Method, device and equipment for model updating, medium and program product

Abstract

According to the embodiment of the invention, a model updating method, device and equipment, a medium and a program product are provided. In the method, first interaction data associated with a user is acquired, the first interaction data comprises data related to an interaction operation performed by the user and an interaction application, and the interaction operation is performed based on a machine learning model associated with the interaction application; operation feature representation corresponding to the first interaction data is determined at least based on time and preference parameters corresponding to the interaction operation, and the preference parameters are determined based on the interaction operation and the time corresponding to the interaction operation; and updating the machine learning model based on the operational feature representation.

Description

Technical Field

[0001] Example embodiments of the present disclosure generally relate to the field of computers, and in particular, to methods, apparatuses, devices, computer-readable storage media, and computer program products for interaction. Background Art

[0002] With the development of machine learning technology, interaction systems based on machine learning models can relatively maturely achieve multi-faceted interactions with users. However, many interaction systems based on machine learning models are difficult to achieve personalized services for users, resulting in poor user experience. Summary of the Invention

[0003] In a first aspect of the present disclosure, a method for model update is provided. The method includes: obtaining first interaction data associated with a user, the first interaction data including data related to an interaction operation between the user and an interaction application, the interaction operation being based on a machine learning model associated with the interaction application; determining an operation feature representation corresponding to the first interaction data at least based on the time corresponding to the interaction operation and a preference parameter, where the preference parameter is determined based on the interaction operation and the time corresponding to the interaction operation; and updating the machine learning model based on the operation feature representation.

[0004] In a second aspect of the present disclosure, an apparatus for model update is provided. The apparatus includes: an obtaining module configured to obtain first interaction data associated with a user, the first interaction data including data related to an interaction operation between the user and an interaction application, the interaction operation being based on a machine learning model associated with the interaction application; a determining module configured to determine an operation feature representation corresponding to the first interaction data at least based on the time corresponding to the interaction operation and a preference parameter, where the preference parameter is determined based on the interaction operation and the time corresponding to the interaction operation; and an updating module configured to update the machine learning model based on the operation feature representation.

[0005] In a third aspect of the present disclosure, an electronic device is provided. The device includes at least one processor; and at least one memory, the at least one memory being coupled to the at least one processor and storing instructions for execution by the at least one processor. The instructions, when executed by the at least one processor, cause the device to execute the method of the first aspect.

[0006] In a fourth aspect of the present disclosure, a computer-readable storage medium is provided. Computer-executable instructions are stored on the medium, and when the computer-executable instructions are executed by a processor, the method of the first aspect is implemented.

[0007] In a fifth aspect of the present disclosure, a computer program product is provided. The computer program product is tangibly stored in a computer storage medium and includes computer-executable instructions that, when executed by a device, cause the device to perform the method of the first aspect.

[0008] It should be understood that the content described in this content part is not intended to define the key features or important features of the embodiments of the present disclosure, nor is it used to limit the scope of the present disclosure. Other features of the present disclosure will become readily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] In conjunction with the accompanying drawings and with reference to the following detailed description, the above and other features, advantages, and aspects of the embodiments of the present disclosure will become more apparent. In the drawings, the same or similar reference numerals denote the same or similar elements, where: Figure 1 A schematic diagram showing an example environment in which embodiments of the present disclosure can be implemented is shown; Figure 2 A flowchart showing a process for model update according to some embodiments of the present disclosure is shown; Figure 3 A schematic structural block diagram showing an example device for model update according to some embodiments of the present disclosure is shown; and Figure 4 A block diagram of an electronic device capable of implementing one or more embodiments of the present disclosure is shown. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0010] Embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although some embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure can be implemented in various forms and should not be construed as limited to the embodiments set forth herein. On the contrary, these embodiments are provided to more thoroughly and completely understand the present disclosure. It should be understood that the drawings and embodiments of the present disclosure are for exemplary purposes only and are not used to limit the protection scope of the present disclosure.

[0011] In the description of the embodiments of the present disclosure, the term "including" and its like shall be understood as an open inclusion, i.e., "including but not limited to". The term "based on" shall be understood as "at least partially based on". The term "one embodiment" or "the embodiment" shall be understood as "at least one embodiment". The term "some embodiments" shall be understood as "at least some embodiments". There may also be other explicit and implicit definitions hereinafter.

[0012] It can be understood that the data involved in the technical solution (including but not limited to the data itself, the acquisition or use of the data) should comply with the requirements of the corresponding laws, regulations, and related provisions.

[0013] It should be understood that before using the technical solutions disclosed in the embodiments of the present disclosure, the types, usage scopes, usage scenarios, etc. of the personal information involved in the present disclosure should be informed to users and the authorization of users should be obtained through appropriate means in accordance with relevant laws and regulations.

[0014] For example, when responding to receiving an active request from a user, a prompt message is sent to the user to clearly prompt the user that the operation requested by the user will require obtaining and using the user's personal information, so that the user can autonomously choose whether to provide personal information to software or hardware such as an electronic device, an application program, a server, or a storage medium that performs the operations of the technical solutions of the present disclosure according to the prompt message.

[0015] As an optional but non-limiting implementation manner, the manner of sending a prompt message to the user in response to receiving an active request from the user can be, for example, in the form of a pop-up window, and the prompt message can be presented in text in the pop-up window. In addition, the pop-up window can also carry a selection control for the user to choose "agree" or "disagree" to provide personal information to the electronic device.

