized data may include: the user's terminal device may display a prompt message, and the prompt message is used to indicate the acquisition of personalized data. The user may select their non-private data from the data stored on the terminal device as their personalized data, and the terminal device may use the selected data as the user's personalized data in response to the user's selection operation.

[0152] In an optional embodiment, the computing device may process the user's personalized data using a preset data enhancement method before using the user's personalized data to train the initial model, so as to increase the amount of the user's personalized data.

[0153] In the embodiments of the present application, the preset data augmentation method is not specifically limited. For example, the preset data augmentation method can be any one of instruction data generation, reverse generation, label reversal, data synthesis, error analysis, or large model data augmentation.

[0154] Specifically, taking the preset data enhancement method of large model data enhancement as an example, the computing device can generate a corresponding prompt based on the user's personalized data, and input the prompt into the large model data enhancement to obtain data similar in style to the user's personalized data, thereby increasing the data volume of the user's personalized data.

[0155] In an optional embodiment, the computing device may include Figure 6 The data enhancement module shown is used to process the user's personalized data using a preset data enhancement method.

[0156] Through the above technical solution, the user's personalized data can be enhanced to increase the data volume of the user's personalized data. In this way, the first personalized model trained using the user's personalized data after data enhancement can be more in line with the user portrait, and better personalized intelligent services can be provided to the user.

[0157] The above mainly introduces the solution provided by the embodiment of the present application from the perspective of method. In order to realize the above functions, the reasoning device includes a hardware structure and / or software module corresponding to the execution of each function. It should be easy for those skilled in the art to realize that, in combination with the units and algorithm steps of each example described in the embodiments disclosed herein, the present application can be implemented in the form of hardware or a combination of hardware and computer software. Whether a function is executed in the form of hardware or computer software driving hardware depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0158] In the embodiment of the present application, the functional modules of the reasoning device can be divided according to the above method. For example, the reasoning device can include functional modules corresponding to the functional divisions, or two or more functions can be integrated into one processing module. The above-mentioned integrated modules can be implemented in the form of hardware or in the form of software functional modules. It should be noted that the division of modules in the embodiment of the present application is schematic and is only a logical functional division. There may be other division methods in actual implementation.

[0159] For example, Figure 7A schematic diagram of a possible structure of the inference device involved in the above embodiments is shown. The device includes an acquisition unit 701, a processing unit 702, and an output unit 703. Acquisition unit 701 is configured to determine an inference result for a first inference task input by a user. Processing unit 702 is configured to personalize the inference result based on the user profile to obtain a first personalized result. Output unit 703 is configured to output the first personalized result.

[0160] Optionally, a first personalized model may be deployed in the computing device, and the first personalized model is used to personalize the inference result based on the user profile. Furthermore, the inference device may further include an adjustment unit, specifically configured to adjust the first personalized model based on the degree of personalization of the first personalized result.

[0161] Optionally, the acquisition unit 701 is further configured to: acquire a second personalized result corresponding to the second reasoning task, and determine a second similarity between the first personalized result and the second personalized result. Accordingly, the adjustment unit is specifically configured to: adjust the first personalized model based on the second similarity.

[0162] The second reasoning task refers to a historical reasoning task for which the first similarity with the first reasoning task is greater than a first similarity threshold and the first satisfaction is greater than a satisfaction threshold. The first satisfaction refers to the user's satisfaction with the second personalized result. The second similarity is positively correlated with the degree of personalization of the first personalized result.

[0163] Optionally, the adjustment unit is specifically configured to: when the second similarity is greater than a second similarity threshold, fuse the first personalized model with the second personalized model; when the second similarity is less than a third similarity threshold, train the first personalized model using newly added training samples.

[0164] The personalized processing capability of the second personalized model is lower than the personalized processing capability of the first personalized model. The third similarity threshold is lower than the second similarity threshold.

[0165] Optionally, the manner of fusing the first personalized model with the second personalized model may include at least one of horizontal superposition, direct addition, weighted addition, and vertical superposition.

[0166] Optionally, the acquisition unit 701 is further configured to: acquire a second personalized result corresponding to the second reasoning task, and determine the difference between the first personalized result and the second personalized result. Accordingly, the adjustment unit is specifically configured to: adjust the first personalized model based on the difference.

[0167] The difference is negatively correlated with the personalization degree of the first personalized result, and the second reasoning task refers to a reasoning task in the historical reasoning task, in which the first similarity between the task and the first reasoning task is greater than the first similarity threshold.

[0168] Optionally, the adjustment unit is specifically configured to: when the difference is greater than a first difference threshold, use newly added training samples to train the first personalized model; when the difference is less than a second difference threshold, fuse the first personalized model with the second personalized model.

