Method, device, electronic device and storage medium for controlling conversation reply content

Abstract

The embodiment of the present application provides a method, device, electronic device and storage medium for controlling the content of a conversation reply, wherein the method includes: obtaining a plurality of additional feature information; determining the user behavior characteristics according to the conversation context between the robot and the user; determining the target additional feature information that matches the user behavior characteristics from the plurality of additional feature information; modifying the robot's identity prompt using the target additional feature information, and controlling the robot's conversation reply content according to the target additional feature information. The method for controlling the content of a conversation reply provided by this embodiment uses the target additional feature information that matches the user behavior characteristics to modify the robot's identity prompt, and controls the robot's conversation reply content through the target additional feature information, so that the robot's conversation reply content is controllable, and the implementation process is convenient, which has significant advantages in terms of development and deployment costs.

Description

Technical Field

[0001] The present application relates to the field of computer technology, and in particular to a method, device, electronic device and storage medium for controlling conversation reply content. Background Art

[0002] Text generation schemes based on pre-trained language models have been widely used in various downstream tasks in recent years. Open-domain dialogue systems, commonly known as chatbots, are one of the representative tasks. Unlike task-based chatbots, the user voice or text input received by chatbots is not limited to any preset task domain, and its output text is generated word by word by a probabilistic language model. The difficulty of existing open-domain dialogue systems based on pre-trained language models is: how to regulate the content of dialogue replies so that the replies meet specific needs. Summary of the invention

[0003] In order to solve the above technical problems, the embodiments of the present application provide a method, device, electronic device and storage medium for controlling conversation reply content.

[0004] In a first aspect, an embodiment of the present application provides a method for controlling a conversation reply content, the method comprising:

[0005] Acquire a plurality of additional feature information, wherein the additional feature information includes at least one of the following: emotion feature information, action feature information, scene feature information, and character feature information;

[0006] Determine user behavior characteristics based on the conversation context between the robot and the user;

[0007] Determining target additional feature information matching the user behavior feature from the multiple types of additional feature information;

[0008] The target additional characteristic information is used to modify the identity prompt of the robot, and the dialogue reply content of the robot is controlled according to the target additional characteristic information.

[0009] In one embodiment, controlling the dialogue reply content of the robot according to the target additional feature information includes:

[0010] In a case where the target additional characteristic information is the emotional characteristic information, the emotional characteristic corresponding to the dialogue reply content is controlled according to the emotional characteristic information.

[0011] In one implementation, the emotion characteristic information includes emotion adverbial information or emotion attributive information, and controlling the emotion characteristic corresponding to the dialogue reply content according to the emotion characteristic information includes:

[0012] The reply content matching the emotional adverbial information or the emotional attributive information is generated through a pre-trained language model.

[0013] In one embodiment, controlling the dialogue reply content of the robot according to the target additional feature information includes:

[0014] In a case where the target additional feature information is the action feature information, the action feature corresponding to the dialogue reply content is controlled according to the action feature information.

[0015] In one implementation, the action feature information includes action adverbial information and action information, and the controlling of the action feature corresponding to the dialogue reply content according to the action feature information includes:

[0016] A reply content matching the action adverbial information and the action information is generated through a pre-trained language model.

[0017] In one embodiment, controlling the dialogue reply content of the robot according to the target additional feature information includes:

[0018] In the case where the target additional feature information is the scene feature information, the scene feature corresponding to the dialogue reply content is controlled according to the scene feature information.

[0019] In one embodiment, controlling the dialogue reply content of the robot according to the target additional feature information includes:

[0020] In a case where the target additional feature information is the personality feature information, the personality feature embodied in the dialogue reply content is controlled according to the personality feature information.

[0021] In a second aspect, an embodiment of the present application provides a device for controlling a conversation reply content, the device comprising:

[0022] An acquisition module, used to acquire a variety of additional feature information, wherein the additional feature information includes at least one of the following: emotion feature information, action feature information, scene feature information, and character feature information;

[0023] A determination module, used to determine user behavior characteristics based on the conversation context between the robot and the user;

[0024] A matching module, used for matching target additional feature information corresponding to the user behavior feature from the multiple additional feature information;

[0025] A control module is used to modify the identity prompt of the robot using the target additional characteristic information, and to control the dialogue reply content of the robot according to the target additional characteristic information.