[0016] It should be understood that the above processes of notifying and obtaining user authorization are only illustrative and do not constitute a limitation on the implementation manners of the present disclosure, and other manners that meet relevant laws and regulations can also be applied to the implementation manners of the present disclosure.

[0017] As used herein, the term "model" can learn the corresponding association relationship between input and output from training data, so that after training is completed, for a given input, a corresponding output can be generated. The generation of the model can be based on machine learning techniques. Deep learning is a machine learning algorithm that processes inputs and provides corresponding outputs by using multiple processing units. A neural network model is an example of a model based on deep learning. In this article, "model" can also be referred to as "machine learning model", "learning model", "machine learning network", or "learning network", and these terms can be used interchangeably in this article.

[0018] A "neural network" is a machine learning network based on deep learning. A neural network can process inputs and provide corresponding outputs, and it generally includes an input layer and an output layer and one or more hidden layers between the input layer and the output layer. Neural networks used in deep learning applications usually include many hidden layers, thereby increasing the depth of the network. The layers of the neural network are connected in sequence, so that the output of the previous layer is provided as the input of the next layer, where the input layer receives the input of the neural network, and the output of the output layer is the final output of the neural network. Each layer of the neural network includes one or more nodes (also referred to as processing nodes or neurons), and each node processes the input from the previous layer.

[0019] Generally, machine learning can roughly include three stages, namely, a training stage, a testing stage, and an application stage (also referred to as an inference stage). In the training stage, a given model can be trained using a large amount of training data, and the parameter values are continuously iteratively updated until the model can obtain consistent inferences that meet the expected goals from the training data. Through training, the model can be considered to be able to learn the association from input to output (also referred to as the mapping from input to output) from the training data. The parameter values of the trained model are determined. In the testing stage, the test input is applied to the trained model to test whether the model can provide the correct output, thereby determining the performance of the model. In the application stage, the model can be used to process the actual input based on the parameter values obtained through training and determine the corresponding output.

[0020] Figure 1 FIG. shows a schematic diagram of an example environment 100 in which embodiments of the present disclosure can be implemented. The environment 100 may include a robot device 120 and a terminal device 150. Interaction applications 125-1 and 125-2 may be installed in the robot device 120 and the terminal device 150 respectively, and are collectively referred to as the interaction application 125. The interaction application 125 may be associated with a machine learning model 130. The user 140 may interact with the interaction application 125 associated with the machine learning model 130 through the robot device 120 and the terminal device 150 respectively. As an example, the machine learning model 130 may be a trained model.

[0021] In some embodiments, the machine learning model 130 may be configured in the server device 110. According to application requirements, the machine learning model 130 may be constructed to include one or more appropriate types of model architectures. Embodiments of the present disclosure do not limit the specific type and structure of the machine learning model 130. In other embodiments, the machine learning model 130 may also be configured in the robot device 120 or the terminal device 150. Embodiments of the present disclosure are described by taking the machine learning model 130 being configured in the server device 110 as an example.

[0022] As an example, in addition to the interaction application 125-1, the robot device 120 may also be installed with hardware components such as a display screen, a microphone, a speaker, a camera, a sensor, etc., so as to receive information about the surrounding environment and some information about the user 140. In an example scenario, the user 140 may perform interaction operations with the robot device 120 through the interaction application and the hardware components. The interaction operations may at least include a conversation between the user 140 and the robot device 120. For the robot device 120 installed with components such as a robotic arm, robotic legs, and a foot mechanism that can support movement, services such as labor may also be provided for the user 140. The robot device 120 discussed herein may include, but is not limited to, personal / home service robots, public service robots, etc.

[0023] In environment 100, the terminal device 150 can be any type of mobile terminal, fixed terminal, or portable terminal, including mobile phones, desktop computers, laptop computers, notebook computers, netbook computers, tablet computers, media computers, multimedia tablets, personal communication system (PCS) devices, personal navigation devices, personal digital assistants (PDAs), audio / video players, digital cameras / video cameras, positioning devices, television receivers, radio broadcast receivers, e-book devices, gaming devices, or any combination of the foregoing, including accessories and peripherals of these devices or any combination thereof. In some embodiments, the terminal device 150 is also capable of supporting any type of user interface (such as "wearable" circuitry, etc.). The server device 110 can be implemented, for example, in various types of computing systems / servers capable of providing computing capabilities, including but not limited to mainframes, edge computing nodes, computing devices in cloud environments, and the like.

[0024] It should be understood that the structures and functions of the various elements in environment 100 are described only for exemplary purposes and do not imply any limitation on the scope of the present disclosure.

[0025] As mentioned above, many current interaction systems based on machine learning models are difficult to achieve personalized services for users, resulting in poor user experience. Therefore, it is desirable that interaction systems or interaction applications based on machine learning models can be more flexible to better serve users.

[0026] In view of this, embodiments of the present disclosure propose a solution for model update. Specifically, in embodiments of the present disclosure, first interaction data associated with a user is obtained. The first interaction data includes data related to interaction operations performed by the user with an interaction application, and the interaction operations are based on a machine learning model associated with the interaction application. At least based on the time corresponding to the interaction operation and the preference parameter, an operation feature representation corresponding to the first interaction data is determined. The preference parameter is determined based on the interaction operation and the time corresponding to the interaction operation. Then, based on the operation feature representation, the machine learning model is updated.