[0169] The personalized processing capability of the second personalized model is lower than the personalized processing capability of the first personalized model, and the second difference threshold is lower than the first difference threshold.

[0170] Optionally, the acquiring unit 701 is further configured to: acquire a first satisfaction level and a second satisfaction level, and determine a difference between the first personalized result and the second personalized result based on the first satisfaction level and the second satisfaction level.

[0171] The second satisfaction level refers to the user's satisfaction with the first personalized result, and the first satisfaction level refers to the user's satisfaction with the second personalized result.

[0172] Optionally, a correspondence between similarity and satisfaction may be stored in the computing device, and the acquisition unit 701 is specifically configured to determine a second similarity between the first personalized result and the second personalized result, and from the correspondence, take the satisfaction corresponding to the second similarity as the second satisfaction.

[0173] For the detailed description of the above optional methods, please refer to the above method embodiments, which will not be repeated here. In addition, the explanation of any of the above-mentioned reasoning devices and the description of the beneficial effects can refer to the above-mentioned corresponding method embodiments, which will not be repeated here.

[0174] The present application also provides a computing device, which includes a processor and a memory, the processor being connected to the memory, the memory storing computer-executable instructions, and the processor implementing the data processing method in the above embodiment when executing the computer-executable instructions. The present application does not impose any restrictions on the specific form of the computing device. For example, the computing device can be a terminal device or a network device. The terminal device can be referred to as: terminal, user equipment (UE), terminal device, access terminal, user unit, user station, mobile station, remote station, remote terminal, mobile device, user terminal, wireless communication device, user agent or user device, etc. The terminal device can be a mobile phone, augmented reality (AR) device, virtual reality (VR) device, tablet computer, laptop computer, ultra-mobile personal computer (UMPC), netbook, personal digital assistant (PDA), etc. The network device can be a server, etc. The server can be a physical or logical server, or two or more physical or logical servers sharing different responsibilities and cooperating with each other to implement the various functions of the server.

[0175] An embodiment of the present application further provides a computer-readable storage medium having a computer program stored thereon. When the computer program is run on a computer, the computer is caused to execute the method executed by any one of the computing devices provided above.

[0176] For explanations of the relevant contents and descriptions of the beneficial effects of any of the computer-readable storage media provided above, reference may be made to the corresponding embodiments described above, and no further details will be given here.

[0177] The embodiment of the present application also provides a chip. The chip integrates a control circuit and one or more ports for implementing the functions of the above-mentioned computing device. Optionally, the functions supported by the chip can be referred to above and will not be repeated here. A person of ordinary skill in the art will understand that all or part of the steps of implementing the above-mentioned embodiment can be completed by a program to instruct the relevant hardware. The program can be stored in a computer-readable storage medium. The storage medium mentioned above can be a read-only memory, a random access memory, etc. The above-mentioned processing unit or processor can be a central processing unit, a general-purpose processor, an application specific integrated circuit (ASIC), a microprocessor (digital signal processor, DSP), a field programmable gate array (FPGA) or other programmable logic device, transistor logic device, hardware component or any combination thereof.

[0178] The present application also provides a computer program product comprising instructions, which, when executed on a computer, causes the computer to perform any of the methods described in the above embodiments. The computer program product comprises one or more computer instructions. When the computer program instructions are loaded and executed on a computer, the process or function according to the embodiment of the present application is generated in whole or in part. The computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions may be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions may be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via a wired (e.g., coaxial cable, optical fiber, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) method. The computer-readable storage medium may be any available medium that a computer can access or a data storage device such as a server or data center that includes one or more available media integrated therein. The available media may be magnetic media (e.g., floppy disk, hard disk, tape), optical media (e.g., DVD), or semiconductor media (e.g., SSD).

[0179] It should be noted that the above-mentioned devices for storing computer instructions or computer programs provided in the embodiments of the present application, such as but not limited to the above-mentioned memories, computer-readable storage media and communication chips, etc., are all non-transitory.

[0180] In the above embodiments, it can be implemented in whole or in part by software, hardware, firmware or any combination thereof. When implemented using a software program, it can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, the process or function according to the embodiment of the present application is generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions can be transmitted from a website, computer, server or data center to another website, computer, server or data center by wired (e.g., coaxial cable, optical fiber, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that can be accessed by a computer or a data storage device such as a server or data center that contains one or more media that can be integrated. The available medium may be a magnetic medium (eg, a floppy disk, a hard disk, a magnetic tape), an optical medium (eg, a DVD), or a semiconductor medium (eg, a solid state disk (SSD)).