[0026] In a third aspect, an embodiment of the present application provides an electronic device, including a memory and a processor, wherein the memory is used to store a computer program, and when the computer program is run by the processor, the method for controlling the conversation reply content provided in the first aspect is executed.

[0027] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium storing a computer program, which, when running on a processor, executes the method for controlling the conversation reply content provided in the first aspect.

[0028] The control method, device, electronic device and storage medium for the dialogue reply content provided by the present application above obtain a variety of additional feature information; determine the user behavior characteristics according to the dialogue context between the robot and the user; determine the target additional feature information that matches the user behavior characteristics from the multiple additional feature information; use the target additional feature information to modify the identity prompt of the robot, and control the dialogue reply content of the robot according to the target additional feature information. The control scheme for the dialogue reply content provided by this embodiment uses the target additional feature information that matches the user behavior characteristics to modify the identity prompt of the robot, and uses the target additional feature information to regulate the dialogue reply content of the robot, so that the dialogue reply content of the robot is controllable, and the implementation process is convenient, without changing the structure of the original language model, and without retraining the language model. Compared with the higher development and deployment costs of existing solutions, it has significant advantages in terms of development and deployment costs. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] In order to more clearly illustrate the technical solution of the present application, the following is a brief introduction to the drawings required for use in the embodiments. It should be understood that the following drawings only illustrate certain embodiments of the present application and should not be regarded as limiting the scope of protection of the present application. In each of the drawings, similar components are numbered similarly.

[0030] Figure 1 A schematic diagram showing a flow chart of a method for controlling the content of a conversation reply provided in an embodiment of the present application;

[0031] Figure 2 A structural schematic diagram of a device for controlling conversation reply content provided in an embodiment of the present application is shown. DETAILED DESCRIPTION

[0032] The technical solutions in the embodiments of the present application will be described clearly and completely below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all of the embodiments.

[0033] The components of the embodiments of the present application generally described and shown in the drawings herein may be arranged and designed in various configurations. Therefore, the following detailed description of the embodiments of the present application provided in the drawings is not intended to limit the scope of the application claimed for protection, but merely represents the selected embodiments of the present application. Based on the embodiments of the present application, all other embodiments obtained by those skilled in the art without making creative work belong to the scope of protection of the present application.

[0034] Hereinafter, the terms "including", "having" and their cognates, which may be used in various embodiments of the present application, are intended only to indicate specific features, numbers, steps, operations, elements, components, or combinations of the foregoing items, and should not be understood as first excluding the existence of one or more other features, numbers, steps, operations, elements, components, or combinations of the foregoing items or adding the possibility of one or more features, numbers, steps, operations, elements, components, or combinations of the foregoing items.

[0035] Furthermore, the terms “first”, “second”, “third”, etc. are merely used for distinguishing descriptions and are not to be understood as indicating or implying relative importance.

[0036] Unless otherwise defined, all terms (including technical terms and scientific terms) used herein have the same meanings as those generally understood by those skilled in the art to which the various embodiments of the present application belong. The terms (such as those defined in generally used dictionaries) will be interpreted as having the same meanings as the contextual meanings in the relevant technical field and will not be interpreted as having idealized meanings or overly formal meanings unless clearly defined in the various embodiments of the present application.

[0037] Example 1

[0038] The disclosed embodiment provides a method for controlling the content of a dialogue reply. The method is applied to a robot capable of performing dialogue replies, and the robot can provide an open domain dialogue system that supports multiple rounds of interaction, and conduct dialogues with users through the open domain dialogue system.

[0039] For details, see Figure 1 , the control methods of the dialogue reply content include:

[0040] Step S101, obtaining a variety of additional feature information.

[0041] In one embodiment, the additional feature information includes at least one of the following: emotion feature information, action feature information, scene feature information, and character feature information.

[0042] Specifically, the additional feature information can be set according to the specific needs of the user for the robot. Emotional feature information includes emotional adverbial information or emotional attributive information. Emotional adverbial information can be adverbials expressing emotions such as excitedly and happily, and emotional attributive information can be definitions of expressing emotions such as happy, pleased, lively, reserved, etc.