[0027] In this way, by using the interaction application with the updated machine learning model to implement interaction with the user, embodiments of the present disclosure can quickly capture changes in user needs, improve the timeliness and accuracy of personalized services for users. It can also use the interaction application in different scenarios to interact with the user in the most appropriate way. This can better meet the user's needs when interacting with the user, thereby enhancing the user's interaction experience.

[0028] Some exemplary embodiments of the present disclosure will be described below with reference to the accompanying drawings.

[0029] Figure 2 FIG. 200 is a schematic diagram of a process for model update according to some embodiments of the present disclosure. For ease of discussion, these embodiments will be described in conjunction with Figure 1 environment 100. In some examples, these embodiments may be implemented at Figure 1 server device 110. In such a case, machine learning model 130 may be configured in server device 110. In other examples, these embodiments may also be implemented at Figure 1 robot device 120 or terminal device 150. In such a case, machine learning model 130 may be configured in robot device 120 or terminal device 150. The following specific embodiments take being implemented at server device 110 as an example.

[0030] In block 210, server device 110 obtains first interaction data associated with user 140. The first interaction data includes data related to the interaction operations that user 140 performs with interaction application 125. The interaction operations are based on machine learning model 130 associated with interaction application 125. Here, machine learning model 130 may be a trained model. Machine learning model 130 may be configured to receive interaction data generated by user 140 through interaction application 125, and output interaction feedback data for the corresponding interaction, and then feedback it to user 140 through interaction application 125.

[0031] In some embodiments, interaction application 125-1 may be configured in robot device 120. As an example, in addition to installing interaction application 125, robot device 120 may also be installed with hardware components such as a display screen, a microphone, a speaker, a camera, a sensor, etc., so as to receive information about the surrounding environment and some information about user 140. In an example scenario, user 140 may perform interaction operations with robot device 120 through the interaction application and the hardware components. The interaction operations may at least include a conversation between user 140 and robot device 120. For robot device 120 installed with components such as a robotic arm, robotic legs, and a foot mechanism that can support movement, services such as labor may also be provided for user 140. Robot device 120 discussed herein may include, but is not limited to, personal / home service robots, public service robots, etc.

[0032] Alternatively or additionally, interaction application 125-2 may be configured in terminal device 150. In such a case, user 140 may implement interaction operations with interaction application 125-2 through terminal device 150.

[0033] It should be noted that the acquisition of interaction data associated with users discussed in this article is carried out under the authorization of users and complies with relevant laws and regulations. For the first interaction data, or the second interaction data to be discussed below, both can include explicit data and implicit data. Explicit data can include, for example, data actively set by user 140 through interaction application 125, such as preference parameters. Implicit data can include, for example, conversation logs, operation data, third-party data, etc. on interaction application 125 after being authorized by user 140.

[0034] For example, the conversation log can include, for example, recording the question content of user 140, the response time interval of interaction application 125 to the question, and correction operations such as "the previous answer is incorrect" feedback by user 140. The interaction information of user 140 can include, for example, click preferences on interaction application 125 (such as the browsing duration of recommended content, etc.), operation information on terminal device 150 and / or robot device 120 (such as whether text input or voice input is usually used in a certain period of time, etc.), or other operation data. Third-party data can include, for example, data in various open-source databases such as knowledge databases about medical knowledge, literary knowledge, etc., and e-commerce product feature libraries.

[0035] It should be understood that the above-listed multiple types of interaction data are only examples, given for the purpose of easy understanding, and the embodiments of the present disclosure are not limited in this regard. In actual applications, the interaction data can be any appropriate type of data set.

[0036] As an example, server device 110 can uniformly store interaction data of different structures for heterogeneous data integration. For example, structured data such as tables, semi-structured data such as JSON format logs, and unstructured data such as the voice input by user 140 can be uniformly stored. As an example, server device 110 can also mark through interaction application 125 whether the interaction data is from display data or implicit data.

[0037] In some embodiments, the server device 110 may preprocess the acquired interaction data. As an example, the preprocessing may include data cleaning, data augmentation, and data desensitization. For data cleaning, for example, the Isolation Forest algorithm may be used to detect interaction operations that deviate from the norm (such as 500 questions in a single day or other abnormal operation data). A score value calculated by the Isolation Forest algorithm close to 1 may indicate deviation from the norm, and close to 0 may indicate compliance with the norm. Data cleaning may also, for example, fill in missing values in the interaction data. For example, for an age parameter that is not set, the median age of users in the same region may be used for completion. For data augmentation processing, for example, in a translation scenario, the question-and-answer corpus of a certain type of knowledge may be expanded through the back-translation technique of translating Chinese into English and then back-translating from English into Chinese. Of course, there are other data augmentation processes, which will not be elaborated here. Through data cleaning and data augmentation, the error rate of the data can be effectively reduced.

[0038] In some embodiments, the interaction data that has been cleaned and augmented may be subjected to hierarchical desensitization processing. As an example, hierarchical desensitization may include strong desensitization and weak desensitization of the interaction data. For example, through strong desensitization processing, the sensitive privacy data of user 140 can be made irrecoverable. The implementation methods of strong desensitization processing may include, for example, the hash algorithm, the mobile phone number random replacement method, the invalidation processing method of the address field, etc. The implementation methods of weak desensitization processing may include, for example, encryption transformation that retains data recoverability, such as adding Laplace noise, generating a fixed-length hash value, etc.

