JP7551970B2 - Artificial intelligence model updating method, apparatus, electronic device and storage medium - Google Patents

Description

The present disclosure relates to the field of artificial intelligence technology, and in particular to methods, apparatus, electronic devices, and storage media for updating artificial intelligence models in fields such as deep learning and natural language processing.

As technology advances, deep learning models obtained through training are being used in a variety of situations, such as classifying images using an image classification model or recognizing emotions in text sentences using an emotion recognition model. To improve the performance of a model, it is possible to update the model, for example by optimizing an existing model.

The present disclosure provides an artificial intelligence model updating method, device, electronic device, and storage medium.

An artificial intelligence model updating method is provided, which includes a dialogue system having a natural interaction with a user, acquiring interaction content including M (M is a positive integer greater than 1) dialogue contents performed for any one training sample, determining an update rule for a model to be processed based on the interaction content in response to the dialogue system determining that the dialogue with the user for the training sample has ended, and updating the model to be processed using the update rule.

An artificial intelligence model updating device is provided that is applied to a dialogue system and includes an interaction module and a first update module, in which the interaction module naturally interacts with a user to obtain interaction content including M (M is a positive integer greater than 1) dialogue contents performed for any one training sample, and the first update module determines an update rule for a model to be processed based on the interaction content in response to determining that the dialogue with the user for the training sample has ended, and updates the model to be processed using the update rule.

An electronic device is provided that includes at least one processor and a memory communicatively connected to the at least one processor, the memory storing commands executable by the at least one processor, and causing the at least one processor to execute the method when the commands are executed by the at least one processor.

A non-transitory computer-readable storage medium is provided that stores computer commands for causing a computer to execute the method.

A computer program product is provided that includes computer programs/commands that, when executed by a processor, implement the method.

It should be understood that the contents described in this section are not intended to identify key or essential features of the embodiments of the present disclosure, nor are they intended to limit the scope of the present disclosure. Other features of the present disclosure will be readily understood from the following specification.

The drawings are for better understanding of the present technical solution and are not intended to limit the present application.
1 is a flow chart of an embodiment of an artificial intelligence model updating method described in the present disclosure. FIG. 1 is a schematic diagram of an embodiment of the dialogue system described in the present disclosure and where the model to be processed is two independent models. FIG. 1 is a schematic diagram of an embodiment in which the dialogue system described in the present disclosure and the model to be processed are integrated into the same model. FIG. 4 is a schematic diagram of the configuration of a first embodiment 400 of an artificial intelligence model updating apparatus described in the present disclosure. FIG. 5 is a schematic diagram of the configuration of a second embodiment 500 of an artificial intelligence model updating apparatus described in the present disclosure. FIG. 6 is a schematic block diagram of an electronic device 600 in which embodiments of the present disclosure can be implemented.

The following describes exemplary embodiments of the present application based on the drawings. For ease of understanding, various details of the embodiments of the present application are included and should be considered as merely illustrative. Therefore, those skilled in the art should recognize that various changes and modifications can be made to the embodiments described herein without departing from the scope and spirit of the present application. Similarly, for the sake of simplicity, the following description omits descriptions of well-known functions and structures.

Furthermore, the term "and/or" in this specification is simply a relational relationship describing related objects, and means that three relations can exist. For example, A and/or B can mean three situations: A exists alone, A and B exist simultaneously, and B exists alone. Also, the character "/" in this specification generally means that the related objects before and after it are in an "or" relationship.

Figure 1 is a flowchart of an embodiment of the artificial intelligence model update method described in this disclosure. As shown in Figure 1, the following specific implementation methods are included:

In step 101, the dialogue system performs natural interactions with the user to obtain interaction content including M dialogue contents performed for any one of the training samples. M is a positive integer greater than 1.

In step 102, in response to determining that the dialogue with the user for the training sample has ended, the dialogue system determines an update rule for the model to be processed based on the interaction content, and updates the model to be processed using the update rule.

In conventional methods, when a model needs to be updated, it is generally updated based on collected training samples. However, according to the technical proposal described in the embodiment of the above-mentioned method, natural interaction with the user is possible, and the update rules for the model to be processed can be determined based on the content of the interaction, and the model to be processed can be updated based on the update rules. In other words, the user can achieve the purpose of updating the model through natural interaction alone, eliminating the need to collect and label training samples, reducing human and time costs and improving processing efficiency. Furthermore, updating the model can improve the model's performance.