[0043] Action feature information includes action adverbial information and action information. The action adverbial information can be excitedly, exclamation, comforting, etc., and the action information can be information on specific robot actions such as speaking, talking, waving, etc. When the action adverbial information and the action information are combined together, examples are as follows: excitedly speaking, exclamation, comforting, etc.

[0044] The scene feature information includes scene adverbial information and action information. The scene adverbial information can be information about scenes such as exhibitions, counters, and podiums, and the action information can be information about specific robot actions such as speaking, talking, and waving. The character feature information includes robot identity setting information, which can be receptionist, movie fan, foodie, etc. The corresponding identity can be determined according to the scene where the robot is located.

[0045] It should be noted that the emotion feature information, action feature information, scene feature information, and character feature information can all be customized according to user needs. The above examples are for ease of understanding and can also be other words, which are not limited here.

[0046] Step S102, determining user behavior characteristics according to the conversation context between the robot and the user.

[0047] In this embodiment, the robot can provide an open domain dialogue system that supports multiple rounds of interaction, and conduct dialogues with the user through the open domain dialogue system. The input data of the pre-trained language model can be several rounds of historical dialogues between the robot and the user. The beginning of each round of historical dialogue usually contains the identity prompt of the speaker, such as using "A:" and "B:" for identity prompts, where A can refer to the user and B can refer to the robot. Or "User:" and "Robot:" can be used for identity prompts. The identity prompt can indicate which party participating in the dialogue provides the statement.

[0048] Specifically, user behavior characteristics can be identified based on historical conversations, that is, user behavior characteristics can be identified based on the conversation context. For example, when a robot and a user are having a conversation and the user posts a message with negative emotions, such as saying that he is unhappy today, the robot can identify the user's behavior characteristics as negative emotional characteristics.

[0049] Step S103: determining target additional feature information matching the user behavior feature from the multiple types of additional feature information.

[0050] In the prior art, if a user posts negative emotional information and there is no control, the robot may reply with the same negative emotion, for example, the robot also says that it feels pessimistic and responds negatively. In this embodiment, in order to avoid negative responses and respond positively, and to reflect positive and optimistic emotions with a relatively high probability, the target additional feature information is determined to reflect positive and optimistic emotional features. Then, based on the target additional feature information, the robot dialogue reply content is controlled to ensure that the reply content output by the robot is positive.

[0051] Step S104, using the target additional characteristic information to modify the identity prompt of the robot, and controlling the dialogue reply content of the robot according to the target additional characteristic information.

[0052] It should be noted that after the target additional feature information is used to modify the identity prompt of the robot, the dialogue context between the robot and the user also includes the corresponding target additional feature information.

[0053] In this embodiment, the robot identity prompt is modified by the target additional feature information, and the robot can control the dialogue reply content according to the target additional feature information through the pre-trained language model, thereby achieving the purpose of controlling the robot's dialogue behavior.

[0054] When the voice model generates a dialogue reply, after the historical dialogue with the target additional feature information added is input into the language model, the reply generated by the language model will reflect the emotion, action, persona or scene features corresponding to the target additional feature information with a high probability. The robot can control the dialogue reply content according to the emotion features, action features, persona features and scene features set by the user.

[0055] It should be noted that since the language model is pre-trained on a large-scale training corpus, the training corpus includes novels, news and other content. The training corpus usually contains dialogue corpus with target additional feature information, so that the language model is able to learn the relationship between the target additional feature information and the dialogue response content, thereby controlling the robot's dialogue response content according to the target additional feature information.

[0056] In one embodiment, the controlling of the robot's dialogue reply content according to the target additional feature information in step S104 includes:

[0057] In a case where the target additional characteristic information is the emotional characteristic information, the emotional characteristic corresponding to the dialogue reply content is controlled according to the emotional characteristic information.

[0058] Specifically, by modifying the robot identity prompt with an adverbial or attributive containing emotional information, it is possible to control the emotional characteristics reflected in the robot's reply content.

[0059] In one implementation, the emotion characteristic information includes emotion adverbial information or emotion attributive information, and controlling the emotion characteristic corresponding to the dialogue reply content according to the emotion characteristic information includes:

[0060] The reply content matching the emotional adverbial information or the emotional attributive information is generated through a pre-trained language model.