[0039] In some embodiments, the desensitized interaction data may be isolated using a sandbox. For example, a trusted execution environment may be created through virtualization technology to ensure that sensitive data is only processed in an encrypted memory area. This method through the hardware isolation layer can allocate independent encrypted containers for each user, and the keys can be generated by various existing means, which will not be elaborated here. The verification mechanism may, for example, verify the hash value when the container is started to prevent tampering during runtime. Additionally, a dual-channel transmission system may be constructed to achieve data stream isolation. For example, for the first type of data, such as the channel for data with high confidentiality, it may only flow within the local sandbox. For the second type of data, such as data with low confidentiality requirements, it may be uploaded to the cloud according to corresponding requirements, such as through an appropriate protocol.

[0040] Through the above data preprocessing and data desensitization, the security of user data can be ensured. By storing the interaction data in the sandbox execution environment, the localization processing of confidential data and the upload of differential data parameters can be achieved.

[0041] At block 220, based at least on the time corresponding to the interaction operation and the preference parameter, the server device 110 determines an operation feature representation corresponding to the first interaction data. The preference parameter is determined based on the interaction operation and the time corresponding to the interaction operation. As an example, the preference parameter determined based on the interaction operation of the user 140 and the corresponding time can characterize the preference selection of the user 140 at that time, or the operation that the user hopes to occur, etc. How to determine the operation feature representation corresponding to the first interaction data based on the time corresponding to the interaction operation and the preference parameter will be discussed in detail below.

[0042] In some embodiments, to determine the operation feature representation corresponding to the first interaction data, based on the time period corresponding to the interaction operation and the attenuation coefficient, the server device 110 may determine an attenuation feature parameter. Based on the preference parameter corresponding to the interaction operation and the enhancement coefficient, the server device 110 may determine an enhancement feature parameter. Then, based on the attenuation feature parameter and the enhancement feature parameter, the server device 110 may determine the operation feature representation of the first interaction data.

[0043] As an example, the attenuation coefficient can measure the attenuation rate of the preference of the user 140, and it can reflect the attenuation rate of the preference of the user 140 over time. The enhancement coefficient can be used to adjust the contribution degree of the preference parameter to the feature representation to be generated. As an example, the attenuation coefficient can be dynamically adjusted to adapt to different types of preferences of the user 140. The enhancement coefficient can also be dynamically adjusted. How to determine the attenuation coefficient and the enhancement coefficient will be discussed in detail below.

[0044] In some embodiments, based on the operation attribute corresponding to the interaction operation, the server device 110 may determine at least one of the attenuation coefficient and the enhancement coefficient. The operation attribute may indicate a positive example attribute or a negative example attribute of the interaction operation.

[0045] As an example, for the interaction operation with the positive example attribute, it can be a positive and active operation. For example, the user 140 actively shares content to the social network. For another example, the user 140 continuously learns exercise courses according to the plan for multiple days in the fitness application, which meets the goal of the fitness application for the expected user to develop continuous fitness, and can belong to a positive operation. It should be understood that these are only examples and are not intended to be any limitation, and the interaction operation with the positive example attribute can be specifically defined according to the actual application situation.

[0046] As an example, for the interaction operation of negative example attributes, it can be an operation opposite to the target operation, or an operation that does not meet the conditions or does not show the expected enthusiasm. For example, when recommending content to user 140 on robot device 120 or terminal device 150, if the user clicks the "Don't show again" button, such a negative feedback operation can be an operation of negative example attributes. Another example is that in a reading application, if user 140 never opens the application for reading after registration, or quickly exits after opening without any reading, it belongs to a negative example that is contrary to the expected reading behavior of the corresponding user of the application. It should be understood that these are only examples and are not intended to impose any limitations. Specifically, the interaction operations of negative example attributes can be defined according to the actual application situation.

[0047] In this way, based on the positive example attributes corresponding to the interaction operation, a decay coefficient with a lower decay rate can be determined, and / or an enhancement coefficient with a higher enhancement rate can be determined. Based on the positive example attributes corresponding to the interaction operation, a decay coefficient with a higher decay rate can be determined, and / or an enhancement coefficient with a lower enhancement rate can be determined.

[0048] In some embodiments, in order to determine the decay characteristic parameter based on the time period corresponding to the interaction operation and the decay coefficient, the server device 110 can determine whether the time period corresponding to the interaction operation exceeds a predetermined threshold. As an example, the predetermined threshold can be a threshold for defining whether user 140's preference is long-term or short-term. The predetermined threshold can be any appropriate time period threshold, which can be specifically set according to the actual application, and the embodiments of the present disclosure do not limit this.

[0049] Furthermore, if the time period does not exceed the predetermined threshold, the server device 110 can determine that the decay characteristic parameter has a first decay form. The time period and the decay coefficient can be influencing factors of the first decay form. In some embodiments, the first decay form can be associated with an exponential function, where the time period and the decay coefficient are used to determine the exponent in the exponential function.

[0050] As an example, assume that is used to represent the time period corresponding to a certain interaction operation, and is used to represent the decay coefficient. Then the decay characteristic parameter with the first decay form can be represented, for example, as . Among them, can represent the start time of the interaction operation, and can represent the end time of the interaction operation. Thus, the time period and the decay coefficient can be used to determine the exponent in the decay characteristic parameter in the form of an exponential function.