The technical solutions described in the embodiments of the above-mentioned method can be applied to various situations such as image classification and emotion recognition. For example, for an image classification model in an image classification situation, the image classification model can be updated according to the technical solutions described in the embodiments of the above-mentioned method, and image classification can be performed using the updated image classification model, thereby improving the accuracy of the classification result. For an emotion recognition model in an emotion recognition situation, the emotion recognition model can be updated according to the technical solutions described in the examples of the above-mentioned method, and emotion recognition can be performed using the updated emotion recognition model, thereby improving the accuracy of the recognition result.

In practical applications, the dialogue system can naturally interact with a user, such as an expert, to obtain the interaction content, and then determine the update rules for the model to be processed based on the interaction content.

The interaction content may include M conversation contents performed for any one of the training samples, where M is a positive integer greater than 1. Accordingly, the update rule may be generated in response to determining that the conversation with the user for the training sample has ended.

Examples of the interaction content are as follows:

1) Image Classification Situation Suppose an image classification model performs poorly on a set of new data, then a user can naturally interact with the dialogue system as follows.
User: "Hello, your effect is not good in this picture." At the same time, attach the corresponding picture like picture a.
Dialogue system: "Looking at this picture, my classification is Pothos because the leaf and stem shapes all match those of a Pothos."
User: "This picture is actually a golden vine because it has gold stripes."
Dialogue system: "I've never come across a goldenrod before, is it a type of pothos?"
User: "No, it doesn't belong to the pothos."
Dialogue system: "I have documented a new type of goldenrod, its characteristics and its golden stripes. I will never make the same mistake again."
User: "Here's the next picture..."
As can be seen from the above, the user has multiple interactions with the dialogue system regarding the training sample picture a, and when it is determined that the dialogue with the user regarding picture a has ended, an update rule can be generated based on the content of the multiple interactions.
Here, according to the user's expression "Next is this picture...", it can be determined that the interaction with the user for picture a has ended.

2) Emotion Recognition Situation In this scenario, the user can naturally interact with the dialogue system as follows:
User: "Why do you classify the emotion as "neutral" for the statement, "Everyone who has been to this restaurant knows it, and as soon as they see it, they spit out their overnight rice?"
Dialogue system: "I think the phrase 'Everyone who's been there knows' is 60% positive, but the part at the end, 'They spit out the rice they had left over from the night before', seems negative, so I classified it as neutral."
User: "No, that's not true. 'Spit out the rice you ate the night before' is clearly a negative thing. The previous sentence, 'Everyone who's been there knows', isn't clear on its own whether it's positive or negative, but when combined with the following words it becomes a clear criticism."
Dialogue system: "Does the true meaning of 'Everyone who's been there knows' depend on the context?"
User: "That's true."
Dialogue system: "I remembered. I fixed it."
As can be seen from the above, the user has multiple interactions with the dialogue system regarding the training sample, "Everyone who has been to this restaurant knows it. When they see it, they immediately spit out the overnight meal," and when it is determined that the dialogue with the user regarding the training sample has ended, an update rule can be generated based on the content of the multiple interactions.
By combining the contents of multiple conversations, the meaning of what the user said can be better understood, and the accuracy of the generated update rules can be improved.

Furthermore, the obtained update rule can be used to update the model to be processed. In one embodiment of the present disclosure, updating the model to be processed can mean updating the neuron states and/or neuron connections of the model to be processed. That is, the update rule can be used to directly and permanently change the neuron states and/or connections of the model to be processed, thereby achieving the effect of model update, which can be realized simply and conveniently.

In one embodiment of the present disclosure, each of the dialogue contents may include a user input content and a response generated in response to the user input content. Here, the response may include a response generated based on the acquired neuron state and neuron connection information of the model to be processed.

For example, the following is the content of one conversation:
User: "Hello, your effect is not good in this picture." At the same time, attach the corresponding picture like picture a.
Dialogue system: "Looking at this picture, my classification is Pothos because the leaf and stem shapes all match those of a Pothos."
Here, the content corresponding to the “user” is the user input content, and the content corresponding to the “dialogue system” is the response generated to the user input content. For example, in combination with the acquired neuron state and neuron connection information of the model to be processed, it is possible to determine that the classification result of the picture is pothos, and generate the response accordingly.

