Information processing method and related device

Abstract

The embodiment of the invention discloses an information processing method and a related device, after to-be-replied information provided by an information provider for a non-player character in a game is obtained, an information style label used for indicating the information style of the reply information provided by the non-player character can be determined firstly, and then the reply information provided by the non-player character is obtained through an information generation model. Generating reply information used for replying the to-be-replied information according to the to-be-replied information and the information style label, and providing the reply information for the information provider through the non-player character. Based on the to-be-replied information, the information generation model can enable the information content of the reply information to be used for replying the to-be-replied information, and based on the information style label, the information generation model can enable the information style of the reply information to meet the requirement of the information interaction, so that the reply information with more real and accurate information style can be generated through the mode; the diversified requirements of the diversified game information interaction process for information styles are met, and the information interaction experience of players is guaranteed.

Description

Technical Field

[0001] The present application relates to the field of data processing technology, and in particular to a testing method and related devices. Background Art

[0002] Electronic games are one of the most important forms of entertainment for people today. Players can experience a rich and diverse gaming world in electronic games. Among them, non-player characters (NPCs) are an important part of the game. Whether non-player characters can interact with players in a real and effective way is a key factor affecting the player's gaming experience.

[0003] In the related art, in order to ensure the information interaction experience between players and non-player characters, a script for information interaction can be manually written. The script sets the reply information that the player can input when interacting with the non-player characters, and sets the reply information that the non-player characters need to reply to when receiving each reply information, so that information interaction between players and non-player characters can be achieved based on the script.

[0004] However, since the information content that can be set through the script is limited, it can only support fixed-content information interaction between non-player characters and players, and it is difficult to meet diverse information interaction needs. Summary of the invention

[0005] In order to solve the above technical problems, the present application provides an information processing method that can generate reply messages with more diverse information styles and optimize the information interaction experience.

[0006] The embodiments of the present application disclose the following technical solutions:

[0007] In a first aspect, an embodiment of the present application discloses an information processing method, the method comprising:

[0008] Obtain the pending reply information provided by the information provider for the non-player characters in the game;

[0009] Determining an information style tag corresponding to the non-player character, the information style tag being used to indicate an information style of reply information provided by the non-player character in response to the information to be replied;

[0010] Generate reply information corresponding to the information to be replied according to the information to be replied and the information style label through the information generation model, wherein the reply information is used to reply to the information to be replied, and the information style of the reply information matches the information style corresponding to the information style label;

[0011] The reply information is provided to the information provider through the non-player character.

[0012] In a second aspect, an embodiment of the present application discloses an information processing method, the method comprising:

[0013] Acquire sample information to be replied and sample reply information, wherein the sample reply information is used to reply the sample information to be replied, the sample information to be replied has a corresponding sample information style label, the sample reply information corresponds to the sample information style, and the sample information style label corresponds to the sample information style;

[0014] Generate pending reply information according to the sample pending reply information and the sample information style label through the initial information generation model;

[0015] According to the difference between the pending reply information and the sample reply information, the model parameters corresponding to the initial information generation model are adjusted to obtain an information generation model, wherein the information generation model is used to generate reply information for replying to the pending reply information provided by an information provider for a non-player character in the game and an information style tag corresponding to the non-player character, wherein the information style corresponding to the reply information matches the information style corresponding to the corresponding information style tag, and the reply information is used to be provided to the information provider through the non-player character.

[0016] In a possible implementation, the derivation process includes an information generation process and a first information adjustment process, the information generation process is used to generate information for replying to the information to be replied, the first information adjustment process is used to adjust the information style corresponding to the information generated by the information generation process according to the information style label, the sample intermediate information includes sample first initial information, the sample first initial information is information generated by the information generation process, the initial information generation model includes an initial information generation module and an initial information adjustment module, and the to-be-determined reply information and the to-be-determined intermediate information corresponding to the to-be-determined reply information are generated according to the sample to-be-replied information and the sample information style label through the initial information generation model, including:

[0017] Generate pending first initial information according to the sample information to be replied by the initial information generation module, the pending first initial information is used to reply the sample information to be replied, and the pending intermediate information includes the pending first initial information;

[0018] Generate the pending reply information according to the pending first initial information and the sample information style label through the initial information adjustment module;

[0019] The step of adjusting the model parameters corresponding to the initial information generation model according to the difference between the pending reply information and the sample reply information, and the difference between the sample intermediate information and the pending intermediate information, to obtain the information generation model includes:

[0020] According to the difference between the pending first initial information and the sample first initial information, the model parameters corresponding to the initial information generation module are adjusted to obtain the information generation module, and according to the difference between the pending reply information and the sample reply information, the model parameters corresponding to the initial information adjustment module are adjusted to obtain the information adjustment module. The information generation module and the information adjustment module are used to constitute the information generation model.

[0021] In a possible implementation, the derivation process further includes a second information adjustment process, the second information adjustment process is used to adjust information content corresponding to the information generated by the first information adjustment process according to historical information in the information interaction process in which the information to be replied is located, the sample information to be replied is information in the sample information interaction process, the sample intermediate information also includes sample certificate second initial information, the sample second initial information is information generated by adjusting the sample first initial information according to the sample information style label through the first information adjustment process, and the sample reply information is information generated by adjusting the sample second initial information according to historical information in the sample information interaction process through the second information adjustment process;

[0022] The initial information adjustment module includes an initial first submodule and an initial second submodule. The initial information adjustment module generates the pending reply information according to the pending first initial information and the sample information style label, including:

[0023] Generate the pending second initial information according to the pending first initial information and the sample information style label through the initial first submodule;

[0024] Generate the pending reply information through the initial second submodule according to the historical information in the process of interaction between the pending second initial information and the sample information;

[0025] The method of adjusting the model parameters corresponding to the initial information generation module according to the difference between the pending first initial information and the sample first initial information to obtain the information generation module, and adjusting the model parameters corresponding to the initial information adjustment module according to the difference between the pending reply information and the sample reply information to obtain the information adjustment module includes:

[0026] According to the difference between the pending first initial information and the sample first initial information, adjusting the model parameters corresponding to the initial information generation module to obtain an information generation module;

[0027] According to the difference between the pending second initial information and the sample second initial information, the model parameters corresponding to the initial first submodule are adjusted to obtain the first submodule, and according to the difference between the pending reply information and the sample reply information, the model parameters corresponding to the initial second submodule are adjusted to obtain the second submodule. The first submodule and the second submodule are used to constitute the information adjustment module.

[0028] In a possible implementation, the pending reply information corresponds to a pending information style, and the method further includes:

[0029] Analyzing the pending information style label corresponding to the pending reply information through a label recognition model, wherein the pending information style label corresponds to the pending information style;

[0030] The step of adjusting the model parameters corresponding to the initial information generation model according to the difference between the pending reply information and the sample reply information to obtain the information generation model comprises:

[0031] According to the difference between the pending reply information and the sample reply information, and the difference between the pending information style label and the sample information style label, the model parameters corresponding to the initial information generation model are adjusted to obtain the information generation model.

[0032] In one possible implementation, the sample information style label includes any one or more combinations of a sample scene label, a sample character label, and a sample context label. The sample scene label is used to identify the game scene corresponding to the sample reply information, the sample character label is used to identify the character character of the non-player character corresponding to the sample reply information, and the sample context label is used to identify the context corresponding to the sample information interaction process in which the sample information to be replied is located.

[0033] In a possible implementation, the sample reply information is any one of a plurality of sample reply information, and the character personalities identified by the sample personality tags corresponding to different sample reply information are different.

[0034] In a possible implementation, the sample reply information is any one of a plurality of sample reply information, and the game scenes identified by the sample scene tags corresponding to different sample reply information are different.

[0035] In a third aspect, an embodiment of the present application discloses an information processing device, the device comprising a first acquisition unit, a first determination unit, a first generation unit and a providing unit:

[0036] The first acquisition unit is used to acquire the to-be-replied information provided by the information provider for the non-player character in the game;

[0037] The first determining unit is used to determine an information style tag corresponding to the non-player character, wherein the information style tag is used to control the information style of the reply information provided by the non-player character in response to the information to be replied;

[0038] The first generating unit is used to generate reply information corresponding to the information to be replied according to the information to be replied and the information style label through an information generation model, wherein the reply information is used to reply to the information to be replied, and the information style of the reply information matches the information style corresponding to the information style label;

[0039] The providing unit is used to provide the reply information to the information provider through the non-player character.

[0040] In a possible implementation manner, the first determining unit is specifically configured to:

[0041] Determining a game scene in which the non-player character is located;

[0042] Obtaining a scene tag for identifying the game scene;

[0043] The scene tag is determined as an information style tag corresponding to the non-player character.

[0044] In a possible implementation, the information style tag includes a personality tag corresponding to the non-player character, and the personality tag is used to identify the character personality of the non-player character.

[0045] In a possible implementation manner, the determining unit is specifically configured to:

[0046] Acquire historical information provided by the information provider before the information provider provides the information to be replied during the information interaction between the information provider and the non-player character;

[0047] Determining, according to the historical information, a context label corresponding to the information interaction process, wherein the context label is used to identify a context corresponding to the information interaction process;

[0048] The context tag is determined as an information style tag corresponding to the non-player character.

[0049] In a possible implementation manner, the device further includes a second determining unit and a first adjusting unit:

[0050] The second determining unit is used to determine the matching degree between the information style corresponding to the reply information and the information style corresponding to the information style tag;

[0051] The first adjustment unit is used to adjust the model parameters corresponding to the information generation model according to the information to be replied based on the fact that the matching degree is less than a matching degree threshold, so that the matching degree between the information style corresponding to the reply information generated by the information generation model according to the information to be replied and the information style corresponding to the information style label is not less than the matching degree threshold.

[0052] In a possible implementation, the information provider is a player character controlled by a player, the information style tag is used to display to the player, and the second determining unit is specifically used to:

[0053] Obtaining the matching degree of the player input for the reply information;

[0054] The matching degree of the player's input for the reply information is determined as the matching degree between the information style corresponding to the reply information and the information style label.

[0055] In a possible implementation, the information generation model includes an information generation module and an information adjustment module, the information generation module is used to execute an information generation process, the information adjustment module is used to execute a first information adjustment process, the information generation process is used to generate information to reply to the information to be replied, the first information adjustment process is used to adjust the information style corresponding to the information generated by the information generation process according to the information style label, and the first generation unit is specifically used to:

[0056] Generate, by the information generating module, first initial information according to the information to be replied, wherein the first initial information is used to reply to the information to be replied;

[0057] The reply information is generated according to the first initial information and the information style tag through the information adjustment module.

[0058] In a possible implementation, the information adjustment module includes a first submodule and a second submodule, the first submodule is used to execute the first information adjustment process, the second submodule is used to execute the second information adjustment process, the second information adjustment process is used to adjust the information content corresponding to the information generated by the first information adjustment process according to the historical information in the information interaction process of the to-be-replied information, and the device also includes a second acquisition unit:

[0059] The second acquisition unit is used to acquire the historical information provided by the information provider before the information provider provides the information to be replied during the information interaction between the information provider and the non-player character.

[0060] The first generating unit is specifically used for:

[0061] Generate second initial information according to the first initial information and the information style label through the first submodule;

[0062] The reply information is generated according to the historical information and the second initial information through the second submodule.

[0063] In a fourth aspect, an embodiment of the present application discloses an information processing device, the device comprising a third acquisition unit, a second generation unit, and a second adjustment unit:

[0064] The third acquisition unit is used to acquire sample information to be replied and sample reply information, wherein the sample reply information is used to reply to the sample information to be replied, the sample information to be replied has a corresponding sample information style label, the sample reply information corresponds to the sample information style, and the sample information style label corresponds to the sample information style;

[0065] The second generating unit is used to generate pending reply information according to the sample to-be-replied information and the sample information style label by using the initial information generating model;

[0066] The second adjustment unit is used to adjust the model parameters corresponding to the initial information generation model according to the difference between the pending reply information and the sample reply information to obtain the information generation model, wherein the information generation model is used to generate reply information for replying to the pending reply information provided by the information provider for the non-player character in the game and the information style tag corresponding to the non-player character, wherein the information style corresponding to the reply information matches the information style corresponding to the corresponding information style tag, and the reply information is used to be provided to the information provider through the non-player character.