[0061] For example, when the robot's identity prompt is "B:", it can be modified to "B said excitedly:", "B exclaimed:", "B comforted:", etc., and the pre-trained language model is used to generate reply content that matches the emotional adverbial information such as "excited", "exclamation", and "comfort". For example, when the robot's identity prompt is "B:", it can be modified to "excited B:", "sad B:", "cheerful B:", etc., and the pre-trained language model is used to generate reply content that matches the emotional adverbial information such as "excited", "sad", and "cheerful". In this way, the emotional characteristics reflected in the robot's reply content can be controlled.

[0062] The following is an example of an application scenario of robot chat. During a conversation between a robot and a user, the user's conversation output content is "I am not happy today". The robot in this embodiment determines the user's negative emotion based on "unhappy", determines positive emotional words from a variety of pre-set additional feature information, and then modifies the robot's identity prompt based on the positive emotional words, controls the conversation reply content according to the positive emotional words, ensures that the robot's output reply is positive, and realizes the control of the robot's conversation content.

[0063] In another embodiment, the controlling the dialogue reply content of the robot according to the target additional feature information in step S104 includes:

[0064] In a case where the target additional feature information is the action feature information, the action feature corresponding to the dialogue reply content is controlled according to the action feature information.

[0065] Specifically, the robot's identity prompt is modified by including action adverbial information and verb information, thereby achieving control over the action features reflected in the conversation reply content.

[0066] In one implementation, the action feature information includes action adverbial information and action information, and the controlling of the action feature corresponding to the dialogue reply content according to the action feature information includes:

[0067] A reply content matching the action adverbial information and the action information is generated through a pre-trained language model.

[0068] For example, when the robot's identity prompt is "B:", it can be modified to "B waved and said:", "B laughed and said:", "B roared and said:", etc. The pre-trained language model generates dialogue response content that matches adverbials and verbs such as "waved", "laughed", "roared", "said", and "said".

[0069] It should be noted that action adverbials and verbs can be customized by users according to specific application scenarios. Specific application scenarios will limit the robot to perform corresponding actions. Through settings, the robot's dialogue reply content can be matched with the robot's actions, so as to achieve control of the action characteristics reflected in the dialogue reply content.

[0070] In another embodiment, the controlling the dialogue reply content of the robot according to the target additional feature information in step S104 includes:

[0071] In the case where the target additional feature information is the scene feature information, the scene feature corresponding to the dialogue reply content is controlled according to the scene feature information.

[0072] Specifically, the robot's identity prompt is modified by adverbials and verbs containing scene information, and the scene features reflected in the robot's dialogue response content are controlled based on the adverbials and verbs of the scene information.

[0073] Specifically, the scene feature information includes scene adverbial information and action information, and the controlling the scene feature corresponding to the dialogue reply content according to the scene feature information includes:

[0074] The pre-trained language model is used to generate reply content that matches the scene adverbial information and action information.

[0075] For example, when the robot's identity prompt is "B:", it can be modified to "B said at the exhibition:", "B said in front of the counter:", "B said on the podium:", etc. It should be noted that the scene adverbial information such as "in front of the counter", "at the exhibition", "on the podium" can be extracted according to the robot's location information, and can also be determined according to the scene information set by the user.

[0076] In another embodiment, the controlling the dialogue reply content of the robot according to the target additional feature information in step S104 includes:

[0077] In a case where the target additional feature information is the personality feature information, the personality feature embodied in the dialogue reply content is controlled according to the personality feature information.

[0078] In this way, the robot's identity prompt is modified through additional personality characteristic information, thereby achieving control over the personality characteristics reflected in the dialogue response content.

[0079] Specifically, the character feature information robot identity setting information, and the character feature corresponding to the control of the dialogue reply content according to the character feature information, include:

[0080] Generate response content that matches the robot’s identity setting information through a pre-trained language model.

[0081] For example, when the robot's identity prompt is "B:", it can be modified to "receptionist B:", "movie fan B:", "foodie B:", "friend B:", "psychologist B:", etc. It should be noted that the robot's personality characteristic information is determined according to the robot's application scenario. For example, in a restaurant scenario, the robot's personality characteristic information is determined to be receptionist B, and in a supermarket shopping scenario, the robot's personality characteristic information is determined to be foodie B.