[0051] Conversely, if the time period exceeds a predetermined threshold, the server device 110 may determine that the attenuation characteristic parameter has a second attenuation form. The time period and the attenuation coefficient may be influencing factors of the second attenuation form. In some embodiments, the second attenuation form may be associated with a logarithmic function, where the time period and the attenuation coefficient are used to determine the argument in the logarithmic function.

[0052] As an example, assume that is still used to represent the time period corresponding to a certain interaction operation, and is used to represent the attenuation coefficient. Then, the attenuation characteristic parameter with the second attenuation form can be represented as, for example, log(1 + λ (t − t 0 )). Thus, the time period and the attenuation coefficient are used to determine the argument in the attenuation characteristic parameter in the form of a logarithmic function.

[0053] Since the attenuation rate of short-term interests is higher than that of long-term preferences, for the short-term interests of user 140, an exponential form of attenuation characteristic parameter can be adopted. For the long-term preferences of user 140, a logarithmic form of attenuation characteristic parameter can be adopted. For example, for the short-term interests of user 140 in browsing promotional items, an exponential form of attenuation characteristic parameter can be adopted. For example, for the photography hobby of user 140 that has lasted for several years, an exponential form of attenuation characteristic parameter can be adopted.

[0054] As an example, for the operation characteristic representation of the first interaction data based on the attenuation characteristic parameter with the first attenuation form, the corresponding formula can be as shown in the following formula (1): (1) Wherein, can represent an enhancement coefficient, can represent a preference parameter for a certain interaction operation, can represent the preference parameters of n (n is an integer greater than or equal to 1) interaction operations within the time period. Additionally, can represent the initial interest weight, can represent the operation characteristic representation corresponding to the first interaction data generated within the time period.

[0055] As an example, for the operation characteristic representation of the first interaction data based on the attenuation characteristic parameter with the second attenuation form, it can, for example, correspond to the following formula (2): (2) In formula (2), , , , , The meanings represented respectively are as explained for the corresponding parameters in the above formula (1), and will not be elaborated here. As an example, the initial interest weight can be obtained by calculating the emotional intensity of operation data through a BERT model (a pre-trained language model).

[0056] Thus, by such a method of determining the operation feature representation of the first interaction data, the short-term preference changes of the user can be effectively obtained. In addition, in this way, the personalized characteristics of the user can be significantly distinguished.

[0057] In some embodiments, the server device 110 can construct a dual-channel feature extraction framework. The first channel can process explicitly set discrete parameters, such as interest tags, health tags, etc. actively set by the user. In the first channel, a semantic vector encoder can be adopted to convert natural language descriptions into multi-dimensional vectors. The first channel can analyze implicit operation sequences. For multiple types of data sources such as dialogue text sentiment analysis, interaction frequency, device sensor data, etc., a gated attention mechanism can be adopted to dynamically allocate weights. In addition, a cross-scenario transfer mechanism can also be adopted. For example, a domain adapter can be adopted to map some preferences learned in the e-commerce scenario to the home recommendation scenario through transfer learning to achieve cross-scenario transfer. It should be understood that the semantic vector encoder, gated attention mechanism, domain adapter, etc. used in the dual-channel feature extraction are only examples, and appropriate tools can be selected in combination with the type of interaction data in actual applications.

[0058] In some embodiments, a multi-modal coordination control algorithm can be adopted to coordinate the feature representations related to text, speech, facial expressions, actions, etc. As an example, an emotional intensity quantization model can be adopted, as shown in the following formula (3): (3) Wherein, are adjustable parameters, and the sum of is 1; can represent the text emotional intensity, can represent the facial expression emotional intensity, can represent the speech emotional intensity, can represent the quantization value of the audio emotional intensity. The purpose of this model can be to make the text generation, speech synthesis, and actions of the robot device 120 (such as facial expressions and body movements) coordinated in emotional expression by adjusting these parameters.

[0059] In addition, a delay alignment mechanism can also be designed to ensure the timing synchronization of text generation, speech synthesis, and virtual avatar actions. For example, when the text emotional intensity When it is 0.8 (or other value), it will trigger the bowing action of the robot device 120, and the duration of this action is, for example, 1.2 (or other time period) seconds, and an error range of 0.3 seconds (or other time period) is allowed. This mechanism helps to improve the naturalness and consistency of the emotional expression of the robot device 120. Thus, through the emotional intensity quantification model, the adaptive dialogue style adjustment can be realized, and the dialogue style can be automatically adjusted according to the dialogue situation and the user's emotion, so that the robot device 120 can communicate with the user 140 in the most appropriate way in different scenarios.

[0060] In block 230, the server device 110 updates the machine learning model 130 based on the operation feature representation. The update of the machine learning model 130 refers to the update of the trained machine learning model 130 discussed above. As an example, the update of the machine learning model 130 can be performed at a predetermined time interval. For example, the machine learning model 130 can be updated once a month (or other time interval). It can be understood that for the next update of the machine learning model 130, it can be based on the machine learning model 130 updated this time.

[0061] In this way, the embodiments of the present disclosure can realize the interaction with the user by using the interaction application of the updated machine learning model, can quickly capture the change of the user's needs, and improve the timeliness and accuracy of the personalized service for the user. It can also use the interaction application to interact with the user in the most appropriate way in different scenarios. This can better meet the user's needs when interacting with the user, thereby enhancing the user's interaction experience.