[0067] In a possible implementation manner, the sample reply information has corresponding sample intermediate information, and the sample intermediate information is used to characterize a derivation process of generating the sample reply information with respect to the sample to-be-replied information, and the second generating unit is specifically used to:

[0068] Generate the pending reply information and the pending intermediate information corresponding to the pending reply information according to the sample pending reply information and the sample information style label through the initial information generation model, wherein the pending intermediate information is used to characterize the derivation process of the initial information generation model to generate the pending reply information;

[0069] The second adjusting unit is specifically used for:

[0070] According to the difference between the pending reply information and the sample reply information, and the difference between the sample intermediate information and the pending intermediate information, the model parameters corresponding to the initial information generation model are adjusted to obtain the information generation model.

[0071] In a possible implementation, the derivation process includes an information generation process and a first information adjustment process, the information generation process is used to generate information for replying to information to be replied, the first information adjustment process is used to adjust the information style corresponding to the information generated by the information generation process according to the information style label, the sample intermediate information includes sample first initial information, the sample first initial information is information generated by the information generation process, the initial information generation model includes an initial information generation module and an initial information adjustment module, and the second generation unit is specifically used to:

[0072] Generate pending first initial information according to the sample information to be replied by the initial information generation module, the pending first initial information is used to reply the sample information to be replied, and the pending intermediate information includes the pending first initial information;

[0073] Generate the pending reply information according to the pending first initial information and the sample information style label through the initial information adjustment module;

[0074] The second adjusting unit is specifically used for:

[0075] According to the difference between the pending first initial information and the sample first initial information, the model parameters corresponding to the initial information generation module are adjusted to obtain the information generation module, and according to the difference between the pending reply information and the sample reply information, the model parameters corresponding to the initial information adjustment module are adjusted to obtain the information adjustment module. The information generation module and the information adjustment module are used to constitute the information generation model.

[0076] In a possible implementation, the derivation process further includes a second information adjustment process, the second information adjustment process is used to adjust information content corresponding to the information generated by the first information adjustment process according to historical information in the information interaction process in which the information to be replied is located, the sample information to be replied is information in the sample information interaction process, the sample intermediate information also includes sample certificate second initial information, the sample second initial information is information generated by adjusting the sample first initial information according to the sample information style label through the first information adjustment process, and the sample reply information is information generated by adjusting the sample second initial information according to historical information in the sample information interaction process through the second information adjustment process;

[0077] The initial information adjustment module includes an initial first submodule and an initial second submodule, and the second generation unit is specifically used for:

[0078] Generate the pending second initial information according to the pending first initial information and the sample information style label through the initial first submodule;

[0079] Generate the pending reply information through the initial second submodule according to the historical information in the process of interaction between the pending second initial information and the sample information;

[0080] The second adjusting unit is specifically used for:

[0081] According to the difference between the pending first initial information and the sample first initial information, adjusting the model parameters corresponding to the initial information generation module to obtain an information generation module;

[0082] According to the difference between the pending second initial information and the sample second initial information, the model parameters corresponding to the initial first submodule are adjusted to obtain the first submodule, and according to the difference between the pending reply information and the sample reply information, the model parameters corresponding to the initial second submodule are adjusted to obtain the second submodule. The first submodule and the second submodule are used to constitute the information adjustment module.

[0083] In a possible implementation, the pending reply information corresponds to a pending information style, and the device further includes an analyzing unit:

[0084] The analyzing unit is used to analyze the pending information style label corresponding to the pending reply information through a label recognition model, wherein the pending information style label corresponds to the pending information style;

[0085] The second adjusting unit is specifically used for:

[0086] According to the difference between the pending reply information and the sample reply information, and the difference between the pending information style label and the sample information style label, the model parameters corresponding to the initial information generation model are adjusted to obtain the information generation model.

[0087] In one possible implementation, the sample information style label includes any one or more combinations of a sample scene label, a sample character label, and a sample context label. The sample scene label is used to identify the game scene corresponding to the sample reply information, the sample character label is used to identify the character character of the non-player character corresponding to the sample reply information, and the sample context label is used to identify the context corresponding to the sample information interaction process in which the sample information to be replied is located.

[0088] In a possible implementation, the sample reply information is any one of a plurality of sample reply information, and the character personalities identified by the sample personality tags corresponding to different sample reply information are different.

[0089] In a possible implementation, the sample reply information is any one of a plurality of sample reply information, and the game scenes identified by the sample scene tags corresponding to different sample reply information are different.

[0090] In a fifth aspect, an embodiment of the present application discloses a computer device, wherein the computer device includes a processor and a memory:

[0091] The memory is used to store a computer program and transmit the computer program to the processor;

[0092] The processor is used to execute the information processing method described in any one of the first aspects, or execute the information processing method described in any one of the second aspects according to the instructions in the computer program;

[0093] In a sixth aspect, an embodiment of the present application discloses a computer-readable storage medium, wherein the computer-readable storage medium is used to store a computer program, wherein the computer program is used to execute the information processing method described in any one of the first aspects, or execute the information processing method described in any one of the second aspects;

[0094] In the seventh aspect, an embodiment of the present application discloses a computer program product including a computer program, which, when executed on a computer device, enables the computer device to execute the information processing method described in any one of the first aspect, or execute the information processing method described in any one of the second aspect.

[0095] It can be seen from the above technical scheme that when generating reply information for a non-player character, the present application will first analyze the information style tag corresponding to the non-player character, and the information style tag is used to indicate the information style of the reply information provided by the non-player character, so that the reply information provided by the non-player character can be more in line with the information interaction needs. After obtaining the information to be replied provided by the information provider for the non-player character, the computer device can provide the information to be replied and the information style tag corresponding to the non-player character to the information generation model, so that the information generation model generates reply information based on these information. Among them, based on the information to be replied, it can be analyzed what kind of information content can be used to reply to the information to be replied, so that the reply information can be used to reply to the information to be replied; based on the information style tag, it can be analyzed how to adjust the information so that the information style corresponding to the reply information matches the information style corresponding to the information style tag, so that the reply information can simultaneously meet the information content and information style required for information interaction, and then when the reply information is provided to the information provider through the non-player character, it can bring a more real and accurate information interaction experience to the information provider. At the same time, through this method, when designing non-player characters in the game, it is only necessary to set a variety of information style tags for the non-player characters to enable the non-player characters to have the ability to reply to information in various styles. Different information style tags can also be set for different game scenes and different game processes, so that non-player characters can reply to information with matching information styles in different game scenes, different information interaction contexts and different role settings, further improving the authenticity and diversity of the player's gaming experience and enhancing the immersion brought by the game. BRIEF DESCRIPTION OF THE DRAWINGS

[0096] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0097] Figure 1 A schematic diagram of an information processing method in a practical application scenario provided by an embodiment of the present application;

[0098] Figure 2 A flowchart of an information processing method provided in an embodiment of the present application;

[0099] Figure 3 A flowchart of an information processing method provided in an embodiment of the present application;

[0100] Figure 4 A flowchart of an information processing method provided in an embodiment of the present application;

[0101] Figure 5 A flowchart of an information processing method provided in an embodiment of the present application;

[0102] Figure 6 A flowchart of an information processing method provided in an embodiment of the present application;

[0103] Figure 7 A flowchart of an information processing method provided in an embodiment of the present application;

[0104] Figure 8 A schematic diagram of an information processing method provided in an embodiment of the present application;

[0105] Fig. 9 A flowchart of an information processing method provided in an embodiment of the present application;

[0106] Fig.10 A flowchart of an information processing method provided in an embodiment of the present application;

[0107] Fig.11 A schematic diagram of an information processing method provided in an embodiment of the present application;

[0108] Fig.12 A flowchart of an information processing method provided in an embodiment of the present application;

[0109] Fig.13 A flowchart of an information processing method provided in an embodiment of the present application;

[0110] Fig.14 A schematic diagram of an information processing method provided in an embodiment of the present application;

[0111] Fig.15 A schematic diagram of an information processing method provided in an embodiment of the present application;

[0112] Fig.16 A flowchart of an information processing method provided in an embodiment of the present application;

[0113] Fig.17 A flowchart of an information processing method provided in an embodiment of the present application;

[0114] Fig.18 A flowchart of an information processing method in a practical application scenario provided by an embodiment of the present application;

[0115] Fig.19 A structural block diagram of an information processing device provided in an embodiment of the present application;

[0116] Fig. 20 A structural block diagram of an information processing device provided in an embodiment of the present application;

[0117] Fig.21 A structural diagram of a terminal provided in an embodiment of the present application;

[0118] Fig. 22 A structural diagram of a server provided in an embodiment of the present application. DETAILED DESCRIPTION

[0119] The embodiments of the present application are described below in conjunction with the accompanying drawings.

[0120] Interacting with non-player characters in the game is a common way to play games. For example, when players want to accept game tasks and participate in game matches, they need to control the player character to interact with the corresponding non-player characters; when experiencing the game plot, they need to control the player character to interact with the non-player characters in the plot to promote the development of the game plot, etc. Therefore, whether the player can interact with non-player characters in a real and accurate manner is the key to ensuring the player's gaming experience.

[0121] In the related art, the reply information provided by non-player characters during the information interaction process is mainly generated through two methods: script writing and model generation. Among them, through script writing, the information to be replied that the player can select when interacting with information can be pre-set, and the reply information corresponding to each information to be replied can be set. After the player selects the information to be replied, the non-player character can automatically provide the reply information corresponding to the information to be replied. When information interaction is performed through model generation, the computer device can input the acquired information to be replied into the information generation model, and generate the reply information for replying to the information to be replied through the information generation model.

[0122] However, the script writing method requires manual setting of a large number of pending reply messages and reply messages, which is difficult to cover the information interaction process involved in each game. It takes a lot of time and effort, resulting in low game development efficiency, and the information diversity in the script is poor, which easily leads to relatively single information interaction content and difficulty in bringing a good information interaction experience. Although the model generation method can generate diverse reply messages, it only focuses on the information content of the pending reply message itself, and does not pay attention to the information style required for the information interaction process, resulting in the information style of the reply message not matching the information interaction needs, and thus it is difficult to bring a real and accurate information interaction experience to players.

[0123] In order to solve the above technical problems, the present application provides an information processing method. After obtaining the information to be replied provided by the information provider for the non-player character in the game, the information style tag used to indicate the information style of the reply information provided by the non-player character can be determined first, and then the reply information for replying to the information to be replied is generated according to the information to be replied and the information style tag through the information generation model, and the reply information is provided to the information provider through the non-player character. Based on the information to be replied, the information generation model can make the information content of the reply information able to be used to reply to the information to be replied, and based on the information style tag, the information generation model can make the information style of the reply information meet the needs of this information interaction, so that in this way, the reply information with more real and accurate information style can be generated, which meets the diverse needs of the diverse game information interaction process for information style and ensures the player's information interaction experience.

[0124] It is understandable that the method can be applied to a computer device, which is a computer device capable of performing information processing, such as a terminal device or a server. The method can be executed independently by a terminal device or a server, or it can be applied to a network scenario in which a terminal device and a server communicate, and is executed in cooperation with a terminal device and a server. Among them, the terminal device can be a mobile phone, a tablet computer, a laptop computer, a desktop computer and other devices. The terminal device can also include a variety of virtual reality devices, such as augmented reality (AR) devices, such as AR glasses, AR screens and other devices, and virtual reality (VR) technology (VR) devices, such as head-mounted VR glasses and other devices. The server can be understood as an application server, or a Web server. In actual deployment, the server can be an independent server, a cluster server, or a cloud server.

[0125] In order to facilitate understanding of the technical solution provided by this application, the information processing method provided by this application will be introduced below in combination with an actual application scenario.

[0126] See also Figure 1 , Figure 1 A schematic diagram of an information processing method in an actual application scenario provided in an embodiment of the present application. In this actual application scenario, the computer device may be a server 101 with information processing capabilities. Figure 1 The game screen in shows a battle scene in the game. The information provider is the player character controlled by the player. The player can instruct the player character to fight with the non-player character. When the player character wins, the non-player object will be provided with the message "I won!" as a waiting reply message.

[0127] After obtaining the information to be replied provided by the information provider for the non-player character, the server 101 will first determine the information style tag corresponding to the non-player character. The information style tag is used to indicate the information style of the reply information provided by the non-player character. In this actual application scenario, the information style tag includes a scene tag for identifying the game scene in which the non-player character is located and a character tag for identifying the character personality of the set non-player character. The scene tag is "battle scene" and the character tag is "serious personality". If the reply information is generated directly based on the information to be replied without referring to the information style tag, a blessing message such as "Congratulations!" may be generated as a reply message, which is obviously not suitable for the battle scene; by inputting the information to be replied and the information style tag into the information generation model, the information generation model can generate information whose content can be used to reply to the information to be replied, and on the other hand, the generated information can match the information style corresponding to the information style tag.