[0082] In this way, the robot's dialogue reply content is controllable and the implementation process is convenient, without changing the structure of the original language model or retraining the language model. Compared with the existing technical solutions that often require the use of special corpus to retrain the language model, or modify the structure of the language model, increasing the development and deployment costs of the system, the control solution for the dialogue reply content provided by this implementation has significant advantages in terms of development and deployment costs.

[0083] The control method for the dialogue reply content provided in this embodiment obtains a variety of additional feature information; determines the user behavior characteristics according to the dialogue context between the robot and the user; determines the target additional feature information that matches the user behavior characteristics from the multiple additional feature information; uses the target additional feature information to modify the identity prompt of the robot, and controls the dialogue reply content of the robot according to the target additional feature information. Through the control scheme for the dialogue reply content provided in this embodiment, the identity prompt of the robot is modified by using the target additional feature information that matches the user behavior characteristics, and the dialogue reply content of the robot is regulated by the target additional feature information, so that the dialogue reply content of the robot is controllable, and the implementation process is convenient, without changing the structure of the original language model, and without retraining the language model. Compared with the higher development and deployment costs of existing solutions, it has significant advantages in terms of development and deployment costs.

[0084] Example 2

[0085] In addition, an embodiment of the present disclosure provides a device for controlling conversation reply content.

[0086] Specifically, Figure 2 As shown, the control device 200 for dialogue reply content provided by the embodiment of the present disclosure includes:

[0087] An acquisition module 201 is used to acquire a variety of additional feature information, wherein the additional feature information includes at least one of the following: emotion feature information, action feature information, scene feature information, and character feature information;

[0088] A determination module 202, configured to determine user behavior characteristics according to a conversation context between the robot and the user;

[0089] A matching module 203, configured to match target additional feature information corresponding to the user behavior feature from the plurality of additional feature information;

[0090] The control module 204 is used to modify the identity prompt of the robot using the target additional characteristic information, and control the dialogue reply content of the robot according to the target additional characteristic information.

[0091] In one implementation, the control module 204 is further configured to control the emotional features corresponding to the dialogue reply content according to the emotional feature information when the target additional feature information is the emotional feature information.

[0092] In one implementation, the emotional feature information includes emotional adverbial information or emotional attributive information, and the control module 204 is further configured to generate reply content matching the emotional adverbial information or the emotional attributive information through a pre-trained language model.

[0093] In one implementation, the control module 204 is further configured to control the action feature corresponding to the dialogue reply content according to the action feature information when the target additional feature information is the action feature information.

[0094] In one embodiment, the action feature information includes action adverbial information and action information, and the control module 204 is further configured to generate reply content matching the action adverbial information and the action information through a pre-trained language model.

[0095] In one implementation, the control module 204 is further configured to control the scene feature reflected in the dialogue reply content according to the scene feature information when the target additional feature information is the scene feature information.

[0096] In one embodiment, the control module 204 is further configured to control the personality characteristics embodied in the dialogue reply content according to the personality characteristic information when the target additional characteristic information is the personality characteristic information.

[0097] The apparatus 200 for controlling the conversation reply content provided in this embodiment can implement the method for controlling the conversation reply content provided in Embodiment 1, and will not be described again to avoid repetition.

[0098] The control device for the dialogue reply content provided in this embodiment obtains a variety of additional feature information; determines the user behavior characteristics according to the dialogue context between the robot and the user; determines the target additional feature information that matches the user behavior characteristics from the multiple additional feature information; uses the target additional feature information to modify the identity prompt of the robot, and controls the dialogue reply content of the robot according to the target additional feature information. Through the control scheme for the dialogue reply content provided in this embodiment, the identity prompt of the robot is modified by the target additional feature information that matches the user behavior characteristics, and the dialogue reply content of the robot is regulated by the target additional feature information, so that the dialogue reply content of the robot is controllable, and the implementation process is convenient, without changing the structure of the original language model, and without retraining the language model. Compared with the higher development and deployment costs of existing solutions, it has significant advantages in terms of development and deployment costs.

[0099] Example 3

[0100] In addition, an embodiment of the present disclosure provides an electronic device, including a memory and a processor, wherein the memory stores a computer program, and when the computer program runs on the processor, the method for controlling the conversation reply content provided in Example 1 is executed.