In other words, it is possible to read the state of the model to be processed, obtain its neuron state and neuron connection information, and then accurately generate a response based on the user input based on that information, thereby improving the quality of the dialogue, etc.

In practical applications, a dialogue system can include an encoder and a decoder, and the encoder can be used to process the acquired interaction content, such as by encoding it, to obtain the necessary update rules. The decoder can also be used to generate responses to user input content.

The specific form of the update rule is not limited. For example, the "Citrus sinensis" in the above example is used as the label of picture a, and corresponding objective functions, learning rates, etc. are simultaneously generated, and these pieces of information are used as update rules, which can be used to update the neuron states and/or neuron connections of the model to be processed using the gradient back method. As another example, an update method for the neuron states and/or neuron connections is directly generated and stored in memory to be read by the model to be processed, and the neuron states and/or neuron connections are updated accordingly.

In practical applications, the update rules can be used to update only the neuron states of the model to be processed, or only the neuron connections of the model to be processed, or to simultaneously update the neuron states and neuron connections of the model to be processed, depending on practical needs.

In one embodiment of the present disclosure, the dialogue system can obtain a prediction result of a model to be processed and update according to the prediction result, where the prediction result is a prediction result of the training sample generated by the model to be processed updated using an update rule.

As described above, the interaction content can include M conversation contents performed for any one training sample. Therefore, after updating the model to be processed using the update rule, the training sample is further processed using the updated model to be processed, and, for example, image classification is performed on the picture a, thereby obtaining a prediction result, i.e., the image classification result of the picture a. In response, the dialogue system obtains this prediction result, and based on this prediction result, updates the dialogue system based on a method such as reinforcement learning, for example, updating the encoder and decoder, etc.

Still taking picture a as an example, after updating the model to be processed using the update rule, the model's performance will theoretically improve, so when re-predicting picture a using the model to be processed, the obtained prediction result may be different from before. For example, if the previous prediction result was pothos, the re-prediction result may be pothos or may be goldenrod. If the re-prediction result is still pothos, it indicates that the generated update rule may not be accurate, and the dialogue system can be updated accordingly to improve the accuracy of the subsequently generated update rule.

In other words, the above processing makes it possible to update the dialogue system, i.e., improve the performance of the dialogue system, and further improve the accuracy of the update rules generated based on the dialogue system, thereby improving the effectiveness of model updates, etc.

In one embodiment of the present disclosure, the dialogue system and the model to be processed may be two independent models, or may be integrated into the same model. The specific method to be adopted can be determined according to actual needs, which is very flexible and convenient.

2 is a schematic diagram of the dialogue system described in the present disclosure and the implementation method when the model to be processed is two independent models. As shown in FIG. 2, the dialogue system may include an encoder and a decoder. The dialogue system receives information, i.e., obtains user input contents including natural language (e.g., characters entered by the user), pictures, voice, etc., and outputs information, i.e., generates a response to the user input contents to dialogue with the user. Here, a response can be generated based on the neuron state and neuron connection information of the model to be processed obtained by reading the state. In addition, the dialogue system can generate an update rule from multiple dialogue contents with the user, and update the neuron state and/or neuron connection of the model to be processed using the update rule. In addition, a training sample corresponding to the multiple dialogue contents can be input into the updated model to be processed to obtain a prediction result, and the dialogue system can be updated using the prediction result.

The model to be processed may be any model, such as a model to be optimized that is already in use, a model that has not been used but has been trained to some extent in advance, or an untrained model.

Figure 3 is a schematic diagram of an implementation method when the dialogue system described in this disclosure and the model to be processed are integrated into the same model. Compared to the method shown in Figure 2, the implementation principle is the same, but the information exchange between the dialogue system and the model to be processed is only implicit. Also, in order to simplify the drawing, "training samples" and "prediction results" are not shown.

Note that, for the sake of simplicity, the embodiments of the above-mentioned methods are expressed as a combination of a series of operations. However, as will be appreciated by those skilled in the art, some steps of the present application can be performed in other orders or simultaneously, and therefore the present application is not limited to the order of operations described. Next, those skilled in the art should understand that any of the embodiments described in the specification are preferred embodiments, and related operations and modules are not necessarily required by the present application.

In short, according to the technical solution described in the embodiment of the method of the present disclosure, the user can achieve the purpose of updating the model only through natural interaction methods, and thus improve the performance of the model. Furthermore, the updating method is simple and convenient, and can allow the model to quickly adapt to actual changes and needs, etc.