[0128] For example, the reply message "Damn it, I will definitely beat you next time!" generated in this way has information content that can be used to reply to the reply message "I won!" On the other hand, the reply message has a serious and confrontational information style, so it matches the information style corresponding to the information style tags "battle scene" and "serious personality", which is more suitable for the information replied by the non-player character in this game scene. Therefore, by providing the reply information to the information provider through the non-player character, the player can experience a more realistic information interaction experience in the battle scene, which can enhance the immersion of the game and bring a better game experience to the player.

[0129] Next, the technical solution provided by this application will be introduced in detail with reference to the accompanying drawings.

[0130] See also Figure 2 , Figure 2 A flowchart of an information processing method provided in an embodiment of the present application. In this embodiment, the computer device may be any computer device having an information processing function. The method includes:

[0131] S201: Obtaining the information to be replied to provided by the information provider for the non-player character in the game.

[0132] The information interaction process targeted by this application is mainly the information interaction process for non-player characters in the game, and the information provider can be any provider that can provide information to non-player characters, such as players, player characters controlled by players in the game, other non-player characters in the game, etc. For example, a player can control a player character to talk to a non-player character to provide information to be replied; multiple non-player characters in the game can also automatically talk to each other and provide each other with information to be replied.

[0133] S202: Determine the information style tag corresponding to the non-player character.

[0134] In the present application, in order to make the reply information provided by the non-player character to the information provider more real and accurate, so as to bring a better information interaction experience, the present application provides a technology for indicating the information style of the generated information based on the information style tag. The computer device can first determine the information style tag corresponding to the non-player character. The information style tag is used to indicate the information style of the reply information provided by the non-player character to the information to be replied. The information style corresponding to the information style tag is the information style that fits this information interaction. There are many ways to determine the information style tag, which will be introduced in detail below and will not be repeated here. Different non-player objects can correspond to different information style tags, and the same non-player object can also correspond to different information style tags in different game scenes and different information interaction processes to enhance the matching degree and diversity of the information style, which will not be repeated here.

[0135] Among them, the label content of the information style tag is not necessarily the information style itself, but may also be the label content with the corresponding information style. For example, the label content of the information style tag can be directly the information style itself such as "serious style", "humorous style", "brave style", etc., or it can be "battle scene", "humorous personality" and other content that can be analyzed to obtain the corresponding information style. For example, the information style corresponding to the battle scene is usually "serious style" and "brave style", and the information style corresponding to the "humorous personality" is usually "humorous style". A variety of information style tags are also provided below, which will not be repeated here.

[0136] S203: Generate reply information corresponding to the information to be replied to according to the information to be replied and the information style label through the information generation model.

[0137] The information generation model in the present application is used to input the information to be replied and the information style label, and generate a reply message having information content for replying to the information to be replied, and the corresponding information style matches the information style corresponding to the information style label. On the one hand, through the information to be replied, the information generation model can analyze how to generate information that can be used to reply to the information to be replied, first of all, to ensure the effectiveness of the generated reply information in terms of information content; on the other hand, through the information style label, the information generation model can adjust the information style corresponding to the generated information, so that the information style of the generated reply information matches the information style corresponding to the information style label, so as to meet the information style requirements of this information interaction. The reply information finally obtained is used to reply to the information to be replied. Among them, the way to adjust the information style corresponding to the information can include multiple ways, such as adding some modal particles in the information or adjusting the information content expression method corresponding to the information, etc., which are not limited here.

[0138] S204: Providing reply information to the information provider through the non-player character.

[0139] The computer device can provide the reply information to the information provider through the non-player character in the game to complete the information exchange. From the perspective of the information provider, the reply information can also be regarded as the information to be replied to provided by the non-player character for subsequent information exchange.

[0140] As can be seen from the above scheme, when generating reply information, the present application will provide the information to be replied to provided by the non-player character and the information style label corresponding to the non-player character to the information generation model at the same time, so that the reply information generated by the information generation model can simultaneously meet the information content and information style required for information interaction, and then when the reply information is provided to the information provider through the non-player character, it can bring a more realistic and accurate information interaction experience to the information provider. At the same time, through this method, when designing non-player characters in the game, it is only necessary to set a variety of information style labels for the non-player characters, so that the non-player characters can have information reply capabilities with diverse styles. Different information style labels can also be set for different game scenes and different game processes, so that the non-player characters can reply to information with matching information styles in different game scenes, different information interaction contexts and different role settings, further improving the authenticity and diversity of the player's game experience and enhancing the immersion brought by the game.

[0141] Next, the technical details involved in this application will be introduced in detail with reference to the accompanying drawings.

[0142] First, we introduce how to train the above information generation model. Figure 3 , Figure 3 A flowchart of an information processing method provided in an embodiment of the present application. In this embodiment, the computer device may be any computer device that can be used for model training. The method includes:

[0143] S301: Obtain sample pending reply information and sample reply information.

[0144] The sample information to be replied can be any information to be replied, that is, any information that can have reply information. The sample reply information is used to reply the sample information to be replied, the sample information to be replied has a corresponding sample information style label, the sample reply information corresponds to the sample information style, and the sample information style label corresponds to the sample information style. That is, if the information generation model can generate relatively accurate reply information based on the sample information to be replied and the sample information style label, then the reply information should have a high similarity with the sample reply information.

[0145] S302: Generate pending reply information according to the sample pending reply information and the sample information style label through the initial information generation model.

[0146] The initial information generation model may be any model capable of generating information, such as a large language model (LLM). The initial information generation model may analyze the information content required for generating information based on the sample information to be replied, so that the generated information can be used to reply the sample information to be replied, and may adjust the information style corresponding to the information according to the sample information style label, so that the information style corresponding to the final generated reply information can match the sample information style label.

[0147] S303: According to the difference between the pending reply information and the sample reply information, the model parameters corresponding to the initial information generation model are adjusted to obtain the information generation model.

[0148] It can be seen from the above analysis process that if the initial information generation model has strong model capabilities in the two dimensions of constructing the information content of the pending reply information based on the sample pending reply information and adjusting the information style of the pending reply information based on the sample information style label, then the final generated pending reply information should have a high similarity with the sample reply information. On the contrary, if the difference between the pending reply information and the sample reply information is greater, it means that the information generation capability of the initial information generation model is worse. Based on this, the computer device can adjust the model parameters corresponding to the initial information generation model according to the difference between the pending reply information and the sample reply information, so that the information difference between the generated pending reply information and the sample reply information is reduced to the desired degree, thereby obtaining an information generation model, which can be used to perform the above process, according to the information to be replied provided by the information provider for the non-player character in the game and the information style label corresponding to the non-player character, to generate reply information for replying to the information to be replied, the information style corresponding to the reply information matches the information style corresponding to the corresponding information style label, and the reply information is used to provide to the information provider through the non-player character.

[0149] Through the above technical solution, on the one hand, the initial information generation model can learn how to accurately analyze the information content required for the generated reply information based on the information to be replied, so that the generated reply information can be used to accurately reply to the information to be replied; on the other hand, the initial information generation model can learn how to accurately adjust the information style corresponding to the information based on the information style label, so that the reply information finally generated can have an information style that matches the information style corresponding to the information style label. Therefore, the information generation model trained in this way can support non-player objects to provide reply information with diverse styles, high authenticity and high accuracy in the game, and improve the player's information interaction experience in the game.

[0150] Next, various information style tags that may be involved in this application will be introduced.

[0151] The first type: scene label

[0152] In real life, people may speak in different styles in different scenarios. For example, in work scenarios, people usually speak in a more serious style; in entertainment scenarios, people usually speak in a more relaxed and humorous style. Based on this, in order to improve the authenticity of the information provided by the non-player characters in the game, the computer device can instruct the non-player characters to provide reply information in different styles in different game scenarios, so that the information style of the reply information is more in line with the game scene, and the authenticity of the information interaction in the game is improved.

[0153] In one possible implementation, see Figure 4 When executing step S202, the computer device may execute steps S2021-S2023 (not shown in the figure), where steps S2021-S2023 are a possible implementation of step S202, including:

[0154] S2021: Determine the game scene where the non-player character is located.

[0155] The game scenes where the non-player characters are located may include multiple types, such as non-combat scenes, combat scenes, etc. The computer device may determine the game scenes where the non-player characters are located in multiple ways, such as by determining the game scenes where the non-player characters are located through information such as the character status of the non-player characters and the character position of the non-player characters, or the computer device may analyze the context information corresponding to the current game scene, such as relevant plot information, dialogue information, etc., to analyze the game scenes where the non-player characters are located, which is not limited here.

[0156] S2022: Obtain a scene tag for identifying a game scene.

[0157] The computer device may set corresponding scene tags for various game scenes, and the scene tags themselves are used to identify the game scenes. For example, when the game character is in a combat state, the computer device may determine that the non-player character is in a combat scene, and the computer device may obtain "combat scene" as the scene tag.

[0158] S2023: Determine the scene tag as an information style tag corresponding to the non-player character.

[0159] By using scene labels as information style labels, the information generation model can understand the game scenario in which the reply information is provided when generating a reply message, so that the information style corresponding to the reply information can be adjusted based on the scene labels, so that the information style corresponding to the reply information matches the information style corresponding to the scene labels, and then the information style of the reply information can fit the game scenario in which the non-player character is located, thereby improving the scene matching degree of the reply information, and can bring players a more realistic and immersive gaming experience.

[0160] For example, when the computer device analyzes that the non-player character is in a battle scene based on the character status of the non-player character, it can determine "battle scene" as a scene label, which corresponds to a serious and tense information style. The computer device can then adjust the information style of the reply information to make the reply information correspond to the serious and tense information style, thereby better fitting the information interaction experience in the battle scene.

[0161] The second type: personality label

[0162] In real life, people with different personalities have different speaking styles. For example, people with humorous personalities speak more humorously, and people who are easily excited speak more quickly and irritably. Based on this, in order to bring players a more diverse and realistic gaming experience, the computer device can set diverse character personalities for non-player characters in the game, so that players can have a rich information interaction experience when interacting with each other.

[0163] In this implementation, the information style tag may include a personality tag corresponding to the non-player character, and the personality tag is used to identify the character personality of the non-player character, such as humor, seriousness, irritability, and peace. Therefore, according to the personality tag, the information generation model can analyze how to adjust the information style of the reply information, so that the reply information can be close to the information interaction habits of the non-player character with the personality identified by the personality tag, and then the reply information can have an information style that matches the information style corresponding to the personality tag. Through the reply information, the player can have an information interaction experience that is more matched with the character personality set by the non-player character, and then the character personality set by the non-player character can be more accurately simulated, thereby improving the realism and richness of the game experience.

[0164] For example, when the personality tag of the non-player character is "humorous", the computer device can add some humorous information in the reply message to make the corresponding information style of the reply message more humorous; when the personality tag of the non-player character is "easily excited", the computer device can add more interjections that express excitement in the reply message to make the corresponding information style of the reply message more exciting.

[0165] The third type: context label

[0166] In real-life scenarios, people's speaking styles usually change in different chat contexts, such as when chatting about different information content or chatting with people with different chat styles. For example, when chatting about more casual topics such as sports and games, people's speaking styles are usually more relaxed and lively; when chatting about more serious topics such as work, study, science and technology, people's speaking styles are usually more serious and solemn. Based on this, in order to enable non-player characters to further simulate people's speaking styles in different contexts to bring players a more realistic information interaction experience, the computer device can also provide the information generation model with corresponding context labels to adjust the information style.

[0167] In one possible implementation, see Figure 5 When executing step S202, the computer device may execute steps S2024-S2026, where steps S2024-S2026 are a possible implementation of step S202, including:

[0168] S2024: Obtain historical information provided by the information provider before the information provider provides the information to be replied to during the information interaction process between the information provider and the non-player character.

[0169] In order to understand the context of information interaction between the information provider and the non-player character, the computer device may first determine the information interaction process in which the information to be replied is located, and the information interaction process is the information interaction process that the reply information needs to participate in. The information interaction process refers to the interaction process consisting of the information to be replied and the reply information, which may include various types, such as various dialogues in the game, letter interactions, etc.