[0062] In some embodiments, when updating the machine learning model 130 based on the operation feature representation, based on the first interaction data, the server device 110 can determine the area where the user 140 is located, and determine the feature representation corresponding to at least one knowledge base associated with the area. As an example, the area where the user 140 is located can be a geographical area. When obtaining the authorization of the user 140 for the geographical location, if the terminal device 150 or the robot device 120 detects the area where the user 140 is currently located, the IP address corresponding to the area can be sent to the server device 110.

[0063] Further, the server device 110 can associate the feature representation corresponding to at least one knowledge base associated with the area with the operation feature representation. Then, based on the operation feature representation and the feature representation associated with the operation feature representation, the server device 110 can update the machine learning model 130.

[0064] As an example, at least one knowledge base associated with a region may include, for example, a regional dialect knowledge base, a regional culture knowledge base, etc. In this way, by associating the feature representation corresponding to the knowledge base related to the region with the operation feature representation, when the user 140 uses the interaction application 125 associated with the updated machine learning model 130 to interact, the interaction application 125 can provide the user 140 with information more in line with the region where the user is located. For example, through the robot device 120, it is possible to have a conversation with the user 140 in the dialect of the region where the user is located. This can improve the user's interaction interest and effectively enhance the interaction experience.

[0065] In some embodiments, the server device 110 may obtain second interaction data associated with the user 140 within a predetermined time period, and the interaction operation corresponding to the second interaction data is based on the interaction application 125 associated with the updated machine learning model 130. As an example, the predetermined time period may be any suitable time period during which the interaction application 125 associated with the updated machine learning model 130 occurs, and the embodiments of the present disclosure do not limit this.

[0066] Furthermore, based on the second interaction data, the server device 110 may determine description information about the user 140. The description information about the user may also be simply referred to as a user profile, which may be personal attributes used to describe the characteristics, hobbies, etc. of the user.

[0067] Then, the server device 110 may determine the difference between the description information and the reference description information. The reference description information may be determined based on the first interaction data. If it is determined that the difference exceeds a predetermined difference threshold, the server device 110 may indicate that the interaction application 125 is abnormal.

[0068] That is to say, if the difference between the reference description information about the user 140 before the update of the machine learning model 130 and the description information about the user 140 after the model update is large, there may be a problem with the interaction application 125. Because the changes in the characteristics, hobbies, etc. of the user 140 are generally relatively small, if the change is large, it may be that the interaction application 125 is abnormal. Abnormal situations such as problems with the settings of the interaction application 125, etc. As an example, the server device 110 may send a relevant notification message indicating that the interaction application 125 is abnormal to remind the user 140 to check the abnormal situation of the interaction application 125.

[0069] The embodiments of the present disclosure use the interaction application of the updated machine learning model to implement interaction with the user, which can quickly capture changes in user needs, improve the timeliness and accuracy of personalized services for users. It can also use the interaction application in different scenarios to interact with the user in the most appropriate way. This can better meet the user's needs when interacting with the user, thereby enhancing the user's interaction experience.

[0070] Figure 3 FIG. 3 shows a schematic structural block diagram of an exemplary apparatus 300 for object recommendation according to some embodiments of the present disclosure. The apparatus 300 may be implemented as or included in the server device 110. Each module / component in the apparatus 300 may be implemented by hardware, software, firmware, or any combination thereof.

[0071] As shown in the figure, the apparatus 300 includes an acquisition module 310 configured to acquire first interaction data associated with a user, the first interaction data including data related to an interaction operation performed by the user with an interaction application, and the interaction operation being based on a machine learning model associated with the interaction application; a determination module 320 configured to determine an operation feature representation corresponding to the first interaction data based at least on the time corresponding to the interaction operation and a preference parameter, where the preference parameter is determined based on the interaction operation and the time corresponding to the interaction operation; and an update module 330 configured to update the machine learning model based on the operation feature representation.

[0072] In some embodiments, the determination module 320 is further configured to determine an attenuation feature parameter based on a time period corresponding to the interaction operation and an attenuation coefficient; determine an enhancement feature parameter based on the preference parameter corresponding to the interaction operation and an enhancement coefficient; and determine the operation feature representation of the first interaction data based on the attenuation feature parameter and the enhancement feature parameter.

[0073] In some embodiments, the apparatus 300 further includes a coefficient determination module configured to determine at least one of the attenuation coefficient and the enhancement coefficient based on an operation attribute corresponding to the interaction operation, where the operation attribute indicates a positive example attribute or a negative example attribute of the interaction operation.

[0074] In some embodiments, the apparatus 300 is further configured to determine whether the time period corresponding to the interaction operation exceeds a predetermined threshold; in response to the time period not exceeding the predetermined threshold, determine that the attenuation feature parameter has a first attenuation form, where the time period and the attenuation coefficient are influencing factors of the first attenuation form; in response to the time period exceeding the predetermined threshold, determine that the attenuation feature parameter has a second attenuation form, where the time period and the attenuation coefficient are influencing factors of the second attenuation form.

[0075] In some embodiments, the first attenuation form is associated with an exponential function, where the time period and the attenuation coefficient are used to determine the exponent in the exponential function.

[0076] In some embodiments, the second attenuation form is associated with a logarithmic function, where the time period and the attenuation coefficient are used to determine the argument in the logarithmic function.