The above is a description of an embodiment of a method, and below, the technical solutions described in this disclosure will be further explained through an embodiment of an apparatus.

Figure 4 is a block diagram of a first embodiment 400 of an artificial intelligence model update device described in the present disclosure. The device can be applied to a dialogue system. As shown in Figure 4, it includes an interaction module 401 and a first update module 402.

The interaction module 401 acquires interaction content including M conversation contents performed for any one of the training samples by natural interaction with the user. M is a positive integer greater than 1.

In response to determining that the interaction with the user for the training sample has ended, the first update module 402 determines an update rule for the model to be processed based on the content of the interaction, and updates the model to be processed using the update rule.

The technical proposal described in the embodiment of the device described above enables natural interaction with the user, determines update rules for the model to be processed based on the content of the interaction, and further updates the model to be processed based on the update rules. In other words, the user can achieve the purpose of updating the model through natural interaction alone, and there is no need to collect or label training samples, thus reducing human and time costs and improving processing efficiency. Furthermore, model updates can improve model performance.

The first update module 402 can use the obtained update rule to update the model to be processed. In one embodiment of the present disclosure, updating the model to be processed can mean updating the neuron states and/or neuron connections of the model to be processed. That is, the update rule can be used to directly and permanently change the neuron states and/or connections of the model to be processed to achieve the effect of model update.

In other words, the update rule may update only the neuron state of the model to be processed, or may update only the neuron connections of the model to be processed, or may update both the neuron state and the neuron connections of the model to be processed simultaneously, depending on the actual needs.

In one embodiment of the present disclosure, each of the dialogue contents may include a user input content and a response generated in response to the user input content. Here, the response may include a response generated based on the acquired neuron state and neuron connection information of the model to be processed.

In other words, the state of the model to be processed is read to obtain its neuron state and neuron connection information, and a response can be generated based on the information in response to user input.

Figure 5 is a schematic diagram of the configuration of a second embodiment 500 of the artificial intelligence model updating device described in the present disclosure. The device can be applied to a dialogue system. As shown in Figure 5, it includes an interaction module 401, a first update module 402, and a second update module 403.

Here, the interaction module 401 and the first update module 402 are the same as in the embodiment shown in FIG. 4 and will not be described further.

The second update module 403 obtains a prediction result of the model to be processed and updates the dialogue system according to the prediction result, where the prediction result is a prediction result of the training sample generated by the model to be processed updated using the update rule.

As described above, the interaction content may include M dialogue contents performed for any one training sample. Therefore, after updating the model to be processed using the update rule, the training sample may be further processed using the updated model to be processed, and, for example, image classification may be performed on the picture a, thereby obtaining a prediction result, i.e., an image classification result for the image a. Accordingly, the second update module 403 may obtain the prediction result and update the dialogue system according to a method such as reinforcement learning based on the prediction result.

The specific workflow of the device embodiment shown in Figures 4 and 5 can be referred to the relevant description in the method embodiment described above and will not be described further.

In short, according to the technical solution described in the embodiment of the device of the present disclosure, the user can achieve the purpose of updating the model only through natural interaction methods, so that the performance of the model can be improved. Moreover, the updating method is simple and convenient, so that the model can quickly adapt to actual changes and needs, etc.

The technical proposal described in this application can be applied to the field of artificial intelligence, especially to the fields of deep learning and natural language processing. Artificial intelligence is a discipline that studies how computers simulate human thought processes and intelligent behaviors (such as learning, reasoning, thinking, planning, etc.), and includes both hardware-level and software-level technologies. Artificial intelligence hardware technologies generally include technologies such as sensors, dedicated artificial intelligence chips, cloud computing, distributed storage, big data processing, etc. Artificial intelligence software technologies mainly include computer vision technology, voice recognition technology, natural language processing technology, and several directions such as machine learning/deep learning, big data processing technology, and knowledge map technology.

In addition, the interaction content in the embodiments of the present disclosure is not targeted to a specific user, and does not reflect personal information of a specific user. Furthermore, the implementer of the method of the present disclosure can obtain the interaction content in a manner that complies with various disclosure and laws and regulations, for example, by obtaining the content from the user with the user's permission.

In the technical solution disclosed herein, the acquisition, storage, application, processing, transmission, provision and distribution of personal information of the users involved all comply with the provisions of relevant laws and regulations and are not contrary to public order and morals.