[0170] Since the context corresponding to the information interaction process is composed of the information that has been provided in the information interaction process, the computer device can obtain historical information before the information provider provides the information to be replied to. The historical information includes the information to be replied and the reply information provided by the information provider and the non-player characters before the information provider provides the information to be replied to. This information can constitute the context of the information interaction process.

[0171] S2025: Determine the context label corresponding to the information interaction process based on the historical information.

[0172] The computer device can analyze the context corresponding to the information interaction process based on these historical information, so as to determine the context label corresponding to the information interaction process, and the context label is used to identify the context corresponding to the information interaction process. The context in this application refers to the information represented by the information provided in the information interaction process, which can cover multiple aspects of information, for example, it can include the information theme corresponding to the information interaction process (such as sports, learning, work, combat and other themes), the emotion corresponding to the information interaction process (such as the information provider and the non-player character interacting peacefully or fiercely), etc., which is not limited here.

[0173] S2026: Determine the context tag as the information style tag corresponding to the non-player character.

[0174] By determining the context label as the information style label, the information generation model can understand the context of the information interaction process in which the reply information is to participate, so that it can know how to adjust the information style corresponding to the reply information, so that the information style corresponding to the reply information can be more in line with the context of the information interaction process, so that when the reply information is provided during the information interaction process, the entire information interaction process can be made more natural and real, further optimizing the player's information interaction experience.

[0175] For example, when the message to be replied is "How was your game?", if the context formed by the historical information is not analyzed, the information generation model may find it difficult to understand what aspect of the level is being asked about in the reply message, and thus it may be difficult to provide reply information with a corresponding appropriate information style. However, the context analyzed in combination with the historical information can be used to obtain the specific information asked in the reply message. For example, the context label can be "Chatting about basketball topics", then the computer device can know that the message to be replied is asking about the level of basketball, and thus the information generation model can adjust the information style of the reply message to a more lively and relaxed information style to fit the topic corresponding to the information interaction process.

[0176] In addition to adjusting the information style based on the above-mentioned multiple information style tags, the computer device can also analyze the player's preferences based on the information generated by the player's information interaction in the historical period, determine which information styles of non-player characters the player likes to interact with, and thus adjust the reply information to the information style preferred by the player.

[0177] When using the information generation model to achieve information interaction in the game, improper adjustment of the information style or misunderstanding of the information content of the information to be replied may lead to abnormal situations where the generated reply information cannot meet the information interaction requirements. In order to continuously improve the quality of the reply information generated by the information generation model, thereby continuously optimizing the player's information interaction experience, the computer device can also continuously analyze the quality of the reply information generated by the information generation model, and optimize the information generation model based on the information quality.

[0178] See also Figure 6 , the computer device may further perform steps S205-S206, including:

[0179] S205: Determine the degree of match between the information style corresponding to the reply information and the information style corresponding to the information style tag.

[0180] Among them, the method of determining the matching degree mainly includes two aspects. On the one hand, the computer device can analyze the information style of the reply information and the information style corresponding to the information style label by itself, and determine the matching degree between the information style corresponding to the reply information and the information style corresponding to the information style label; on the other hand, when the information provider involves players, the players can also judge the matching degree between the information styles, thereby reflecting the players' true feelings about the information interaction experience.

[0181] S206: Based on the matching degree being less than the matching degree threshold, the model parameters corresponding to the information generation model are adjusted according to the information to be replied, so that the matching degree between the information style corresponding to the reply information generated by the information generation model according to the information to be replied and the information style corresponding to the information style label is not less than the matching degree threshold.

[0182] It can be seen from the above content that the greater the matching degree, the stronger the generation ability of the information generation model based on the information style label for information generation, and vice versa. Therefore, the computer device can preset a matching degree threshold, which is used to measure whether there is a high matching degree between the reply information and the information style corresponding to the information style label. Based on the matching degree is not less than the matching degree threshold, it means that the matching degree is high, and high-quality reply information can be generated at this time, and no adjustment is required. Based on the matching degree is less than the matching degree threshold, it means that the matching degree is low, and the quality of the reply information is poor. The computer device can continuously adjust the model parameters corresponding to the information generation model, and continuously analyze the information style matching degree of the reply information generated by the information generation model according to the information to be replied during the adjustment process, so that the matching degree can be continuously improved until the matching degree is not less than the matching degree threshold.

[0183] In this way, the computer device can dynamically adjust the model parameters of the information generation model based on the information generation effect of the information generation model in actual applications, so that the information generation model can maintain a relatively high-quality information generation capability and meet the diverse information style requirements. At the same time, when there is a need to generate reply information for a new information style in the game, by dynamically adjusting the model parameters, the information generation model can also quickly learn how to generate information for the new information style requirements, so that the information generation model can be applied to a wider range of game application scenarios, bringing players a more comprehensive information interaction experience. In addition, adjusting parameters on the existing information generation model can retain the model knowledge learned by the model during the training process, thereby ensuring the stability of the information generation function.

[0184] Among them, there may be multiple ways to adjust the model parameters. For example, the computer device may adopt a reinforcement learning method to punish reply information with a low matching degree, that is, set a lower model score for reply information with a low matching degree, adjust the model parameters based on the model score, and finally output more accurate reply information; or, the computer device may adopt a direct policy optimization method to directly adjust the policy function corresponding to the model to optimize the reply information generation effect of the model in specific tasks or scenarios; in addition, the computer device may also use reply information with a low matching degree as a negative sample and reply information with a high matching degree as a positive sample, and perform negative sample error correction training on the information generation model in combination with the positive and negative samples, so that the reply information generated by the information generation model can be far away from the negative sample and close to the positive sample, so as to improve the matching degree of the generated reply information.

[0185] As mentioned above, the present application can adjust the model based on the matching degree judged by the player. In a possible implementation, the information provider can be a player character controlled by the player, such as a human character or a non-human character controlled by the player in the game, and the information style label can be used to display to the player, that is, the player can know the information style label based on which the reply information is generated, so as to judge the matching degree between the reply information generated by the information generation model and the information style label in the corresponding information style.

[0186] See also Figure 7 When executing step S205, the computer device may execute steps S2051-S2052, where steps S2051-S2052 are a possible implementation of step S205, including:

[0187] S2051: Obtain the matching degree of the player's input for the reply message.

[0188] The matching degree is the matching degree input by the player for matching degree analysis based on the information style tag and the reply information. The input method may include multiple interactive methods, which are not limited here. Since it is the matching degree input by the player, the matching degree can more accurately reflect the player's satisfaction with the information style corresponding to the reply information, and thus can more accurately reflect the player's satisfaction with the information generation capability of the information generation model.

[0189] S2052: Determine the matching degree of the player's input for the reply information as the matching degree between the information style corresponding to the reply information and the information style label.

[0190] Since the matching degree can truly reflect the player's satisfaction with the information generation ability of the information generation model, adjusting the model parameters based on the matching degree can enable the model knowledge learned by the information generation model to more effectively enhance the player's information interaction experience and provide players with more realistic and highly matched reply information.

[0191] Next, the generation process of the information generation model to generate reply information will be introduced in detail.

[0192] See also Figure 8 In a possible implementation, the information generation model may include an information generation module and an information adjustment module. The information generation module is used to execute the information generation process, and the information adjustment module is used to execute the first information adjustment process. The information generation process is used to generate information for replying to the information to be replied, that is, to analyze the information content of the information to be replied, so that the information content of the generated information can be used to reply to the information to be replied; the first information adjustment process is used to adjust the information style corresponding to the information generated by the information generation process according to the information style label, that is, through information adjustment, the information style corresponding to the information can match the information style corresponding to the information style label.

[0193] See also Fig. 9 When executing step S203, the computer device may execute steps S2031-S2032, where steps S2031-S2032 are a possible implementation of step S203, including:

[0194] S2031: Generate first initial information according to the information to be replied through the information generation module.

[0195] The information generation module can analyze the information content of the information to be replied, so as to analyze what information content can be used to reply the information to be replied, and generate the first initial information based on this part of the information content. The first initial information can be used to reply the information to be replied, and has a more reasonable information content and information structure. For example, when the information to be replied is "You win", the first initial information can be "Congratulations" or "I lose".

[0196] S2032: Generate reply information according to the first initial information and the information style tag through the information adjustment module.

[0197] The information adjustment module can analyze the information style corresponding to the information style tag, and thus analyze how to adjust the first initial information so that the first initial information has the information style corresponding to the information style tag, so as to generate the final reply information. For example, when the first initial information is "I lost", and the information style tags are "battle scene" and "excited personality", the reply information output after adjustment by the information adjustment module can be "I lost! I am not willing!! I will definitely beat you next time!!!!", so that it can fit the information interaction method in the battle scene, and fit the information interaction method of non-player characters with excited personalities.

[0198] Through the above two models, the computer device can output reply information that meets the information interaction requirements in terms of information content and information style, so that the reply information can bring a better information interaction experience. At the same time, by carefully splitting the information generation process into multiple processes, the information analysis process can be refined, so that more accurate and reasonable reply information can be generated.

[0199] Next, we will introduce how to train an information generation model with the above model architecture.

[0200] First, it can be seen from the above content that in order to improve the quality of information generation, the information generation model can refine the generation process of the reply information. Based on this, during model training, in order to further enrich the model knowledge learned by the information generation model for generating reply information, the computer device can not only adjust the model parameters based on the difference of the reply information finally output, but also adjust the parameters based on the difference of the information generated in the intermediate process of generating the reply information.

[0201] In a possible implementation, the sample response information has corresponding sample intermediate information, which is used to characterize the derivation process of generating the sample response information based on the sample pending response information, that is, it can characterize how the model should process information in each link when generating the sample response information based on the pending response information. Fig.10 When executing step S302, the computer device may execute step S3021, which is a possible implementation of step S302, including:

[0202] S3021: Generate pending reply information and pending intermediate information corresponding to the pending reply information through an initial information generation model according to sample pending reply information and sample information style labels.

[0203] The pending intermediate information is used to characterize the derivation process of the initial information generation model generating the pending reply information. For example, the pending intermediate information may be information generated by the initial information generation model in the process of generating the pending reply information.

[0204] When executing step S303, the computer device may execute step S3031. Step S3031 is a possible implementation of step S303, including:

[0205] S3031: According to the difference between the pending reply information and the sample reply information, and the difference between the sample intermediate information and the pending intermediate information, the model parameters corresponding to the initial information generation model are adjusted to obtain the information generation model.

[0206] Since the sample reply information is accurate and reasonable reply information generated for the sample reply information and the sample information style, the sample intermediate information can characterize the accurate derivation method for reply information generation based on the sample reply information and the sample information style label, that is, the sample intermediate information is the information that should be generated in the derivation process when the reply information is accurately generated based on the sample reply information and the sample information style. Therefore, the difference between the sample intermediate information and the pending intermediate information can more accurately characterize the derivation ability of the initial information generation model in the process of generating reply information. The greater the difference, the worse the derivation ability of the initial information generation model in the process of generating reply information. Furthermore, when adjusting the model parameters, the computer device can adjust the parameters based on the difference between the pending reply information and the sample reply information on the one hand, so that the model learns how to generate accurate reply information; on the other hand, the computer device can adjust the parameters based on the difference between the pending reply information and the sample reply information on the other hand, so that the model learns how to accurately derive based on the sample reply information and the sample information style label to generate accurate reply information. To sum up, through the model training method of the present application, the model's ability to generate reply information can be effectively trained from two dimensions: information generation results and information generation process, thereby improving the accuracy of reply information generation from both the results and process dimensions.

[0207] Next, a variety of specific derivation processes will be introduced in detail.

[0208] First, in a possible implementation, as described above, the derivation process may include an information generation process and a first information adjustment process, the information generation process is used to generate information to reply to the information to be replied, and the first information adjustment process is used to adjust the information style corresponding to the information generated by the information generation process according to the information style label. At this time, the intermediate output of the derivation process of the information generation model to generate the reply information is the output of the information generation process, so the sample intermediate information may include the sample first initial information, and the sample first initial information is the information generated by the information generation process.

[0209] See also Fig.11 The initial information generation model may include an initial information generation module and an initial information adjustment module, wherein the initial information generation module is used to learn how to perform the information generation process, and the initial information adjustment module is used to learn how to perform the first information adjustment process. Fig.12 When executing step S3021, the computer device may execute steps S30211-S30212, where steps S30211-S30212 are a possible implementation of step S3021, including:

[0210] S30211: Generate pending first initial information based on the sample information to be replied through the initial information generation module.