[0181] Although the present application is described herein in conjunction with various embodiments, in the process of implementing the claimed application, those skilled in the art may understand and implement other variations of the disclosed embodiments by reviewing the drawings, the disclosure, and the appended claims. In the claims, the word "comprising" does not exclude other components or steps, and "a" or "an" does not exclude multiple situations. A single processor or other unit may implement several functions listed in the claims. Certain measures are recorded in different dependent claims, but this does not mean that these measures cannot be combined to produce good results.

[0182] Although the present application has been described with reference to specific features and embodiments thereof, it is apparent that various modifications and combinations may be made thereto without departing from the spirit and scope of the present application. Accordingly, this specification and the drawings are merely illustrative of the present application as defined by the appended claims and are deemed to cover any and all modifications, variations, combinations or equivalents within the scope of the present application. Obviously, those skilled in the art may make various modifications and variations to the present application without departing from the spirit and scope of the present application. Thus, the present application is intended to include such modifications and variations as fall within the scope of the claims of the present application and their equivalents.

Claims

1. A reasoning method, characterized in that: Applied to a computing device, the method includes: In response to a first reasoning task input by a user, determining a reasoning result of the first reasoning task; Based on the user portrait, personalizing the inference result to obtain a first personalized result; the first personalized result has a greater matching degree with the user portrait than the inference result has with the user portrait; The first personalized result is output.

2. The method according to claim 1, characterized in that A first personalization model is deployed in the computing device, and the first personalization model is used to personalize the inference result based on the user portrait; The method further comprises: The first personalized model is adjusted based on a personalization degree of the first personalized result; the personalization degree is used to indicate a matching degree between the first personalized result and the preference information.

3. The method according to claim 2, characterized in that The method further comprises: Obtaining a second personalized result corresponding to a second reasoning task; the second reasoning task is a reasoning task in the history, wherein the first similarity between the second reasoning task and the first reasoning task is greater than a first similarity threshold, and the first satisfaction is greater than a satisfaction threshold, and the first satisfaction is the user's satisfaction with the second personalized result; Determining a second similarity between the first personalized result and the second personalized result, wherein the second similarity is positively correlated with the personalization degree of the first personalized result; The adjusting the first personalized model based on the personalization degree of the first personalized result includes: Based on the second similarity, the first personalized model is adjusted.

4. The method according to claim 3, characterized in that The adjusting the first personalized model based on the second similarity includes: When the second similarity is greater than a second similarity threshold, the first personalized model and the second personalized model are merged; the personalized processing capability of the second personalized model is lower than the personalized processing capability of the first personalized model; When the second similarity is less than a third similarity threshold, the first personalized model is trained using newly added training samples; and the third similarity threshold is less than the second similarity threshold.

5. The method according to claim 4, characterized in that The method of fusing the first personalized model with the second personalized model includes any one of the following: Horizontal superposition; Direct addition; Weights are added; Vertical stacking.

6. The method according to claim 2, characterized in that The method further comprises: Obtaining a second personalized result corresponding to a second reasoning task; the second reasoning task is a reasoning task in the historical reasoning task, the first similarity between which and the first reasoning task is greater than a first similarity threshold; Determining a degree of difference between the first personalized result and the second personalized result, wherein the degree of difference is negatively correlated with the degree of personalization of the first personalized result; The adjusting the first personalized model based on the personalization degree of the first personalized result includes: Based on the difference, the first personalized model is adjusted.

7. The method according to claim 6, characterized in that The adjusting the first personalized model based on the difference includes: When the difference is greater than a first difference threshold, training the first personalized model using the newly added training samples; When the difference is less than a second difference threshold, the first personalized model and the second personalized model are fused; the personalized processing capability of the second personalized model is lower than the personalized processing capability of the first personalized model, and the second difference threshold is less than the first difference threshold.

8. The method according to claim 6 or 7, characterized in that The determining the difference between the first personalized result and the second personalized result includes: Obtaining a first satisfaction level and a second satisfaction level; wherein the first satisfaction level refers to the user's satisfaction with the second personalized result, and the second satisfaction level refers to the user's satisfaction with the first personalized result; A degree of difference between the first personalized result and the second personalized result is determined based on the first satisfaction level and the second satisfaction level.

9. The method according to claim 8, characterized in that The computing device stores a correspondence between similarity and satisfaction; The obtaining of the second satisfaction level includes: determining a second similarity between the first personalized result and the second personalized result; From the corresponding relationship, the satisfaction level corresponding to the second similarity is used as the second satisfaction level.

10. A computing device, characterized in that include: processor and memory; The processor is connected to a memory, the memory is used to store computer-executable instructions, and the processor executes the computer-executable instructions stored in the memory to enable the computing device to implement the method according to any one of claims 1 to 9.

Images