According to embodiments of the present disclosure, the present disclosure further provides an electronic device, a readable storage medium, and a computer program product.

FIG. 6 is a schematic block diagram of an electronic device 600 that may be used to implement embodiments of the present disclosure. The electronic device may represent various forms of digital computers, such as laptops, desktop computers, workbenches, servers, blade servers, mainframes, and other suitable computers. The electronic device may also represent various forms of mobile devices, such as PDAs, mobile phones, smartphones, wearable devices, and other similar computing devices. The components, their connections and relationships, and their functions shown herein are merely examples and are not intended to limit the implementation of the present disclosure as described and/or claimed herein.

As shown in FIG. 6, the device 600 includes a computing means 601 capable of performing various appropriate operations and processes according to a computer program stored in a read only memory (ROM) 602 or loaded from a storage means 608 into a random access memory (RAM) 603. The RAM 603 may store various programs and data necessary for the operation of the device 600. The computing means 601, the ROM 602, and the RAM 603 are connected via a bus 604. An input/output (I/O) interface 605 is also connected to the bus 604.

Several components of the device 600 are connected to the I/O interface 605, including input means 606, e.g., a keyboard, a mouse, etc., output means 607, e.g., various types of displays, speakers, etc., storage means 608, e.g., a magnetic disk, an optical disk, etc., and communication means 609, e.g., a network card, a modem, a wireless communication transceiver, etc. The communication means 609 allows the device 600 to exchange information/data with other devices via a computer network, e.g., the Internet, and/or various telecommunication networks.

The computing means 601 may be various general-purpose and/or dedicated processing components having processing and computing capabilities. Some examples of the computing means 601 include, but are not limited to, a central processing unit (CPU), a graphic processing unit (GPU), various dedicated artificial intelligence (AI) computing chips, various computing units that execute machine learning model algorithms, digital signal processors (DSPs), and any suitable processor, controller, microcontroller, etc. The computing means 601 executes various methods and processes described above, such as the methods described in the present disclosure. For example, in some embodiments, the methods described in the present disclosure may be implemented as a computer software program physically embodied in a machine-readable medium, such as the storage means 608. In some embodiments, some or all of the computer program may be loaded and/or installed in the device 600 via the ROM 602 and/or the communication means 609. When the computer program is loaded into the RAM 603 and executed by the computing means 601, one or more steps of the methods described in the present disclosure may be executed. Alternatively, in other embodiments, the computing means 601 may be configured in any other suitable manner (e.g., via firmware) to perform the methods described herein.

Various embodiments of the systems and techniques described herein may be implemented in digital electronic circuitry systems, integrated circuit systems, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on chips (SOCs), load programmable logic devices (CPLDs), computer hardware, firmware, software, and/or combinations thereof. These various embodiments may include being embodied in one or more computer programs. The one or more computer programs may be executed and/or interpreted on a programmable system that includes at least one programmable processor. The programmable processor may be a dedicated or general purpose programmable processor that is capable of receiving data and instructions from a storage system, at least one input device, and at least one output device, and transferring data and instructions to the storage system, the at least one input device, and the at least one output device.

Program codes for implementing the methods of the present disclosure can be written using any combination of one or more programming languages. These program codes can be provided to a processor or controller of a general purpose computer, a special purpose computer, or other programmable data processing apparatus such that, when executed by the processor or controller, the program codes perform the functions/operations specified in the flowcharts and/or block diagrams. The program codes can be executed entirely on the machine, partially on the machine, as a stand-alone package partially on the machine and partially on a remote machine, or entirely on a remote machine or server.

In the context of this disclosure, a machine-readable medium is a tangible medium that can contain or store a program for use by or in conjunction with an instruction execution system, apparatus, or device. The machine-readable medium may be a machine-readable signal medium or a machine-readable storage medium. The machine-readable medium may include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any suitable combination thereof. More specific examples of machine-readable storage media include one or more line-based electrical connections, portable computer disks, hard disks, random access memories (RAMs), read-only memories (ROMs), erasable programmable read-only memories (EPROMs or flash memories), optical fibers, portable compact disk read-only memories (CD-ROMs), optical storage devices, magnetic storage devices, or any suitable combination thereof.