[0211] The pending first initial information is used to reply to the sample information to be replied, and the pending intermediate information includes the pending first initial information. The initial information module can analyze what information content can be used to reply to the sample information to be replied based on the information content corresponding to the sample information to be replied, thereby obtaining the pending first initial information.

[0212] S30212: Generate pending reply information according to the pending first initial information and the style label of the sample information through the initial information adjustment module.

[0213] The initial information adjustment module can analyze how to adjust the information style corresponding to the pending first initial information based on the information style required by the sample information style label analysis, so as to match the information style corresponding to the sample information style label and obtain the pending reply information.

[0214] When executing step S3031, the computer device may execute step S30311. Step S30311 is a possible implementation of step S301, including:

[0215] S30311: According to the difference between the pending first initial information and the sample first initial information, the model parameters corresponding to the initial information generation module are adjusted to obtain the information generation module; and according to the difference between the pending reply information and the sample reply information, the model parameters corresponding to the initial information adjustment module are adjusted to obtain the information adjustment module.

[0216] Through the above analysis, it can be seen that the difference between the pending first initial information and the sample first initial information can characterize the accuracy and effectiveness of the information content generated by the initial information generation module when generating information for replying to the sample information to be replied, and the difference is inversely correlated with the accuracy. Therefore, the computer device can adjust the model parameters corresponding to the initial information generation module based on the difference, so that the initial information generation module can learn how to accurately generate information that can be used to reply to the sample information to be replied, so that the information generation module has the ability to accurately execute the information generation process.

[0217] Similarly, the difference between the pending intermediate information and the sample intermediate information can characterize the accuracy of the initial information adjustment module when adjusting the information style based on the sample information style label, and the difference is inversely correlated with the accuracy. Therefore, the computer device can adjust the model parameters corresponding to the initial information adjustment module based on the difference, so that the initial information adjustment module can learn how to accurately adjust the information style based on the information style label, while retaining the information's ability to reply to the information to be replied, so that the obtained information adjustment module has the ability to accurately execute the first information adjustment process. The information generation module and the information adjustment module are used to form an information generation model.

[0218] In addition, in real-life scenarios, the information in an information interaction process is usually relatively coherent in content, and the information content before and after is highly correlated. For example, a conversation usually discusses the information content of a topic, etc. Based on this, in a possible implementation, in order to make the information content of the generated reply information more consistent with the corresponding information interaction process, in addition to adjusting the information based on the information style tag, the computer device can also adjust the information content of the reply information based on the historical information in the information interaction process corresponding to the reply information, so that its information content is more consistent with the information content of the information interaction process, thereby making the entire information interaction process more fluent, accurate and authentic.

[0219] In this implementation, the information adjustment module in the information generation model may include a first submodule and a second submodule, wherein the first submodule is used to execute the above-mentioned first information adjustment process, and the second submodule is used to execute the second information adjustment process, and the second information adjustment process is used to adjust the information content corresponding to the information generated by the first information adjustment process according to the historical information in the information interaction process of the information to be replied, and the historical information is the information that appeared in the information interaction process in the historical period, which can characterize the context, information content, etc. corresponding to the information interaction process.

[0220] See also Fig.13 The computer device may further perform step S1301, including:

[0221] S1301: Acquire historical information provided by the information provider before the information provider provides the information to be replied during the information interaction process between the information provider and the non-player character.

[0222] Similar to the above-mentioned method of analyzing context tags, the computer device can first obtain historical information of the information interaction process in which the information to be replied is located. The information interaction process is an information interaction process that requires reply information to be provided. These historical information may include the information to be replied and the reply information provided during the information interaction process before the information provider provides the information to be replied. The provider can be any party participating in the information interaction process.

[0223] When executing step S2032, the computer device may execute steps S20321-S20322, where steps S20321-S20322 are a possible implementation of step S2032, including:

[0224] S20321: Generate second initial information according to the first initial information and the information style label through the first submodule.

[0225] like Fig.14 As shown, the first submodule can analyze the information style corresponding to the information style label, and thus analyze how to adjust the first initial information so that the first initial information has the information style corresponding to the information style label, so as to generate the second initial information. The second initial information satisfies the requirements for replying to the information to be replied in terms of information content, and matches the information style corresponding to the information style label in terms of information style.

[0226] S20322: Generate reply information according to the historical information and the second initial information through the second submodule.

[0227] The second submodule can analyze the information content corresponding to the information interaction process according to the historical information, and adjust the information content corresponding to the second initial information based on the information content, so that the information content corresponding to the second initial information matches the information content corresponding to the information interaction process, and obtain the final output reply information. In this way, the information content of the output reply information can be made more consistent with the information content corresponding to the information interaction process, so that after the reply information is added to the information interaction process, the consistency and fluency of the information content in the information interaction process can be maintained, and thus a more smooth, accurate and real information interaction experience can be brought to the players.

[0228] For example, when the message to be replied is "Which football stars do you like", by analyzing the historical information, it is possible to analyze which ball sport the information content discussed during the information interaction process is about, and thus the football star information corresponding to the ball sport can be replied to match the information content corresponding to the information interaction process.

[0229] In addition to adjusting information based on information style tags and historical information, computer devices can also adjust information through other information in the game. For example, for some more critical game tasks in the game, when replying to information, the information generation model can enrich the information content included in the reply information to give players more help and prompts so that players can successfully complete these game tasks; for some non-critical game tasks, the information content included in the reply information can be simplified to improve the simplicity and fluency of information interaction.

[0230] Next, we will introduce how to train an information generation model that can perform the above information generation process.

[0231] In a possible implementation, the derivation process also includes the above-mentioned second information adjustment process, which is used to adjust the information content corresponding to the information generated by the first information adjustment process according to the historical information in the information interaction process of the information to be replied to, so that the information content is more in line with the information content corresponding to the information interaction process.

[0232] The sample information to be replied is the information in the sample information interaction process, that is, the reply information generated based on the sample information to be replied is used to be added to the sample information interaction process, so the information content of the reply information needs to be consistent with the information content corresponding to the sample information interaction process. The sample intermediate information also includes the sample certificate second initial information, the sample second initial information is the information generated by adjusting the sample first initial information according to the sample information style label through the first information adjustment process, and the sample reply information is the information generated by adjusting the sample second initial information according to the historical information in the sample information interaction process through the second information adjustment process.

[0233] See also Fig.15 The initial information adjustment module in the initial information generation model may include an initial first submodule and an initial second submodule, wherein the initial first submodule is used to learn and execute the first information adjustment process, and the initial second submodule is used to learn and execute the second information adjustment process. Fig.16 When executing step S30212, the computer device may execute steps S302121-S302122, where steps S302121-S302122 are a possible implementation of step S30212, including:

[0234] S302121: Generate pending second initial information according to the pending first initial information and the style label of the sample information through the initial first submodule.

[0235] The initial first submodule can adjust the information style corresponding to the pending first initial information according to the information style corresponding to the sample information style label, so that the information style corresponding to the first initial information can match the information style corresponding to the sample information style label, so as to generate the pending second initial information.

[0236] S302122: Generate pending reply information through the initial second submodule according to historical information in the process of interaction between the pending second initial information and the sample information.

[0237] The initial second submodule can analyze the information content involved in the sample information interaction process based on the historical information in the sample information interaction process, and adjust the information content corresponding to the pending second initial information based on the information content, so that the information content corresponding to the pending second initial information can fit the information content involved in the sample information interaction process, so as to generate the final output pending reply information.

[0238] When executing step S30311, the computer device may execute steps S303111-S303112, where steps S303111-S303112 are a possible implementation of step S30311, including:

[0239] S303111: According to the difference between the pending first initial information and the sample first initial information, the model parameters corresponding to the initial information generation module are adjusted to obtain the information generation module.

[0240] This adjustment process has been introduced in detail in step S30311 and will not be repeated here.

[0241] S303112: According to the difference between the pending second initial information and the sample second initial information, the model parameters corresponding to the initial first submodule are adjusted to obtain the first submodule, and according to the difference between the pending reply information and the sample reply information, the model parameters corresponding to the initial second submodule are adjusted to obtain the second submodule.

[0242] According to the difference between the pending second initial information and the sample second initial information, the accuracy of the initial first submodule in executing the first information adjustment process can be characterized, and the difference is inversely correlated with the accuracy. Therefore, the computer device can adjust the model parameters corresponding to the initial first submodule according to the difference to narrow the difference. In this process, the initial first submodule can learn how to accurately execute the first information adjustment process, and obtain a first submodule that can be used to accurately execute the first information adjustment process.

[0243] According to the difference between the pending reply information and the sample reply information, the accuracy of the initial second submodule in executing the second information adjustment process, and the retention rationality of the information style corresponding to the pending second initial information and the information content corresponding to the pending first initial information when adjusting the information content, can be characterized, and the difference is inversely correlated with the accuracy and the retention rationality. Therefore, the computer device can adjust the model parameters corresponding to the initial second submodule according to the difference, so that the difference is continuously reduced. In this process, the initial second submodule can learn how to accurately adjust the information content of the information based on the information content characterized by the historical information, so that the information content of the information finally generated can fit the information content in the corresponding information interaction process in terms of information content, and maintain the rationality of the information interaction process; at the same time, it can learn how to maintain the information style corresponding to the information while adjusting the information content, and keep the information content corresponding to the information can be used to reply to the information to be replied, so that the second submodule for accurately executing the second information adjustment process can be obtained, and the first submodule and the second submodule are used to constitute the information adjustment module.

[0244] In addition to directly adjusting model parameters through information differences, the computer device can also measure the accuracy of the model's adjustment of information style when generating reply information by directly analyzing the information style corresponding to the reply information generated by the model.

[0245] In one possible implementation, see Fig.17 The pending reply information corresponds to the pending information style. The computer device may further execute step S1701, including:

[0246] S1701: Analyze the pending information style label corresponding to the pending reply information through the label recognition model.

[0247] The pending information style label corresponds to the pending information style. The label recognition model can be any model that can analyze the information style label based on the information. When training the model, a sample pair of information and the information style label can be constructed, and the model can be trained based on the sample pair so that the model learns the model knowledge for identifying the information style label.

[0248] When executing step S303, the computer device may execute step S3032. Step S3032 is a possible implementation of step S303, including:

[0249] S3032: According to the difference between the pending reply information and the sample reply information, and the difference between the pending information style label and the sample information style label, the model parameters corresponding to the initial information generation model are adjusted to obtain the information generation model.

[0250] The principle of adjusting based on the difference between the sample reply information and the pending reply information has been explained in the above content and will not be repeated here. Since the sample information style corresponding to the sample information style label is the information style required for this information generation, and the pending information style corresponding to the pending information style label is the actual information style corresponding to the pending reply information generated by the initial information generation model, the difference between the sample information style label and the pending information style label can characterize the accuracy of the initial information generation model in adjusting the information style corresponding to the information, and the difference is inversely correlated with the accuracy. Thus, the computer device can adjust the model parameters of the initial information generation model according to the difference to narrow the difference. In this process, the model can learn how to accurately adjust the information style corresponding to the information, and then the trained information generation model can accurately adjust the information style of the information based on the information style label.

[0251] Among them, in order to enable the information generation model to accurately adjust the information style under various information interaction requirements, the computer device can set a variety of information style labels for indicating information style for model training. For example, in combination with the above content, it can be seen that the sample information style label can include any one or more combinations of sample scene labels, sample character labels and sample context labels. The sample scene label is used to identify the game scene corresponding to the sample reply information, so as to characterize in what game scene the information reply should be made in the information style corresponding to the sample reply information, and the information style corresponding to the sample reply information is the information style corresponding to the scene label; the sample character label is used to identify the character personality of the non-player character corresponding to the sample reply information, so as to characterize in what character personality the non-player character has, the information reply should be made in the information style corresponding to the sample reply information, and the information style corresponding to the sample reply information is the information style corresponding to the sample character label; the sample context label is used to identify the context corresponding to the sample information interaction process in which the sample to be replied information is located, so as to characterize in what context the information interaction should be made in the information style corresponding to the sample reply information, and the information style corresponding to the sample reply information is the information style corresponding to the sample character label.