[0101] The electronic device provided in this embodiment can implement the method for controlling the conversation reply content provided in Embodiment 1, and will not be described again here to avoid repetition.

[0102] Example 4

[0103] The present application also provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the method for controlling the conversation reply content provided in Example 1 is implemented.

[0104] In this embodiment, the computer-readable storage medium may be a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk.

[0105] The computer-readable storage medium provided in this embodiment can implement the method for controlling the conversation reply content provided in Example 1, and will not be described again here to avoid repetition.

[0106] It should be noted that, in this article, the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, article or terminal including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or terminal. In the absence of further restrictions, an element defined by the sentence "comprises a ..." does not exclude the presence of other identical elements in the process, method, article or terminal including the element.

[0107] Through the description of the above implementation methods, those skilled in the art can clearly understand that the above-mentioned embodiment methods can be implemented by means of software plus a necessary general hardware platform, and of course by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present application, or the part that contributes to the prior art, can be embodied in the form of a software product, which is stored in a storage medium (such as ROM / RAM, a magnetic disk, or an optical disk), and includes a number of instructions for a terminal (which can be a mobile phone, a computer, a server, an air conditioner, or a network device, etc.) to execute the methods described in each embodiment of the present application.

[0108] The embodiments of the present application are described above in conjunction with the accompanying drawings, but the present application is not limited to the above-mentioned specific implementation methods. The above-mentioned specific implementation methods are merely illustrative and not restrictive. Under the guidance of the present application, ordinary technicians in this field can also make many forms without departing from the purpose of the present application and the scope of protection of the claims, all of which are within the protection of the present application.

Claims

1. A method for controlling the content of a conversation reply, characterized in that: The method comprises: Acquire multiple additional feature information, the additional feature information including at least one of the following: emotion feature information, action feature information, scene feature information, and character feature information; the action feature information includes action adverbial information and action information, the action information is information about specific actions of the robot, and the action adverbial information and the action information are combined together; Determine user behavior characteristics based on the conversation context between the robot and the user; Determining target additional feature information matching the user behavior feature from the multiple types of additional feature information; In the case where the target additional feature information is the action feature information, the action adverbial and the action information are used to modify the identity prompt of the robot, and a reply content matching the action adverbial information and the action information is generated through a pre-trained language model.

2. The method according to claim 1, characterized in that The method further comprises: In a case where the target additional characteristic information is the emotional characteristic information, the emotional characteristic corresponding to the dialogue reply content is controlled according to the emotional characteristic information.

3. The method according to claim 2, characterized in that The emotional feature information includes emotional adverbial information or emotional attributive information, and the step of controlling the emotional feature corresponding to the dialogue reply content according to the emotional feature information includes: The reply content matching the emotional adverbial information or the emotional attributive information is generated through a pre-trained language model.

4. The method according to claim 1, characterized in that The method further comprises: In the case where the target additional feature information is the scene feature information, the scene feature corresponding to the dialogue reply content is controlled according to the scene feature information.

5. The method according to claim 1, characterized in that The method further comprises: In a case where the target additional feature information is the personality feature information, the personality feature embodied in the dialogue reply content is controlled according to the personality feature information.

6. A device for controlling the content of a conversation reply, characterized in that: The device comprises: an acquisition module, used to acquire a plurality of additional feature information, wherein the additional feature information includes at least one of the following: emotion feature information, action feature information, scene feature information, and character feature information; the action feature information includes action adverbial information and action information, wherein the action information is information about a specific action of the robot, and the action adverbial information and the action information are combined together; A determination module, used to determine user behavior characteristics based on the conversation context between the robot and the user; A matching module, used for matching target additional feature information corresponding to the user behavior feature from the multiple additional feature information; A control module is used to modify the identity prompt of the robot using the action adverbial and the action information when the target additional feature information is the action feature information, and to generate reply content matching the action adverbial information and the action information through a pre-trained language model.

7. An electronic device, characterized in that: The invention comprises a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the method for controlling the dialogue reply content according to any one of claims 1 to 5 is executed.

8. A computer-readable storage medium, characterized in that: The device stores a computer program, which, when executed on a processor, executes the method for controlling the dialogue reply content according to any one of claims 1 to 5.

Images