[0077] In some embodiments, the update module 330 is further configured to determine the region where the user is located based on the first interaction data; determine the feature representations corresponding to at least one knowledge base associated with the region; associate the feature representations corresponding to the at least one knowledge base with the operation feature representation; and update the machine learning model based on the operation feature representation and the feature representations associated with the operation feature representation.

[0078] In some embodiments, the apparatus 300 further includes a description module configured to obtain second interaction data associated with the user within a predetermined time period, where the interaction operation corresponding to the second interaction data is performed based on an interaction application associated with the updated machine learning model; determine description information about the user based on the second interaction data; determine the difference between the description information and reference description information, where the reference description information is determined based on the first interaction data; and in response to determining that the difference exceeds a predetermined difference threshold, indicate that the interaction application is abnormal.

[0079] In some embodiments, the interaction application may be implemented or configured at least at one of a robotic device and a terminal device.

[0080] Figure 4 The block diagram of an electronic device 400 in which one or more embodiments of the present disclosure may be implemented is shown. It should be understood that Figure 4 the illustrated electronic device 400 is merely exemplary and should not constitute any limitation to the functions and scope of the embodiments described herein. Figure 4 The illustrated electronic device 400 may be used to implement Figure 1 the server device 110 or Figure 3 the apparatus 300.

[0081] As Figure 4 shown, the electronic device 400 is in the form of a general-purpose computing device. The components of the electronic device 400 may include, but are not limited to, one or more processors or processing units 410, a memory 420, a storage device 430, one or more communication units 440, one or more input devices 450, and one or more output devices 460. The processing unit 410 may be an actual or virtual processor and is capable of performing various processes according to the programs stored in the memory 420. In a multi-processor system, multiple processing units execute computer-executable instructions in parallel to improve the parallel processing ability of the electronic device 400.

[0082] The electronic device 400 generally includes multiple computer storage media. Such media can be any accessible media available to the electronic device 400, including but not limited to volatile and non-volatile media, removable and non-removable media. The memory 420 can be volatile memory (such as registers, caches, random access memory (RAM)), non-volatile memory (such as read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory), or some combination thereof. The storage device 430 can be removable or non-removable media and can include machine-readable media such as a flash drive, a magnetic disk, or any other medium that can be capable of storing information and / or data (such as training data for training) and can be accessed within the electronic device 400.

[0083] The electronic device 400 can further include additional removable / non-removable, volatile / non-volatile storage media. Although not shown in Figure 4 it, a disk drive for reading from or writing to a removable, non-volatile magnetic disk (such as a "floppy disk") and an optical disk drive for reading from or writing to a removable, non-volatile optical disk can be provided. In these cases, each drive can be connected to a bus (not shown) by one or more data media interfaces. The memory 420 can include a computer program product 425 having one or more program modules that are configured to execute the various methods or actions of the various embodiments of the present disclosure.

[0084] The communication unit 440 enables communication with other electronic devices via a communication medium. Additionally, the functions of the components of the electronic device 400 can be implemented by a single computing cluster or multiple computer machines that are capable of communicating via a communication connection. Thus, the electronic device 400 can operate in a networked environment using a logical connection with one or more other servers, network personal computers (PCs), or another network node.

[0085] The input device 450 can be one or more input devices such as a mouse, a keyboard, a trackball, etc. The output device 460 can be one or more output devices such as a display, a speaker, a printer, etc. The electronic device 400 can also communicate with one or more external devices (not shown) as needed via the communication unit 440, external devices such as storage devices, display devices, etc., communicate with one or more devices that enable a user to interact with the electronic device 400, or communicate with any device that enables the electronic device 400 to communicate with one or more other electronic devices (such as a network card, a modem, etc.). Such communication can be performed via an input / output (I / O) interface (not shown).

[0086] According to an exemplary implementation of the present disclosure, there is provided a computer-readable storage medium having computer-executable instructions stored thereon, wherein the computer-executable instructions are executed by a processor to implement the method described above. According to an exemplary implementation of the present disclosure, there is also provided a computer program product, the computer program product being tangibly stored on a non-transitory computer-readable medium and including computer-executable instructions, and the computer-executable instructions being executed by a processor to implement the method described above.

[0087] Aspects of the present disclosure are described herein with reference to the flowcharts and / or block diagrams of methods, apparatuses, devices, and computer program products implemented according to the present disclosure. It should be understood that each block of the flowcharts and / or block diagrams, and combinations of blocks in the flowcharts and / or block diagrams, can be implemented by computer-readable program instructions.

[0088] These computer-readable program instructions can be provided to a processing unit of a general-purpose computer, a special-purpose computer, or other programmable data processing device, thereby producing a machine such that when these instructions are executed by the processing unit of the computer or other programmable data processing device, a device is produced that implements the functions / actions specified in one or more blocks of the flowchart and / or block diagram. These computer-readable program instructions can also be stored in a computer-readable storage medium, and these instructions cause a computer, a programmable data processing device, and / or other devices to work in a specific manner, so that the computer-readable medium storing the instructions includes a manufactured article, which includes instructions for implementing various aspects of the functions / actions specified in one or more blocks of the flowchart and / or block diagram.

[0089] The computer-readable program instructions can be loaded onto a computer, other programmable data processing device, or other device, so that a series of operation steps are executed on the computer, other programmable data processing device, or other device to produce a computer-implemented process, so that the instructions executed on the computer, other programmable data processing device, or other device implement the functions / actions specified in one or more blocks of the flowchart and / or block diagram.