[0252] In addition, in order to enable the information generation model to adjust information according to diverse information styles to meet diverse information style requirements, the computer device can provide sample response information corresponding to diverse information styles for model training.

[0253] On the one hand, in a possible implementation, the sample reply information is any one of a plurality of sample reply information, and the character personalities identified by the sample personality labels corresponding to different sample reply information are different, that is, the plurality of sample reply information is reply information provided by non-player characters with different character personalities. Thus, when the model is trained based on a plurality of sample reply information, the information generation model can learn how to adjust the information style corresponding to the information for non-player characters with different character personalities, so that the information style of the generated reply information can fit various different character personalities, thereby supporting more effective and realistic information interaction with non-player characters with various different character personalities in the game.

[0254] Among them, there are multiple ways to generate sample reply information corresponding to different sample personality labels. For example, by inputting different personality information into the language model, the language model can simulate different character personalities to conduct conversations, thereby obtaining information on conversations with characters with different personalities. The computer device can extract sample reply information and sample reply information corresponding to different sample personality labels from this information.

[0255] On the other hand, in a possible implementation, the sample reply information can be any one of a plurality of sample reply information, and the sample scene labels corresponding to different sample reply information have different game scenes. Thus, by training the model through a plurality of sample reply information, the information generation model can learn how to accurately adjust the information style of the reply information for a variety of game scenes, and then in the actual application process, no matter what game scene the non-player character is in, the information generation model can generate reply information whose information style matches the game scene, so as to support more realistic and effective information interaction with non-player characters in different game scenes in the game, and improve the generalization of the information generation ability of the information generation model in multiple game scenes.

[0256] It should be emphasized that multiple sample reply messages can correspond to the same sample pending reply message, or can correspond to multiple sample pending reply messages. As long as the information interaction between the sample pending reply message and the sample reply message conforms to the information interaction method in the game, no limitation is made here.

[0257] In order to facilitate understanding of the technical solution provided by this application, the information processing method provided by this application will be introduced below in combination with an actual application scenario.

[0258] See also Fig.18 , Fig.18A flowchart of an information processing method in an actual application scenario provided by an embodiment of the present application. In the actual application scenario, the computer device may be any computer device that supports information processing, wherein steps S1801-S1806 are a model training process, and steps S1807-S1811 are a model application process. The method includes:

[0259] S1801: Construct sample pending reply information and sample reply information corresponding to a plurality of sample personality labels, sample scene labels, and sample context labels.

[0260] In order to enable the information generation model to generate reply information with a relatively matching information style in a variety of information interaction scenarios, when constructing training samples, the computer device can construct training samples corresponding to multiple information style labels. In this implementation, the information style labels can include personality labels, scene labels, and context labels.

[0261] S1802: Obtain historical information in the sample information interaction process corresponding to the sample information to be replied.

[0262] The historical information is the historical pending reply information and reply information in the sample information interaction process. The context and information content corresponding to the sample information interaction process can be represented through the historical information.

[0263] S1803: Clean and enhance the data of sample pending reply information, sample reply information and historical information.

[0264] Data cleaning is used to remove some duplicate information and noise information, and enhancement is used to further generate diversified training samples using existing training samples through techniques such as synonym replacement and syntactic transformation, thereby increasing the number of training samples while ensuring the effectiveness of training samples.

[0265] S1804: Perform semantic segmentation and label alignment on the sample response information.

[0266] In order to enable the information generation model to have a more detailed and accurate understanding of the information, the computer device can perform semantic segmentation and annotation alignment on the sample response information. Semantic segmentation is used to split the sample response information into smaller semantic units, so that the model can understand the structure and content of the sample response information in detail. Annotation alignment ensures that each semantic unit is correctly matched with the corresponding information style label, that is, the information style corresponding to each semantic unit is annotated, so that the information generation model can learn how to adjust the information style of the information more accurately and fine-grainedly.

[0267] S1805: Based on the sample pending reply information, the sample reply information and the historical information, an initial information generation model is trained to obtain an information generation model.

[0268] The initial information generation model can generate pending reply information based on sample pending reply information, sample reply information and historical information, and sample information style labels corresponding to the sample reply information. When adjusting the model parameters, the computer device can adjust the model parameters based on information style label differences, based on intermediate information differences, and based on reply information differences, so that the trained information generation model can learn how to generate reply information whose information content can be used to reply to pending reply information, whose information content matches the information interaction process corresponding to the pending reply information, and whose information style matches the information style corresponding to the information style label.

[0269] S1806: Obtain the to-be-replied information provided by the player character controlled by the player to the non-player character.

[0270] In this practical application, the information provider can be a player character controlled by the player, and the information to be replied can be information extracted from the game plot and task text, or information input by the player through the player character, etc., which is not limited here. The computer device can pre-set the information that the player can input in the game, so that the player can select and input from this part of the information, or it can support the player to freely input information to interact with non-player characters.

[0271] S1807: Determine the personality tag and scene tag corresponding to the non-player character, and determine the context tag and historical information corresponding to the information interaction process corresponding to the message to be replied.

[0272] The personality tags corresponding to the non-player characters can be preset by the computer device, or can be set in real time based on the player's preferences. The scene tags can be determined by analyzing the status of the non-player characters, or analyzing the relevant information of the player characters in the game, and the historical information can be the information generated by the information interaction between players in the historical period.

[0273] S1808: Input personality tags, scene tags, context tags, information to be replied and historical information into the information generation model to generate reply information.

[0274] For the specific generation process, see steps S18081-S18083. Steps S18081-S18083 are a possible implementation of step S1808, including:

[0275] S18081: Generate first initial information according to the information to be replied through the information generation module.

[0276] The information generation module can analyze what information content can be used to reply to the information to be replied based on the information to be replied, and obtain the first initial information that can be used to reply to the information to be replied in terms of information content.

[0277] S18082: Generate second initial information according to the first initial information, the personality label, the scene label and the context label through the first submodule.

[0278] The first submodule can adjust the information style corresponding to the first initial information according to the multiple information style tags, so that the information style can meet the information style requirements of this information interaction.

[0279] S18083: Generate reply information according to the second initial information and the historical information through the second submodule.

[0280] The second submodule can analyze the information content corresponding to the information interaction process corresponding to the information to be replied to based on historical information, so as to adjust the information content corresponding to the second initial information, so that the information content corresponding to the reply information obtained after adjustment can match the information content corresponding to the information interaction process, so that after providing reply information in the information interaction process, it can ensure that the information content corresponding to multiple information in the information interaction process is relatively consistent, giving players a more coherent and smooth information interaction experience.

[0281] S1809: Obtain the matching degree of the player's input for the reply message.

[0282] After generating and providing various reply messages, the computer device can display the reply messages to the player while displaying the information style tags used to generate the reply messages, so that the player can analyze the degree of match between the information style corresponding to the reply messages and the required information style, and then input the degree of match considered by the player to accurately measure the information generation effect of the information generation model.

[0283] S1810: Based on the matching degree being less than the matching degree threshold, the model parameters corresponding to the information generation model are adjusted according to the information to be replied, so that the matching degree between the information style corresponding to the reply information generated by the information generation model according to the information to be replied and the information style corresponding to the information style label is not less than the matching degree threshold.

[0284] If the matching degree is less than the matching degree threshold, it means that the player is not satisfied with the reply information generated based on the information generation model. At this time, the computer device can adjust the model parameters corresponding to the information generation model so that the matching degree between the information style corresponding to the reply information generated by the model based on the information to be replied and the information style corresponding to the information style label is not less than the matching degree threshold, thereby further optimizing the player's information interaction experience.

[0285] It can be seen from the above content that this application has the following technical effects:

[0286] 1. When generating reply information, the present application will provide the information to be replied to provided by the non-player character and the information style label corresponding to the non-player character to the information generation model at the same time, so that the reply information generated by the information generation model can simultaneously meet the information content and information style required for information interaction, and then when the reply information is provided to the information provider through the non-player character, it can bring a more realistic and accurate information interaction experience to the information provider. At the same time, through this method, when designing non-player characters in the game, it is only necessary to set a variety of information style labels for the non-player characters, so that the non-player characters can have information reply capabilities with diverse styles. It is also possible to set different information style labels for different game scenes and different game processes, so that the non-player characters can reply to information with matching information styles in different game scenes, different information interaction contexts and different role settings, further improving the authenticity and diversity of the player's game experience and enhancing the immersion brought by the game.

[0287] 2. This application provides a variety of information style tags, supporting the information generation model to generate style-matching reply information for various game scenarios, various non-player character personalities, and various information interaction contexts, comprehensively optimizing the player's information interaction experience in the game and improving the versatility of the information generation model.

[0288] 3. The information generation model can accurately analyze the information content corresponding to the information interaction process based on the historical information in the information interaction process corresponding to the information to be replied, such as multiple rounds of information to be replied and reply information, so that the generated reply information is more closely matched with the information content corresponding to the overall information interaction process in terms of information content. After providing the reply information, the consistency and coherence of the content of the information interaction process is maintained, thereby ensuring the player's information interaction experience in the entire information interaction process.

[0289] 4. This application divides the information generation process into multiple information generation processes, and performs more detailed model parameter adjustments in the three processes of content generation, information style adjustment, and information content adjustment, so that the information generation model can more fully learn the model knowledge used to generate reply information, thereby improving the accuracy and rationality of the reply information.

[0290] 5. During the application of the information generation model, this application can continuously detect the information style matching degree corresponding to the generated reply information to measure the information generation effect of the model, and continuously optimize the information generation effect to ensure the player's information interaction experience.

[0291] Based on the information processing method on the application side provided in the above embodiment, the present application also provides an information processing device, see Fig.19 , Fig.19This is a structural block diagram of an information processing device provided in an embodiment of the present application. The device 1900 includes a first acquisition unit 1901, a first determination unit 1902, a first generation unit 1903 and a providing unit 1904:

[0292] The first acquisition unit 1901 is used to acquire the to-be-replied information provided by the information provider for the non-player character in the game;

[0293] The first determining unit 1902 is used to determine an information style tag corresponding to the non-player character, wherein the information style tag is used to indicate an information style of a reply message provided by the non-player character in response to the message to be replied;

[0294] The first generating unit 1903 is used to generate reply information corresponding to the information to be replied according to the information to be replied and the information style label through an information generation model, wherein the reply information is used to reply to the information to be replied, and the information style of the reply information matches the information style corresponding to the information style label;

[0295] The providing unit 1904 is configured to provide the reply information to the information provider through the non-player character.

[0296] In a possible implementation manner, the first determining unit 1902 is specifically configured to:

[0297] Determining a game scene in which the non-player character is located;

[0298] Obtaining a scene tag for identifying the game scene;

[0299] The scene tag is determined as an information style tag corresponding to the non-player character.

[0300] In a possible implementation, the information style tag includes a personality tag corresponding to the non-player character, and the personality tag is used to identify the character personality of the non-player character.

[0301] In a possible implementation manner, the determining unit is specifically configured to:

[0302] Acquire historical information provided by the information provider before the information provider provides the information to be replied during the information interaction between the information provider and the non-player character;

[0303] Determining, according to the historical information, a context label corresponding to the information interaction process, wherein the context label is used to identify a context corresponding to the information interaction process;

[0304] The context tag is determined as an information style tag corresponding to the non-player character.

[0305] In a possible implementation manner, the device further includes a second determining unit and a first adjusting unit:

[0306] The second determining unit is used to determine the matching degree between the information style corresponding to the reply information and the information style corresponding to the information style tag;

[0307] The first adjustment unit is used to adjust the model parameters corresponding to the information generation model according to the information to be replied based on the fact that the matching degree is less than a matching degree threshold, so that the matching degree between the information style corresponding to the reply information generated by the information generation model according to the information to be replied and the information style corresponding to the information style label is not less than the matching degree threshold.

[0308] In a possible implementation, the information provider is a player character controlled by a player, the information style tag is used to display to the player, and the second determining unit is specifically used to:

[0309] Obtaining the matching degree of the player input for the reply information;

[0310] The matching degree of the player's input for the reply information is determined as the matching degree between the information style corresponding to the reply information and the information style label.

[0311] In a possible implementation, the information generation model includes an information generation module and an information adjustment module, the information generation module is used to execute an information generation process, the information adjustment module is used to execute a first information adjustment process, the information generation process is used to generate information to reply to the information to be replied, the first information adjustment process is used to adjust the information style corresponding to the information generated by the information generation process according to the information style label, and the first generation unit 1903 is specifically used to:

[0312] Generate, by the information generating module, first initial information according to the information to be replied, wherein the first initial information is used to reply to the information to be replied;

[0313] The reply information is generated according to the first initial information and the information style tag through the information adjustment module.