To provide interaction with a user, the systems and techniques described herein may be implemented on a computer that includes a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user, and a keyboard and pointing device (e.g., a mouse or trackball) for providing input to the computer by the user. Other types of devices may also be used to provide interaction with a user. For example, feedback provided to the user may be any form of sensing feedback (e.g., visual feedback, auditory feedback, or haptic feedback), and input from the user may be received in any form (including sound input, speech input, or haptic input).

The systems and techniques described herein may be implemented in a computing system that includes back-end components (e.g., a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes front-end components (e.g., a client computer having a graphical user interface or a web browser through which a user can interact with embodiments of the systems and techniques described herein), or any combination of such back-end, middleware, or front-end components. The components of the system may be interconnected by any form or medium of digital data communication (e.g., a communications network). Examples of communications networks include a local area network ("LAN"), a wide area network ("WAN"), and an internetwork.

The computer system may include clients and servers. The clients and servers are generally remote from each other and typically interact through a communication network. The relationship between the clients and servers arises by computer programs running on the corresponding computers and having a client-server relationship with each other. The servers may be cloud servers, servers in a distributed system, or servers in combination with a blockchain.

It should be understood that steps may be rearranged, added, or removed using the various types of flows shown above. For example, each step described in this application may be performed in a parallel or sequential order, or may be performed in a different order, and is not limited thereto, so long as the desired results of the technical proposal disclosed in this application are achieved.

The above specific embodiments do not constitute limitations on the scope of protection of this application. Those skilled in the art should understand that various modifications, combinations, sub-combinations and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions and improvements made within the spirit and principles of this application should be included in the scope of protection of this application.

Claims (10)

1. A method for updating an artificial intelligence model, comprising:
The dialogue system performs natural interactions with the user to obtain interaction contents including M (M is a positive integer greater than 1) dialogue contents performed for any one of the training samples;
In response to the dialogue system determining that the dialogue with the user for the training sample has ended, determining an update rule for a model to be processed based on the interaction content, and updating the model to be processed using the update rule;
Including,
The update rule includes update method information of neuron states and/or neuron connections of the model to be processed;
updating the model to be processed using the update rule comprises updating neuronal states and/or neuronal connections of the model to be processed using the update rule.
How to update artificial intelligence models. Each of the dialogue contents includes a user input content and a response generated in response to the user input content,
The response includes a response generated based on the acquired information of the neuron states and the neuron connections of the model to be processed.
The method for updating an artificial intelligence model according to claim 1 . The dialogue system further includes obtaining a prediction result of the model to be processed, the prediction result being updated using the update rule, of the training sample generated by the model to be processed, and updating according to the prediction result.
3. A method for updating an artificial intelligence model according to claim 1 or 2 . The dialogue system and the model to be processed are two independent models;
Alternatively, the dialogue system and the model to be processed are integrated into the same model.
3. A method for updating an artificial intelligence model according to claim 1 or 2 . An update device for an artificial intelligence model applied to a dialogue system, comprising: an interaction module and a first update module;
The interaction module performs natural interactions with a user to obtain interaction content including M (M is a positive integer greater than 1) conversation contents performed for any one training sample;
The first update module, in response to determining that the interaction with the user on the training sample is completed, determines an update rule for a model to be processed based on the interaction content, and updates the model to be processed using the update rule, the update rule including update method information of neuron states and/or neuron connections of the model to be processed;
updating the model to be processed using the update rule comprises updating neuronal states and/or neuronal connections of the model to be processed using the update rule.
Artificial intelligence model updater. Each of the dialogue contents includes a user input content and a response generated in response to the user input content,
The response includes a response generated based on the acquired information of the neuron states and the neuron connections of the model to be processed.
6. An apparatus for updating an artificial intelligence model according to claim 5 . a second update module for obtaining a prediction result of the model to be processed, the prediction result being updated using the update rule, of the training sample generated by the model to be processed, and updating the dialogue system according to the prediction result;
7. An apparatus for updating an artificial intelligence model according to claim 5 or 6 . At least one processor;
a memory communicatively coupled to the at least one processor;
An electronic device having stored in the memory instructions executable by the at least one processor, the commands, when executed by the at least one processor, causing the at least one processor to perform the artificial intelligence model updating method described in claim 1. A non-transitory computer-readable storage medium storing computer commands for causing a computer to execute the artificial intelligence model updating method described in claim 1. A computer program that, when executed by a processor, realizes the method for updating an artificial intelligence model according to claim 1.

Images