[0090] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various implementations of the present disclosure. In this regard, each block in the flowchart or block diagram may represent a module, a segment of a program, or a portion of an instruction, which contains one or more executable instructions for implementing the specified logical function. In some alternative implementations, the functions noted in the blocks may occur in a different order than noted in the accompanying drawings. For example, two consecutive blocks may in fact be executed substantially in parallel, or they may sometimes be executed in the reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can be implemented by a dedicated hardware-based system that performs the specified functions or actions, or can be implemented by a combination of dedicated hardware and computer instructions.

[0091] The various implementations of the present disclosure have been described above. The above description is exemplary, not exhaustive, and is not limited to the disclosed implementations. Many modifications and variations will be apparent to those of ordinary skill in the art in the field of the present technology without departing from the scope and spirit of the described implementations. The choice of terms used herein is intended to best explain the principles of the implementations, the practical application, or the improvement of the technology in the market, or to enable other ordinary skilled artisans in the field of the present technology to understand the various implementation manners disclosed herein.

Claims

1. A method for model updating, characterized in that: The method comprises: Acquire first interaction data associated with a user, the first interaction data comprising data related to an interaction operation performed by the user with an interactive application, the interaction operation being performed based on a machine learning model associated with the interactive application; determining an operation characteristic representation corresponding to the first interaction data based at least on a time corresponding to the interaction operation and a preference parameter, wherein the preference parameter is determined based on the interaction operation and the time corresponding to the interaction operation; and Based on the operational feature representation, the machine learning model is updated.

2. The method according to claim 1, characterized in that Determining the operation characteristic representation corresponding to the first interaction data based at least on the time and preference parameters corresponding to the interaction operation includes: Determining an attenuation characteristic parameter based on a time period and an attenuation coefficient corresponding to the interactive operation; Determining enhanced feature parameters based on the preference parameters and enhancement coefficients corresponding to the interactive operation; and The operational characteristic representation of the first interaction data is determined based on the attenuation characteristic parameter and the enhancement characteristic parameter.

3. The method according to claim 2, characterized in that The method further comprises: At least one of the attenuation coefficient and the enhancement coefficient is determined based on an operation attribute corresponding to the interactive operation, wherein the operation attribute indicates a positive example attribute or a negative example attribute of the interactive operation.

4. The method according to claim 2, characterized in that: Determining the attenuation characteristic parameter based on the time period and the attenuation coefficient corresponding to the interactive operation includes: Determining whether the time period corresponding to the interactive operation exceeds a predetermined threshold; In response to the time period not exceeding the predetermined threshold, determining that the attenuation characteristic parameter has a first attenuation form, wherein the time period and the attenuation coefficient are influencing factors of the first attenuation form; In response to the time period exceeding the predetermined threshold, it is determined that the attenuation characteristic parameter has a second attenuation form, wherein the time period and the attenuation coefficient are influencing factors of the second attenuation form.

5. The method according to claim 4, characterized in that The first decay profile is associated with an exponential function, wherein the time period and the decay coefficient are used to determine an exponent in the exponential function.

6. The method according to claim 4, characterized in that The second decay form is associated with a logarithmic function, wherein the time period and the decay coefficient are used to determine a real number in the logarithmic function.

7. The method according to claim 1, characterized in that Based on the operational feature representation, updating the machine learning model comprises: Based on the first interaction data, determining a region where the user is located; determining a feature representation corresponding to at least one knowledge base associated with the region; Associating the feature representation corresponding to the at least one knowledge base to the operation feature representation; and The machine learning model is updated based on the operational feature representation and the feature representation associated with the operational feature representation.

8. The method according to claim 1, characterized in that The method further comprises: Acquire second interaction data associated with the user within a predetermined time period, where the interaction operation corresponding to the second interaction data is performed based on the interaction application associated with the updated machine learning model; Determining description information for the user based on the second interaction data; determining a difference between the description information and reference description information, the reference description information being determined based on the first interaction data; and In response to determining that the difference exceeds a predetermined difference threshold, indicating that the interactive application is abnormal.

9. The method according to claim 1, characterized in that The interactive application is configured in at least one of the following devices: Robotic devices, and Terminal device.

10. A device for model updating, characterized in that: The device comprises: an acquisition module configured to acquire first interaction data associated with a user, wherein the first interaction data includes data related to an interaction operation performed by the user with an interactive application, wherein the interaction operation is performed based on a machine learning model associated with the interactive application; a determination module configured to determine an operation characteristic representation corresponding to the first interaction data based at least on a time corresponding to the interaction operation and a preference parameter, wherein the preference parameter is determined based on the interaction operation and the time corresponding to the interaction operation; and An updating module is configured to update the machine learning model based on the operational feature representation.

11. An electronic device comprising: at least one processor; as well as At least one memory, the at least one memory is coupled to the at least one processor and stores instructions for execution by the at least one processor, the instructions causing the electronic device to perform the method according to any one of claims 1 to 9 when executed by the at least one processor.

12. A computer-readable storage medium having computer-executable instructions stored thereon, wherein the computer-executable instructions can be executed by a processor to implement the method according to any one of claims 1 to 9.

13. A computer program product comprising computer executable instructions, wherein the computer executable instructions, when executed by a processor, implement the method according to any one of claims 1 to 9.

Images