[0314] In a possible implementation, the information adjustment module includes a first submodule and a second submodule, the first submodule is used to execute the first information adjustment process, the second submodule is used to execute the second information adjustment process, the second information adjustment process is used to adjust the information content corresponding to the information generated by the first information adjustment process according to the historical information in the information interaction process of the to-be-replied information, and the device also includes a second acquisition unit:

[0315] The second acquisition unit is used to acquire the historical information provided by the information provider before the information provider provides the information to be replied during the information interaction between the information provider and the non-player character.

[0316] The first generating unit 1903 is specifically used for:

[0317] Generate second initial information according to the first initial information and the information style label through the first submodule;

[0318] The reply information is generated according to the historical information and the second initial information through the second submodule.

[0319] Based on the information processing method on the training side provided in the above embodiment, the present application also provides an information processing device, see Fig. 20 , Fig. 20 This is a structural block diagram of an information processing device provided in an embodiment of the present application. The device 2000 includes a third acquisition unit 2001, a second generation unit 2002, and a second adjustment unit 2003:

[0320] The third acquisition unit 2001 is used to acquire sample information to be replied and sample reply information, wherein the sample reply information is used to reply to the sample information to be replied, the sample information to be replied has a corresponding sample information style label, the sample reply information corresponds to the sample information style, and the sample information style label corresponds to the sample information style;

[0321] The second generating unit 2002 is used to generate pending reply information according to the sample to-be-replied information and the sample information style label by using the initial information generating model;

[0322] The second adjustment unit 2003 is used to adjust the model parameters corresponding to the initial information generation model according to the difference between the pending reply information and the sample reply information to obtain the information generation model, wherein the information generation model is used to generate reply information for replying to the pending reply information provided by the information provider for the non-player character in the game and the information style tag corresponding to the non-player character, wherein the information style corresponding to the reply information matches the information style corresponding to the corresponding information style tag, and the reply information is used to be provided to the information provider through the non-player character.

[0323] In a possible implementation, the sample reply information has corresponding sample intermediate information, and the sample intermediate information is used to characterize a derivation process of generating the sample reply information with respect to the sample information to be replied, and the second generating unit 2002 is specifically used to:

[0324] Generate the pending reply information and the pending intermediate information corresponding to the pending reply information according to the sample pending reply information and the sample information style label through the initial information generation model, wherein the pending intermediate information is used to characterize the derivation process of the initial information generation model to generate the pending reply information;

[0325] The second adjusting unit 2003 is specifically used for:

[0326] According to the difference between the pending reply information and the sample reply information, and the difference between the sample intermediate information and the pending intermediate information, the model parameters corresponding to the initial information generation model are adjusted to obtain the information generation model.

[0327] In a possible implementation, the derivation process includes an information generation process and a first information adjustment process, the information generation process is used to generate information for replying to information to be replied, the first information adjustment process is used to adjust the information style corresponding to the information generated by the information generation process according to the information style label, the sample intermediate information includes sample first initial information, the sample first initial information is information generated by the information generation process, the initial information generation model includes an initial information generation module and an initial information adjustment module, and the second generation unit 2002 is specifically used to:

[0328] Generate pending first initial information according to the sample information to be replied by the initial information generation module, the pending first initial information is used to reply the sample information to be replied, and the pending intermediate information includes the pending first initial information;

[0329] Generate the pending reply information according to the pending first initial information and the sample information style label through the initial information adjustment module;

[0330] The second adjusting unit 2003 is specifically used for:

[0331] According to the difference between the pending first initial information and the sample first initial information, the model parameters corresponding to the initial information generation module are adjusted to obtain the information generation module, and according to the difference between the pending reply information and the sample reply information, the model parameters corresponding to the initial information adjustment module are adjusted to obtain the information adjustment module. The information generation module and the information adjustment module are used to constitute the information generation model.

[0332] In a possible implementation, the derivation process further includes a second information adjustment process, the second information adjustment process is used to adjust information content corresponding to the information generated by the first information adjustment process according to historical information in the information interaction process in which the information to be replied is located, the sample information to be replied is information in the sample information interaction process, the sample intermediate information also includes sample certificate second initial information, the sample second initial information is information generated by adjusting the sample first initial information according to the sample information style label through the first information adjustment process, and the sample reply information is information generated by adjusting the sample second initial information according to historical information in the sample information interaction process through the second information adjustment process;

[0333] The initial information adjustment module includes an initial first submodule and an initial second submodule, and the second generation unit 2002 is specifically used for:

[0334] Generate the pending second initial information according to the pending first initial information and the sample information style label through the initial first submodule;

[0335] Generate the pending reply information through the initial second submodule according to the historical information in the process of interaction between the pending second initial information and the sample information;

[0336] The second adjusting unit 2003 is specifically used for:

[0337] According to the difference between the pending first initial information and the sample first initial information, adjusting the model parameters corresponding to the initial information generation module to obtain the information generation module;

[0338] According to the difference between the pending second initial information and the sample second initial information, the model parameters corresponding to the initial first submodule are adjusted to obtain the first submodule, and according to the difference between the pending reply information and the sample reply information, the model parameters corresponding to the initial second submodule are adjusted to obtain the second submodule. The first submodule and the second submodule are used to constitute the information adjustment module.

[0339] In a possible implementation, the pending reply information corresponds to a pending information style, and the device further includes an analyzing unit:

[0340] The analyzing unit is used to analyze the pending information style label corresponding to the pending reply information through a label recognition model, wherein the pending information style label corresponds to the pending information style;

[0341] The second adjusting unit 2003 is specifically used for:

[0342] According to the difference between the pending reply information and the sample reply information, and the difference between the pending information style label and the sample information style label, the model parameters corresponding to the initial information generation model are adjusted to obtain the information generation model.

[0343] In one possible implementation, the sample information style label includes any one or more combinations of a sample scene label, a sample character label, and a sample context label. The sample scene label is used to identify the game scene corresponding to the sample reply information, the sample character label is used to identify the character character of the non-player character corresponding to the sample reply information, and the sample context label is used to identify the context corresponding to the sample information interaction process in which the sample information to be replied is located.

[0344] In a possible implementation, the sample reply information is any one of a plurality of sample reply information, and the character personalities identified by the sample personality tags corresponding to different sample reply information are different.

[0345] In a possible implementation, the sample reply information is any one of a plurality of sample reply information, and the game scenes identified by the sample scene tags corresponding to different sample reply information are different.

[0346] The present application also provides a computer device, see Fig.21 As shown, the computer device may be a terminal device, and a mobile phone is taken as an example:

[0347] Fig.21 FIG. 1 is a block diagram showing a partial structure of a mobile phone related to a terminal device provided in an embodiment of the present application. Fig.21 The mobile phone includes: a radio frequency (RF) circuit 710, a memory 720, an input unit 730, a display unit 740, a sensor 750, an audio circuit 760, a wireless fidelity (WiFi) module 770, a processor 780, and a power supply 790. Those skilled in the art will understand that Fig.21 The mobile phone structure shown in the figure does not constitute a limitation on the mobile phone, and may include more or fewer components than shown in the figure, or combine certain components, or arrange the components differently.

[0348] Combine the following Fig.21 A detailed introduction to the various components of the mobile phone:

[0349] The RF circuit 710 can be used for receiving and sending signals during the process of sending and receiving information or making calls. In particular, after receiving the downlink information of the base station, it is sent to the processor 780 for processing; in addition, the designed uplink data is sent to the base station. Usually, the RF circuit 710 includes but is not limited to an antenna, at least one amplifier, a transceiver, a coupler, a low noise amplifier (Low Noise Amplifier, referred to as LNA), a duplexer, etc. In addition, the RF circuit 710 can also communicate with the network and other devices through wireless communication. The above-mentioned wireless communication can use any communication standard or protocol, including but not limited to the Global System of Mobile communication (GSM), General Packet Radio Service (GPRS), Code Division Multiple Access (CDMA), Wideband Code Division Multiple Access (WCDMA), Long Term Evolution (LTE), email, Short Messaging Service (SMS), etc.

[0350] The memory 720 can be used to store software programs and modules. The processor 780 executes various functional applications and data processing of the mobile phone by running the software programs and modules stored in the memory 720. The memory 720 can mainly include a program storage area and a data storage area, wherein the program storage area can store an operating system, an application required for at least one function (such as a sound playback function, an image playback function, etc.), etc.; the data storage area can store data created according to the use of the mobile phone (such as audio data, a phone book, etc.), etc. In addition, the memory 720 can include a high-speed random access memory, and can also include a non-volatile memory, such as at least one disk storage device, a flash memory device, or other volatile solid-state storage devices.

[0351] The input unit 730 can be used to receive input digital or character information, and to generate key signal input related to the user settings and function control of the mobile phone. Specifically, the input unit 730 may include a touch panel 731 and other input devices 732. The touch panel 731, also known as a touch screen, can collect the user's touch operation on or near it (such as the user's operation on the touch panel 731 or near the touch panel 731 using any suitable object or accessory such as a finger, stylus, etc.), and drive the corresponding connection device according to a pre-set program. Optionally, the touch panel 731 may include two parts: a touch detection device and a touch controller. Among them, the touch detection device detects the user's touch orientation, detects the signal brought by the touch operation, and transmits the signal to the touch controller; the touch controller receives the touch information from the touch detection device, converts it into contact coordinates, and then sends it to the processor 780, and can receive and execute commands sent by the processor 780. In addition, the touch panel 731 can be implemented in various types such as resistive, capacitive, infrared, and surface acoustic waves. In addition to the touch panel 731, the input unit 730 may further include other input devices 732. Specifically, the other input devices 732 may include but are not limited to one or more of a physical keyboard, function keys (such as volume control keys, switch keys, etc.), a trackball, a mouse, a joystick, and the like.

[0352] The display unit 740 can be used to display information input by the user or information provided to the user and various menus of the mobile phone. The display unit 740 may include a display panel 741. Optionally, the display panel 741 may be configured in the form of a liquid crystal display (Liquid Crystal Display, LCD), an organic light-emitting diode (Organic Light-Emitting Diode, OLED), etc. Further, the touch panel 731 may cover the display panel 741. When the touch panel 731 detects a touch operation on or near it, it is transmitted to the processor 780 to determine the type of touch event. Subsequently, the processor 780 provides a corresponding visual output on the display panel 741 according to the type of touch event. Although in Fig.21 In the embodiment, the touch panel 731 and the display panel 741 are used as two independent components to realize the input and output functions of the mobile phone, but in some embodiments, the touch panel 731 and the display panel 741 can be integrated to realize the input and output functions of the mobile phone.

[0353] The mobile phone may also include at least one sensor 750, such as a light sensor, a motion sensor, and other sensors. Specifically, the light sensor may include an ambient light sensor and a proximity sensor, wherein the ambient light sensor may adjust the brightness of the display panel 741 according to the brightness of the ambient light, and the proximity sensor may turn off the display panel 741 and / or the backlight when the mobile phone is moved to the ear. As a type of motion sensor, the accelerometer sensor can detect the magnitude of acceleration in all directions (generally three axes), and can detect the magnitude and direction of gravity when stationary. It can be used for applications that identify the posture of the mobile phone (such as horizontal and vertical screen switching, related games, magnetometer posture calibration), vibration recognition related functions (such as pedometer, tapping), etc.; as for other sensors that can be configured in the mobile phone, such as gyroscopes, barometers, hygrometers, thermometers, infrared sensors, etc., they will not be repeated here.

[0354] The audio circuit 760, the speaker 761, and the microphone 762 can provide an audio interface between the user and the mobile phone. The audio circuit 760 can transmit the received audio data to the speaker 761 after converting the received audio data into an electrical signal, which is converted into a sound signal for output; on the other hand, the microphone 762 converts the collected sound signal into an electrical signal, which is received by the audio circuit 760 and converted into audio data, and then the audio data is output to the processor 780 for processing, and then sent to another mobile phone through the RF circuit 710, or the audio data is output to the memory 720 for further processing.

[0355] WiFi is a short-range wireless transmission technology. The mobile phone can help users send and receive emails, browse web pages and access streaming media through the WiFi module 770. It provides users with wireless broadband Internet access. Fig.21 A WiFi module 770 is shown, but it is understandable that it is not an essential component of the mobile phone and can be omitted as needed without changing the essence of the invention.

[0356] The processor 780 is the control center of the mobile phone. It uses various interfaces and lines to connect various parts of the entire mobile phone. By running or executing software programs and / or modules stored in the memory 720, and calling data stored in the memory 720, it executes various functions of the mobile phone and processes data, thereby performing overall detection of the mobile phone. Optionally, the processor 780 may include one or more processing units; preferably, the processor 780 may integrate an application processor and a modem processor, wherein the application processor mainly processes the operating system, user interface, and application programs, and the modem processor mainly processes wireless communications. It is understandable that the above-mentioned modem processor may not be integrated into the processor 780.

[0357] The mobile phone also includes a power supply 790 (such as a battery) for supplying power to various components. Preferably, the power supply can be logically connected to the processor 780 through a power management system, so that the power management system can manage charging, discharging, power consumption and other functions.

[0358] Although not shown, the mobile phone may also include a camera, a Bluetooth module, etc., which will not be described in detail here.

[0359] In this embodiment, the processor 780 included in the terminal device also has the following functions:

[0360] Obtain the pending reply information provided by the information provider for the non-player characters in the game;

[0361] Determining an information style tag corresponding to the non-player character, the information style tag being used to indicate an information style of reply information provided by the non-player character in response to the information to be replied;

[0362] Generate reply information corresponding to the information to be replied according to the information to be replied and the information style label through the information generation model, wherein the reply information is used to reply to the information to be replied, and the information style of the reply information matches the information style corresponding to the information style label;

[0363] The reply information is provided to the information provider through the non-player character.

[0364] Alternatively, in this embodiment, the processor 780 included in the terminal device further has the following functions:

[0365] Acquire sample information to be replied and sample reply information, wherein the sample reply information is used to reply the sample information to be replied, the sample information to be replied has a corresponding sample information style label, the sample reply information corresponds to the sample information style, and the sample information style label corresponds to the sample information style;

[0366] Generate pending reply information according to the sample pending reply information and the sample information style label through the initial information generation model;

[0367] According to the difference between the pending reply information and the sample reply information, the model parameters corresponding to the initial information generation model are adjusted to obtain an information generation model, wherein the information generation model is used to generate reply information for replying to the pending reply information provided by an information provider for a non-player character in the game and an information style tag corresponding to the non-player character, wherein the information style corresponding to the reply information matches the information style corresponding to the corresponding information style tag, and the reply information is used to be provided to the information provider through the non-player character.

[0368] The present application also provides a server. Fig. 22 As shown, Fig. 22 The structural diagram of the server 800 provided in the embodiment of the present application, the server 800 may have relatively large differences due to different configurations or performances, and may include one or more central processing units (CPU) 822 (for example, one or more processors) and a memory 832, and one or more storage media 830 (for example, one or more mass storage devices) storing application programs 842 or data 844. Among them, the memory 832 and the storage medium 830 can be temporary storage or permanent storage. The program stored in the storage medium 830 may include one or more modules (not shown in the figure), and each module may include a series of instruction operations on the server. Furthermore, the central processing unit 822 can be configured to communicate with the storage medium 830 and execute a series of instruction operations in the storage medium 830 on the server 800.

[0369] The server 800 may also include one or more power supplies 826, one or more wired or wireless network interfaces 850, one or more input and output interfaces 858, and / or one or more operating systems 841, such as Windows Server 2000. TM , Mac OS X TM , Unix TM ,Linux TM , FreeBSD TM etc.

[0370] The steps performed by the server in the above embodiment can be based on Fig. 22 The server structure shown.

[0371] An embodiment of the present application also provides a computer-readable storage medium for storing a computer program, wherein the computer program is used to execute any one of the information processing methods described in the aforementioned embodiments.

[0372] An embodiment of the present application further provides a computer program product including a computer program, which, when executed on a computer device, enables the computer device to execute the information processing method described in any one of the above embodiments.

[0373] It is understandable that in the specific implementation of the present application, related data such as user information (such as information to be replied to, reply information) is involved. When the above embodiments of the present application are applied to specific products or technologies, user permission or consent is required, and the collection, use and processing of relevant data need to comply with relevant laws, regulations and standards of relevant countries and regions.

[0374] A person of ordinary skill in the art can understand that all or part of the steps of implementing the above-mentioned method embodiment can be completed by hardware related to program instructions, and the above-mentioned program can be stored in a computer-readable storage medium. When the program is executed, it executes the steps of the above-mentioned method embodiment; and the above-mentioned storage medium can be at least one of the following media: read-only memory (English: read-only memory, abbreviated: ROM), RAM, magnetic disk or optical disk, etc. Various media that can store program codes.

[0375] It should be noted that each embodiment in this specification is described in a progressive manner, and the same and similar parts between the embodiments can refer to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the device and system embodiments, since they are basically similar to the method embodiments, the description is relatively simple, and the relevant parts can refer to the partial description of the method embodiments. The device and system embodiments described above are merely schematic, in which the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative work.

[0376] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any changes or substitutions that can be easily thought of by a person skilled in the art within the technical scope disclosed in the present application should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims

1. An information processing method, characterized in that: The method comprises: Obtain the pending reply information provided by the information provider for the non-player characters in the game; Determining an information style tag corresponding to the non-player character, the information style tag being used to indicate an information style of reply information provided by the non-player character in response to the information to be replied; Generate reply information corresponding to the information to be replied according to the information to be replied and the information style label through the information generation model, wherein the reply information is used to reply to the information to be replied, and the information style of the reply information matches the information style corresponding to the information style label; The reply information is provided to the information provider through the non-player character.

2. The method according to claim 1, characterized in that The determining the information style tag corresponding to the non-player character includes: Determining a game scene in which the non-player character is located; Obtaining a scene tag for identifying the game scene; The scene tag is determined as an information style tag corresponding to the non-player character.

3. The method according to claim 1, characterized in that The information style tag includes a personality tag corresponding to the non-player character, and the personality tag is used to identify the character personality of the non-player character.

4. The method according to claim 1, characterized in that: The determining the information style tag corresponding to the non-player character includes: Acquire historical information provided by the information provider before the information provider provides the information to be replied during the information interaction between the information provider and the non-player character; Determining, according to the historical information, a context label corresponding to the information interaction process, wherein the context label is used to identify a context corresponding to the information interaction process; The context tag is determined as an information style tag corresponding to the non-player character.

5. The method according to claim 1, characterized in that The method further comprises: Determining a degree of match between the information style corresponding to the reply information and the information style corresponding to the information style tag; Based on the fact that the matching degree is less than the matching degree threshold, the model parameters corresponding to the information generation model are adjusted according to the information to be replied, so that the matching degree between the information style corresponding to the reply information generated by the information generation model according to the information to be replied and the information style corresponding to the information style label is not less than the matching degree threshold.

6. The method according to claim 5, characterized in that The information provider is a player character controlled by a player, the information style tag is used to display to the player, and the determining of the matching degree between the information style corresponding to the reply information and the information style corresponding to the information style tag includes: Obtaining the matching degree of the player input for the reply information; The matching degree of the player's input for the reply information is determined as the matching degree between the information style corresponding to the reply information and the information style label.

7. The method according to claim 1, characterized in that The information generation model includes an information generation module and an information adjustment module, wherein the information generation module is used to execute an information generation process, and the information adjustment module is used to execute a first information adjustment process, wherein the information generation process is used to generate information to reply to the information to be replied, and the first information adjustment process is used to adjust the information style corresponding to the information generated by the information generation process according to the information style label, and wherein the information generation model generates reply information corresponding to the information to be replied according to the information to be replied and the information style label, including: Generate a first initial message according to the information to be replied by the information generation module, wherein the first initial message is used to reply to the information to be replied; The reply information is generated according to the first initial information and the information style tag through the information adjustment module.

8. The method according to claim 7, characterized in that The information adjustment module includes a first submodule and a second submodule, the first submodule is used to execute the first information adjustment process, the second submodule is used to execute the second information adjustment process, the second information adjustment process is used to adjust the information content corresponding to the information generated by the first information adjustment process according to the historical information in the information interaction process of the to-be-replied information, the method also includes: Acquire the historical information provided by the information provider before the information provider provides the information to be replied during the information interaction between the information provider and the non-player character; The step of generating the reply information according to the first initial information and the information style tag by the information adjustment module includes: Generate second initial information according to the first initial information and the information style label through the first submodule; The reply information is generated according to the historical information and the second initial information through the second submodule.

9. An information processing method, characterized in that: The method comprises: Acquire sample information to be replied and sample reply information, wherein the sample reply information is used to reply the sample information to be replied, the sample information to be replied has a corresponding sample information style label, the sample reply information corresponds to the sample information style, and the sample information style label corresponds to the sample information style; Generate pending reply information according to the sample pending reply information and the sample information style label through the initial information generation model; According to the difference between the pending reply information and the sample reply information, the model parameters corresponding to the initial information generation model are adjusted to obtain an information generation model, wherein the information generation model is used to generate reply information for replying to the pending reply information provided by an information provider for a non-player character in the game and an information style tag corresponding to the non-player character, wherein the information style corresponding to the reply information matches the information style corresponding to the corresponding information style tag, and the reply information is used to be provided to the information provider through the non-player character.

10. The method according to claim 9, characterized in that The sample reply information has corresponding sample intermediate information, and the sample intermediate information is used to characterize the derivation process of generating the sample reply information for the sample information to be replied. The generating model through the initial information generates the pending reply information according to the sample information to be replied and the sample information style label, including: Generate the pending reply information and the pending intermediate information corresponding to the pending reply information according to the sample pending reply information and the sample information style label through the initial information generation model, wherein the pending intermediate information is used to characterize the derivation process of the initial information generation model to generate the pending reply information; The step of adjusting the model parameters corresponding to the initial information generation model according to the difference between the pending reply information and the sample reply information to obtain the information generation model comprises: According to the difference between the pending reply information and the sample reply information, and the difference between the sample intermediate information and the pending intermediate information, the model parameters corresponding to the initial information generation model are adjusted to obtain the information generation model.

11. An information processing device, characterized in that: The device comprises a first acquisition unit, a first determination unit, a first generation unit and a providing unit: The first acquisition unit is used to acquire the to-be-replied information provided by the information provider for the non-player character in the game; The first determining unit is used to determine an information style tag corresponding to the non-player character, wherein the information style tag is used to indicate an information style of reply information provided by the non-player character in response to the information to be replied; The first generating unit is used to generate reply information corresponding to the information to be replied according to the information to be replied and the information style label through an information generation model, wherein the reply information is used to reply to the information to be replied, and the information style of the reply information matches the information style corresponding to the information style label; The providing unit is used to provide the reply information to the information provider through the non-player character.

12. An information processing device, characterized in that: The device comprises a third acquisition unit, a second generation unit and a second adjustment unit: The third acquisition unit is used to acquire sample information to be replied and sample reply information, wherein the sample reply information is used to reply to the sample information to be replied, the sample information to be replied has a corresponding sample information style label, the sample reply information corresponds to the sample information style, and the sample information style label corresponds to the sample information style; The second generating unit is used to generate pending reply information according to the sample to-be-replied information and the sample information style label by using the initial information generating model; The second adjustment unit is used to adjust the model parameters corresponding to the initial information generation model according to the difference between the pending reply information and the sample reply information to obtain the information generation model, wherein the information generation model is used to generate reply information for replying to the pending reply information provided by the information provider for the non-player character in the game and the information style tag corresponding to the non-player character, wherein the information style corresponding to the reply information matches the information style corresponding to the corresponding information style tag, and the reply information is used to be provided to the information provider through the non-player character.

13. A computer device, characterized in that: The computer device comprises a processor and a memory: The memory is used to store a computer program and transmit the computer program to the processor; The processor is used to execute the information processing method described in any one of claims 1 to 8 according to the instructions in the computer program, or to execute the information processing method described in any one of claims 9 to 10.

14. A computer-readable storage medium, characterized in that: The computer-readable storage medium is used to store a computer program, and the computer program is used to execute the information processing method described in any one of claims 1 to 8, or execute the information processing method described in any one of claims 9 to 10.

15. A computer program product comprising a computer program, which, when executed on a computer device, enables the computer device to execute the information processing method according to any one of claims 1 to 8, or the information processing method according to any one of claims 9 to 10.

Images